diff --git a/main/README.html b/main/README.html index a5e3a554b..c4513bd20 100644 --- a/main/README.html +++ b/main/README.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

NVIDIA Merlin

@@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -583,7 +582,7 @@

    Installation

    Example Notebooks and Tutorials#

    -

    A collection of end-to-end examples are available in the form of Jupyter notebooks. +

    A collection of end-to-end examples are available in the form of Jupyter notebooks. The example notebooks demonstrate how to:

    @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -483,7 +482,7 @@

    OverviewMoving Beyond Recommender Models talk at KDD’21 and read more in this blog post.

    +

    To learn more about the four-stage recommender systems, you can listen to Even Oldridge’s Moving Beyond Recommender Models talk at KDD’21 and read more in this blog post.

    Learning objectives#

    @@ -1059,7 +1058,7 @@

    Training a Ranking Model with DLRM(DLRM) architecture is a popular neural network model originally proposed by Facebook in 2019. The model was introduced as a personalization deep learning model that uses embeddings to process sparse features that represent categorical data and a multilayer perceptron (MLP) to process dense features, then interacts these features explicitly using the statistical techniques proposed in here. To learn more about DLRM architetcture please visit Exploring-different-models notebook in the Merlin Models GH repo.

    +

    Deep Learning Recommendation Model (DLRM) architecture is a popular neural network model originally proposed by Facebook in 2019. The model was introduced as a personalization deep learning model that uses embeddings to process sparse features that represent categorical data and a multilayer perceptron (MLP) to process dense features, then interacts these features explicitly using the statistical techniques proposed in here. To learn more about DLRM architetcture please visit Exploring-different-models notebook in the Merlin Models GH repo.

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -1097,8 +1096,8 @@

    Retrieving Recommendations from Triton - + +
    diff --git a/main/examples/Building-and-deploying-multi-stage-RecSys/index.html b/main/examples/Building-and-deploying-multi-stage-RecSys/index.html index 5918012f6..ac58d846f 100644 --- a/main/examples/Building-and-deploying-multi-stage-RecSys/index.html +++ b/main/examples/Building-and-deploying-multi-stage-RecSys/index.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -437,14 +436,14 @@

    Deploying a Multi-Stage Recommender System -
  • Building the Recommender System:

    +
  • Building the Recommender System:

    • Execute the preprocessing and feature engineering pipeline (ETL) with NVTabular on the GPU/CPU.

    • Train a ranking and retrieval model with TensorFlow based on the ETL output.

    • Export the saved models, user and item features, and item embeddings.

  • -
  • Deploying the Recommender System with Triton:

    +
  • Deploying the Recommender System with Triton:

  • @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -1416,8 +1415,8 @@

    Summary - + +
    diff --git a/main/examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction.html b/main/examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction.html index 3f2e2b477..a64fdaf1f 100644 --- a/main/examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction.html +++ b/main/examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -147,6 +147,7 @@ +

    NVIDIA Merlin

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -660,7 +659,7 @@

    Preprocessing with NVTabularfit_transform

  • transform

  • -

    Read more about the NVTabular Operators

    +

    Read more about the NVTabular Operators

    @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -835,8 +834,8 @@

    Splitting into train and validation dataset - + +
    diff --git a/main/examples/getting-started-movielens/02-ETL-with-NVTabular.html b/main/examples/getting-started-movielens/02-ETL-with-NVTabular.html index 780ad4baa..d255abc4a 100644 --- a/main/examples/getting-started-movielens/02-ETL-with-NVTabular.html +++ b/main/examples/getting-started-movielens/02-ETL-with-NVTabular.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -1004,7 +1003,7 @@

    Next StepsDataset

    The next step for learning to use Merlin for creating a recommender system is to train a model. -Refer to Training with TensorFlow, Training with HugeCTR, or Training with PyTorch.

    +Refer to Training with TensorFlow, Training with HugeCTR, or Training with PyTorch.

    @@ -1098,8 +1097,8 @@

    Next Steps - + +
    diff --git a/main/examples/getting-started-movielens/03-Training-with-HugeCTR.html b/main/examples/getting-started-movielens/03-Training-with-HugeCTR.html index f83a78c94..22c84bd9c 100644 --- a/main/examples/getting-started-movielens/03-Training-with-HugeCTR.html +++ b/main/examples/getting-started-movielens/03-Training-with-HugeCTR.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -815,8 +814,8 @@

    Let’s define our model - - + +
    diff --git a/main/examples/getting-started-movielens/03-Training-with-PyTorch.html b/main/examples/getting-started-movielens/03-Training-with-PyTorch.html index 67ea7db94..607461783 100644 --- a/main/examples/getting-started-movielens/03-Training-with-PyTorch.html +++ b/main/examples/getting-started-movielens/03-Training-with-PyTorch.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -53,7 +53,7 @@ - + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -779,11 +778,11 @@

    Defining Neural Network Architecture

    next

    -

    Serve Recommendations from the HugeCTR Model

    +

    Serve Recommendations from the TensorFlow Model

    @@ -844,8 +843,8 @@

    Defining Neural Network Architecture - + +
    diff --git a/main/examples/getting-started-movielens/03-Training-with-TF.html b/main/examples/getting-started-movielens/03-Training-with-TF.html index a76cc251d..fa4ad8122 100644 --- a/main/examples/getting-started-movielens/03-Training-with-TF.html +++ b/main/examples/getting-started-movielens/03-Training-with-TF.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -830,8 +829,8 @@

    Saving the model for inference - + +
    diff --git a/main/examples/getting-started-movielens/04-Triton-Inference-with-TF.html b/main/examples/getting-started-movielens/04-Triton-Inference-with-TF.html index 912e33466..e4fe18305 100644 --- a/main/examples/getting-started-movielens/04-Triton-Inference-with-TF.html +++ b/main/examples/getting-started-movielens/04-Triton-Inference-with-TF.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -54,7 +54,7 @@ - + - + +
    diff --git a/main/examples/getting-started-movielens/index.html b/main/examples/getting-started-movielens/index.html index 76be64ff0..480a5965d 100644 --- a/main/examples/getting-started-movielens/index.html +++ b/main/examples/getting-started-movielens/index.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -432,17 +431,15 @@

    Getting Started with Merlin and the MovieLens Dataset

    Use the Merlin Dataloader with PyTorch.

  • Train a HugeCTR model.

  • Serve recommendations from the Tensorflow model with the Triton Inference Server.

  • -
  • Serve recommendations from the HugeCTR model with the Triton Inference Server.

  • Explore the following notebooks:

    @@ -514,8 +511,8 @@

    Getting Started with Merlin and the MovieLens Dataset - - + +
    diff --git a/main/examples/index.html b/main/examples/index.html index 2ff94cc8a..cf5045c1f 100644 --- a/main/examples/index.html +++ b/main/examples/index.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -446,7 +445,7 @@

    NVIDIA Merlin Example Notebooks

    Inventory#

    -

    Building and Deploying a multi-stage RecSys#

    +

    Building and Deploying a multi-stage RecSys#

    Recommender system pipelines are often based on multiple stages: Retrieval, Filtering, Scoring and Ordering. This example provides an end-to-end pipeline that leverages the Merlin framework:

    -

    Scaling Large Datasets with Criteo#

    +

    Scaling Large Datasets with Criteo#

    Criteo provides the largest publicly available dataset for recommender systems with a size of 1TB of uncompressed click logs that contain 4 billion examples.

    These notebooks demonstrate how to scale NVTabular as well as the following:

    • Use multiple GPUs and nodes with NVTabular for feature engineering.

    • Train recommender system models with the Merlin Models for TensorFlow.

    • Train recommender system models with HugeCTR using multiple GPUs.

    • -
    • Inference with the Triton Inference Server and Merlin Models for TensorFlow or HugeCTR.

    • +
    • Inference with the Triton Inference Server and Merlin Models for TensorFlow.

    -

    Training and Serving with Merlin on AWS SageMaker#

    +

    Training and Serving with Merlin on AWS SageMaker#

    The notebook and scripts demonstrate how to use Merlin components like NVTabular, Merlin Models, and Merlin Systems with Triton Inference Server to build and deploy a sample end-to-end recommender system in AWS SageMaker.

    @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -428,7 +427,7 @@

    Quick-start for MerlinWe provide examples on the usage of those scripts with public datasets, but they are meant to be used with your own dataset without much effort.

    Here are the quick-start guides currently available:

    We are going to work on next releases to add:

    @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -508,7 +507,7 @@

    Preprocessingpreprocessing.py. That script is based on dask_cudf and NVTabular libraries that leverage GPUs for accelerated and distributed preprocessing.
    P.s. NVTabular also supports CPU which is suitable for prototyping in dev environments.

    The preprocessing script outputs preprocessed data as a number of parquet files, as well as a schema that stores output features metadata like statistics and tags.

    -

    In this example, we set some options for preprocessing. Here is the explanation of the main arguments; you can check the full documentation and best practices for preprocessing.

    +

    In this example, we set some options for preprocessing. Here is the explanation of the main arguments; you can check the full documentation and best practices for preprocessing.

    Training a ranking model with multi-task learning#

    @@ -561,17 +560,17 @@

    Training an MMOE modelCUDA_VISIBLE_DEVICES=0 TF_GPU_ALLOCATOR=cuda_malloc_async python -m quick_start.scripts.ranking.ranking --train_data_path $OUT_DATASET_PATH/train --eval_data_path $OUT_DATASET_PATH/eval --output_path ./outputs/ --tasks=click,like,follow,share --model mmoe --mmoe_num_mlp_experts 3 --expert_mlp_layers 128 --gate_dim 32 --use_task_towers=True --tower_layers 64 --embedding_sizes_multiplier 4 --l2_reg 1e-5 --embeddings_l2_reg 1e-8 --dropout 0.05 --lr 1e-3 --lr_decay_rate 0.99 --lr_decay_steps 100 --train_batch_size 65536 --eval_batch_size 65536 --epochs 2 --mtl_pos_class_weight_click=1 --mtl_pos_class_weight_like=2 --mtl_pos_class_weight_share=3 --mtl_pos_class_weight_follow=4 --mtl_loss_weight_click=3 --mtl_loss_weight_like=3 --mtl_loss_weight_follow=1 --mtl_loss_weight_share=1 -

    You can find more quick-start information on multi-task learning and MMOE architecture here.

    +

    You can find more quick-start information on multi-task learning and MMOE architecture here.

    Hyperparameter tuning#

    -

    We provide a tutorial on how to do hyperparameter tuning with Merlin models and Weights&Biases Sweeps.

    -

    We also make it available a benchmark resulted from our own hyperparameter tuning of TenRec dataset. It compares the different single-task and multi-task learning models. It provides also empirical information on what were the improvements obtained with hyperparameter tuning, the curated hypertuning search space for modeling hyperparameters of ranking.py and the most important hyperparameters.

    +

    We provide a tutorial on how to do hyperparameter tuning with Merlin models and Weights&Biases Sweeps.

    +

    We also make it available a benchmark resulted from our own hyperparameter tuning of TenRec dataset. It compares the different single-task and multi-task learning models. It provides also empirical information on what were the improvements obtained with hyperparameter tuning, the curated hypertuning search space for modeling hyperparameters of ranking.py and the most important hyperparameters.

    Model Deployment on Triton Inference Server#

    -

    In the model deployment step, we deploy NVTabular workflow, and the trained and saved ranking model(s) on Triton Inference Server. The inference.py script makes it easy to export model configuration files and the required artifacts to deploy the models on Triton. Moreover, we provide an example notebook to demonstrate how to prepare a batch raw request to sent Triton and receive a response from it. You can find more information about inference step and the scripts here.

    +

    In the model deployment step, we deploy NVTabular workflow, and the trained and saved ranking model(s) on Triton Inference Server. The inference.py script makes it easy to export model configuration files and the required artifacts to deploy the models on Triton. Moreover, we provide an example notebook to demonstrate how to prepare a batch raw request to sent Triton and receive a response from it. You can find more information about inference step and the scripts here.

    @@ -652,8 +651,8 @@

    Model Deployment on Triton Inference Server - + +
    diff --git a/main/examples/quick_start/scripts/inference/index.html b/main/examples/quick_start/scripts/inference/index.html index cbc8e044f..224d9565b 100644 --- a/main/examples/quick_start/scripts/inference/index.html +++ b/main/examples/quick_start/scripts/inference/index.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -147,6 +147,7 @@ +

    NVIDIA Merlin

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -446,7 +445,7 @@

    Deploying a Ranking model on Triton Inference Server

    Creating the Ensemble Graph#

    -

    In order to do model deployment stage, you are required to complete preprocessing and ranking steps already from the Quick-start for Ranking. At the inference step, we might have a collection of multiple (individual) models to be deployed on Triton. In this example, we deploy our NVTabular workflow model to be able to transform raw data the same way as in the dataset preprocessing phase, in order to avoid the training-serving skew.

    +

    In order to do model deployment stage, you are required to complete preprocessing and ranking steps already from the Quick-start for Ranking. At the inference step, we might have a collection of multiple (individual) models to be deployed on Triton. In this example, we deploy our NVTabular workflow model to be able to transform raw data the same way as in the dataset preprocessing phase, in order to avoid the training-serving skew.

    In this context, deploying multiple models is called an ensemble model since it represents a pipeline of one or more models that are sequentially connected, i.e., output of a model is the input of next model. Ensemble models are intended to be used to encapsulate a procedure that involves multiple models, such as “data preprocessing -> inference -> data postprocessing”.

    The Triton Inference Server serves models from one or more model repositories that are specified when the server is started. Each model must include a configuration that provides required and optional information about the model. Merlin Systems simplified that step, so that we can easily export ensemble graph config files and artifacts. We use Ensemble class for that, which is responsible for interpreting the graph and exporting the correct files for the Triton server.

    Exporting an ensemble graph consists of the following steps:

    @@ -570,8 +569,8 @@

    Inputs - - + +
    diff --git a/main/examples/quick_start/scripts/inference/inference.html b/main/examples/quick_start/scripts/inference/inference.html index e7ee5301e..50604bb7e 100644 --- a/main/examples/quick_start/scripts/inference/inference.html +++ b/main/examples/quick_start/scripts/inference/inference.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -147,6 +147,7 @@ +

    NVIDIA Merlin

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -800,8 +799,8 @@

    Summary - + +
    diff --git a/main/examples/quick_start/scripts/preproc/index.html b/main/examples/quick_start/scripts/preproc/index.html index 7ac2b2d6a..53845d997 100644 --- a/main/examples/quick_start/scripts/preproc/index.html +++ b/main/examples/quick_start/scripts/preproc/index.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -147,6 +147,7 @@ +

    NVIDIA Merlin

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -540,9 +539,9 @@

    Command line argumentsIn this section we describe the command line arguments of the preprocessing script.

    The input and format can be CSV, TSV or Parquet, but the latter is recommended for being a columnar format which is faster to preprocess. Output preprocessing format is parquet format.

    -

    You can check how to setup the Docker container to run preprocessing.py script with Docker.

    +

    You can check how to setup the Docker container to run preprocessing.py script with Docker.

    -

    Here is an example command line for running preprocessing for the TenRec dataset in our Docker image, which is explained here. +

    Here is an example command line for running preprocessing for the TenRec dataset in our Docker image, which is explained here. The parameters and their values can be separated by either space or by =.

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -447,10 +446,10 @@

    Contents

    Benchmark of ranking models#

    In this document, we describe a hyperparameter tuning benchmark of the ranking models available in in Merlin Models library. You can use this benchmark as a reference when deciding the models and hyperparameters you want to explore with your own dataset.

    -

    The experiments used the Quick-Start for ranking script on the TenRec dataset (described here). Weights&Biases Sweeps was used for hyperparameter tuning.

    +

    The experiments used the Quick-Start for ranking script on the TenRec dataset (described here). Weights&Biases Sweeps was used for hyperparameter tuning.

    Neural ranking models.#

    -

    This benchmark includes the following neural architectures for ranking, which are described in more detail here. They are divided in two groups of models, which are trained with Single-Task or Multi-Task Learning (MTL).

    +

    This benchmark includes the following neural architectures for ranking, which are described in more detail here. They are divided in two groups of models, which are trained with Single-Task or Multi-Task Learning (MTL).

    • Single-Task Learning (STL): MLP, Wide&Deep, DeepFM, DLRM, DCN-v2, training with a single prediction head.

    • Multi-Task Learning (MTL): MLP, MMOE, PLE. All these models were built with a separate tower (MLP layers) and a head for each task. While MLP model shares bottom layers, MMOE and PLE are specialized MTL models that uses experts and gates designed to control the weight sharing between tasks.

    • @@ -460,15 +459,15 @@

      Neural ranking models.Hyperparameter tuning setup#

      For a fair comparison of the models, we ran a separate hyperparameter tuning process for each model architecture using TenRec dataset, which we call experiment group.

      We use the Weights&Biases Sweeps feature for managing the hypertuning process for each experiment group. The hypertuning uses bayesian optimization (method=bayes) to improve the AUC metric, which is different for STL and MTL, as explained below.

      -

      You can check our tutorial for more details on how to setup the hypertuning of Quick-start for ranking using W&B Sweeps. -We share the hyperparameter space configurations we used for this benchmark, so that you can reuse for your own hyperparameter tuning.

      +

      You can check our tutorial for more details on how to setup the hypertuning of Quick-start for ranking using W&B Sweeps. +We share the hyperparameter space configurations we used for this benchmark, so that you can reuse for your own hyperparameter tuning.

    Single-task learning#

    For benchmarking the ranking models with single-task learning we used the click binary target, as it is the most frequent event in the dataset. It was performed 200 trials for each experiment group.

    STL Benchmark results#

    -

    In Table 1, you can see the models with the best accuracy (AUC) for predicting the click target. You can see the models have a similar level of accuracy, maybe because the dataset contains only 5 basic features (which are presented here). But you can notice that models more advanced than MLP can provide better accuracy.

    +

    In Table 1, you can see the models with the best accuracy (AUC) for predicting the click target. You can see the models have a similar level of accuracy, maybe because the dataset contains only 5 basic features (which are presented here). But you can notice that models more advanced than MLP can provide better accuracy.

    Multi-task learning architectures
    Table 1. Single-task learning ranking models benchmark
    @@ -651,8 +650,8 @@

    Refine the search space - - + +
    diff --git a/main/examples/quick_start/scripts/ranking/hypertuning/tutorial_with_wb_sweeps.html b/main/examples/quick_start/scripts/ranking/hypertuning/tutorial_with_wb_sweeps.html index 16a267394..c7a23fa33 100644 --- a/main/examples/quick_start/scripts/ranking/hypertuning/tutorial_with_wb_sweeps.html +++ b/main/examples/quick_start/scripts/ranking/hypertuning/tutorial_with_wb_sweeps.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -147,6 +147,7 @@ +

    NVIDIA Merlin

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -447,7 +446,7 @@

    1. Creating the Sweeps config fileranking.py script. You will notice that we specify in program: ranking.py the script we want to execute, the method: bayes for bayesian optimization of the hyperparameters, toward maximizing the auc-final metric, which is computed on the evaluation set by the ranking script.
    Then you specify the hyperparameters distribution, which can be categorical, int_uniform, uniform, log_uniform, among others.

    The constant parameters that will not vary in the hyperparameter tuning (e.g. --epochs, --train_data_path, --eval_data_path) can be provided using a categorical distribution with a single value.

    -

    You can learn about more about the W&B sweeps configuration or about the ranking.py script hyperparameters available in its CLI.

    +

    You can learn about more about the W&B sweeps configuration or about the ranking.py script hyperparameters available in its CLI.

    program: ranking.py
     method: bayes
     metric:
    @@ -534,14 +533,14 @@ 

    1. Creating the Sweeps config filesweep config files from our hypertuning of ranking models for TenRec dataset. You can use their search space as a basis for setting up your config files for hypertuning ranking models on your own dataset.

    +

    We provide the sweep config files from our hypertuning of ranking models for TenRec dataset. You can use their search space as a basis for setting up your config files for hypertuning ranking models on your own dataset.

    2. Environment setup#

    -

    We need to prepare the environment configured for running the Quick-start scripts. The easiest way is to pull and run the Merlin Tensorflow image, as explained here, mapping the folder with the TenRec dataset.

    +

    We need to prepare the environment configured for running the Quick-start scripts. The easiest way is to pull and run the Merlin Tensorflow image, as explained here, mapping the folder with the TenRec dataset.

    -

    It assumes that you have already preprocessed the TenRec dataset using preprocessing.py, as explained here.

    +

    It assumes that you have already preprocessed the TenRec dataset using preprocessing.py, as explained here.

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -521,8 +520,8 @@

    Modeling inputs features

    Dealing with high-cardinality categorical features#

    -

    We explain in the Quick-start preprocessing documentation that large services might have categorical features with very high cardinality (e.g. order of hundreds of millions or higher), like user id or item id. They typically require a high memory to be stored (e.g. with embedding tables) or processed (e.g. with one-hot encoding). In addition, most of the categorical values are very infrequent, for which it is not possible to learn good embeddings.

    -

    The preprocessing documentation describes some options to deal with the high-cardinality features: Frequency capping, Filtering out rows with infrequent values and Hashing.

    +

    We explain in the Quick-start preprocessing documentation that large services might have categorical features with very high cardinality (e.g. order of hundreds of millions or higher), like user id or item id. They typically require a high memory to be stored (e.g. with embedding tables) or processed (e.g. with one-hot encoding). In addition, most of the categorical values are very infrequent, for which it is not possible to learn good embeddings.

    +

    The preprocessing documentation describes some options to deal with the high-cardinality features: Frequency capping, Filtering out rows with infrequent values and Hashing.

    You might also decide to keep the original high-cardinality of the categorical features for better personalization level and accuracy.

    The embedding tables are typically responsible for most of the parameters of Recommender System models. For large scale systems, where the number of users and items is in the order of hundreds of millions, it is typically needed to use a distributed embeddings solution, so that embedding embedding tables can be sharded in multiple compute devices (e.g. GPU, CPU).

    @@ -551,7 +550,7 @@

    Classes weights

    Negative sampling#

    If you have only positive interactions in your training data, you can use negative sampling to include synthetic negative examples in the training batch. The negative samples are generated by adding for each positive example N negative examples, keeping user features and replacing features of the target item by other item interacted by another users in the batch. You can easily set the number of negative examples for train (--in_batch_negatives_train) and evaluate (--in_batch_negatives_eval).
    -This functionality require that user and item features are tagged accordingly, as explained in the Quick-start preprocessing documentation.

    +This functionality require that user and item features are tagged accordingly, as explained in the Quick-start preprocessing documentation.

    Multi-task learning#

    @@ -575,9 +574,9 @@

    Setting tasks sample space#

    In this section we describe the command line arguments of the ranking.py script.

    -

    You can check how to setup the Docker image to run ranking.py script with Docker.

    +

    You can check how to setup the Docker image to run ranking.py script with Docker.

    -

    This is an example command line for running the training for the TenRec dataset in our Docker image, which is explained here. +

    This is an example command line for running the training for the TenRec dataset in our Docker image, which is explained here. The parameters and their values can be separated by either space or by =.

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -722,7 +721,7 @@

    Feature Engineering with NVTabular

    Build and Train a DLRM model#

    In this example, we build, train, and export a Deep Learning Recommendation Model (DLRM) architecture. To learn more about how to train different deep learning models, how easily transition from one model to another and the seamless integration between data preparation and model training visit 03-Exploring-different-models.ipynb notebook.

    -

    NVTabular workflow above exports a schema file, schema.pbtxt, of our processed dataset. To learn more about the schema object, schema file and tags, you can explore 02-Merlin-Models-and-NVTabular-integration.ipynb.

    +

    NVTabular workflow above exports a schema file, schema.pbtxt, of our processed dataset. To learn more about the schema object, schema file and tags, you can explore 02-Merlin-Models-and-NVTabular-integration.ipynb.

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -437,7 +436,7 @@

    Training and Serving Merlin on AWS SageMaker -
  • Training and Serving Merlin on AWS SageMaker

  • +
  • Training and Serving Merlin on AWS SageMaker

  • This notebook assumes that readers are familiar with some basic concepts in NVIDIA Merlin, such as:

    @@ -447,7 +446,7 @@

    Training and Serving Merlin on AWS SageMakerDeploying a Multi-Stage Recommender System +Deploying a Multi-Stage Recommender System in this repository or example notebooks in Merlin Models.

    @@ -455,7 +454,7 @@

    Running the Example Notebookmerlin-tensorflow container. -See Running the Example Notebooks +See Running the Example Notebooks for more details.

    Additionally, you need to configure basic AWS settings. For setting up AWS credentials, refer to @@ -463,7 +462,7 @@

    Running the Example Notebook-v $HOME/.aws:/root/.aws to your Docker command in Step 1 of -Running the Example Notebooks:

    +
    Running the Example Notebooks:

    docker run --gpus all --rm -it \
       -p 8888:8888 -p 8797:8787 -p 8796:8786 --ipc=host \
       -v $HOME/.aws:/root/.aws \
    @@ -471,7 +470,7 @@ 

    Running the Example NotebookRunning the Example Notebooks.

    +
    Running the Example Notebooks.

    @@ -557,8 +556,8 @@

    Running the Example Notebook - + +
    diff --git a/main/examples/sagemaker-tensorflow/sagemaker-merlin-tensorflow.html b/main/examples/sagemaker-tensorflow/sagemaker-merlin-tensorflow.html index 1a1f68d02..36e040d80 100644 --- a/main/examples/sagemaker-tensorflow/sagemaker-merlin-tensorflow.html +++ b/main/examples/sagemaker-tensorflow/sagemaker-merlin-tensorflow.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -481,7 +480,7 @@

    Training and Serving Merlin on AWS SageMakerDeploying a Multi-Stage Recommender System +Deploying a Multi-Stage Recommender System in this repository or example notebooks in Merlin Models.

    To run this notebook, you need to have Amazon SageMaker Python SDK installed.

    @@ -1736,8 +1735,8 @@

    Terminate endpoint and clean up artifacts - + +
    diff --git a/main/examples/scaling-criteo/01-Download-Convert.html b/main/examples/scaling-criteo/01-Download-Convert.html index a8a996b3f..b96aaeab9 100644 --- a/main/examples/scaling-criteo/01-Download-Convert.html +++ b/main/examples/scaling-criteo/01-Download-Convert.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -721,8 +720,8 @@

    Conversion Script for Criteo Dataset (CSV-to-Parquet) - - + +
    diff --git a/main/examples/scaling-criteo/02-ETL-with-NVTabular.html b/main/examples/scaling-criteo/02-ETL-with-NVTabular.html index 75e04991d..5cdabb5e5 100644 --- a/main/examples/scaling-criteo/02-ETL-with-NVTabular.html +++ b/main/examples/scaling-criteo/02-ETL-with-NVTabular.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -494,7 +493,7 @@

    ETL with NVTabularCriteo 1TB Click Logs dataset dataset. The following notebooks can use the output to train a deep learning model.

    Data Prep#

    -

    The previous notebook 01-Download-Convert converted the tsv data published by Criteo into the parquet format that our accelerated readers prefer. Accelerating these pipelines on new hardware like GPUs may require us to make new choices about the representations we use to store that data, and parquet represents a strong alternative.

    +

    The previous notebook 01-Download-Convert converted the tsv data published by Criteo into the parquet format that our accelerated readers prefer. Accelerating these pipelines on new hardware like GPUs may require us to make new choices about the representations we use to store that data, and parquet represents a strong alternative.

    We load the required libraries.

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -653,7 +652,7 @@

    Summary

    Next steps#

    -

    The next step is to deploy the NVTabular workflow and DLRM model to production.

    +

    The next step is to deploy the NVTabular workflow and DLRM model to production.

    If you are interested more in different architecture and training models with Merlin Models, we recommend to check out our Merlin Models examples

    @@ -679,11 +678,11 @@

    Next steps

    next

    -

    Scaling Criteo: Triton Inference with HugeCTR

    +

    Scaling Criteo: Triton Inference with Merlin Models TensorFlow

    @@ -744,8 +743,8 @@

    Next steps - + +
    diff --git a/main/examples/scaling-criteo/04-Triton-Inference-with-Merlin-Models-TensorFlow.html b/main/examples/scaling-criteo/04-Triton-Inference-with-Merlin-Models-TensorFlow.html index 2fabe8c49..1f9414f09 100644 --- a/main/examples/scaling-criteo/04-Triton-Inference-with-Merlin-Models-TensorFlow.html +++ b/main/examples/scaling-criteo/04-Triton-Inference-with-Merlin-Models-TensorFlow.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -54,7 +54,7 @@ - + - + +
    diff --git a/main/examples/scaling-criteo/index.html b/main/examples/scaling-criteo/index.html index c3eea540c..e06172077 100644 --- a/main/examples/scaling-criteo/index.html +++ b/main/examples/scaling-criteo/index.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -428,23 +427,22 @@

    Scaling Large Datasets with CriteoMerlin containers for the examples. Each notebook provides the required container.

    +

    Our recommendation is to use our latest stable Merlin containers for the examples. Each notebook provides the required container.

    Explore the following notebooks:

    Training and Deployment with TensorFlow:

    -

    Training and Deployment with HugeCTR:

    +

    Training with HugeCTR:

    @@ -516,8 +514,8 @@

    Scaling Large Datasets with Criteo - + +
    diff --git a/main/examples/traditional-ml/Serving-An-Implicit-Model-With-Merlin-Systems.html b/main/examples/traditional-ml/Serving-An-Implicit-Model-With-Merlin-Systems.html index 8eb69a20e..0972c102c 100644 --- a/main/examples/traditional-ml/Serving-An-Implicit-Model-With-Merlin-Systems.html +++ b/main/examples/traditional-ml/Serving-An-Implicit-Model-With-Merlin-Systems.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -147,6 +147,7 @@ +

    NVIDIA Merlin

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -804,8 +803,8 @@

    Retrieving Recommendations from Triton Inference Server - - + +
    diff --git a/main/examples/traditional-ml/Serving-An-XGboost-Model-With-Merlin-Systems.html b/main/examples/traditional-ml/Serving-An-XGboost-Model-With-Merlin-Systems.html index 9552a262d..6764898cf 100644 --- a/main/examples/traditional-ml/Serving-An-XGboost-Model-With-Merlin-Systems.html +++ b/main/examples/traditional-ml/Serving-An-XGboost-Model-With-Merlin-Systems.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -147,6 +147,7 @@ +

    NVIDIA Merlin

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -862,8 +861,8 @@

    Retrieving Recommendations from Triton Inference Server - - + +
    diff --git a/main/generated/nvcr.io-nvidia-merlin-merlin-hugectr.html b/main/generated/nvcr.io-nvidia-merlin-merlin-hugectr.html index c073dc1c3..5fab3354f 100644 --- a/main/generated/nvcr.io-nvidia-merlin-merlin-hugectr.html +++ b/main/generated/nvcr.io-nvidia-merlin-merlin-hugectr.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -147,6 +147,7 @@ +

    NVIDIA Merlin

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -1089,8 +1088,8 @@

    22.xx Container Images - + +
    diff --git a/main/generated/nvcr.io-nvidia-merlin-merlin-pytorch.html b/main/generated/nvcr.io-nvidia-merlin-merlin-pytorch.html index a9a59b10b..ebf49568a 100644 --- a/main/generated/nvcr.io-nvidia-merlin-merlin-pytorch.html +++ b/main/generated/nvcr.io-nvidia-merlin-merlin-pytorch.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -147,6 +147,7 @@ +

    NVIDIA Merlin

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -1103,8 +1102,8 @@

    22.xx Container Images - + +
    diff --git a/main/generated/nvcr.io-nvidia-merlin-merlin-tensorflow.html b/main/generated/nvcr.io-nvidia-merlin-merlin-tensorflow.html index 2c962f3be..1c0ee4efd 100644 --- a/main/generated/nvcr.io-nvidia-merlin-merlin-tensorflow.html +++ b/main/generated/nvcr.io-nvidia-merlin-merlin-tensorflow.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -147,6 +147,7 @@ +

    NVIDIA Merlin

    @@ -185,7 +186,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -204,7 +204,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -1103,8 +1102,8 @@

    22.xx Container Images - + +
    diff --git a/main/genindex.html b/main/genindex.html index f186cb19e..7a5cc5a0e 100644 --- a/main/genindex.html +++ b/main/genindex.html @@ -18,12 +18,12 @@ - - - + + + - + @@ -35,9 +35,9 @@ - - - + + + @@ -146,6 +146,7 @@ +

    NVIDIA Merlin

    @@ -184,7 +185,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -203,7 +203,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -424,8 +423,8 @@

    Index

    - - + +
    diff --git a/main/guide/recommender_models.html b/main/guide/recommender_models.html index 00308f633..da51f3c85 100644 --- a/main/guide/recommender_models.html +++ b/main/guide/recommender_models.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -601,8 +600,8 @@

    Resources for Session-Based Models: - + +
    diff --git a/main/guide/recommender_system_guide.html b/main/guide/recommender_system_guide.html index f597c9a49..fa51755fa 100644 --- a/main/guide/recommender_system_guide.html +++ b/main/guide/recommender_system_guide.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -502,8 +501,8 @@

    Recommender Systems Guide - + +
    diff --git a/main/index.html b/main/index.html index bb2767902..eba85ec5c 100644 --- a/main/index.html +++ b/main/index.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -148,6 +148,7 @@ +

    NVIDIA Merlin

    @@ -186,7 +187,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -205,7 +205,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -521,8 +520,8 @@

    Related Resources - + +
    diff --git a/main/objects.inv b/main/objects.inv index 9b6a6d709..fe61c991a 100644 Binary files a/main/objects.inv and b/main/objects.inv differ diff --git a/main/search.html b/main/search.html index 2286db288..d1a5651e2 100644 --- a/main/search.html +++ b/main/search.html @@ -17,12 +17,12 @@ - - - + + + - + @@ -34,9 +34,9 @@ - - - + + + @@ -148,6 +148,7 @@ +

    NVIDIA Merlin

    @@ -186,7 +187,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -205,7 +205,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -436,8 +435,8 @@

    Search

    - - + +
    diff --git a/main/searchindex.js b/main/searchindex.js index 6d4ad88d2..7ad9aaea7 100644 --- a/main/searchindex.js +++ b/main/searchindex.js @@ -1 +1 @@ -Search.setIndex({"docnames": ["README", "containers", "examples/Building-and-deploying-multi-stage-RecSys/01-Building-Recommender-Systems-with-Merlin", "examples/Building-and-deploying-multi-stage-RecSys/02-Deploying-multi-stage-RecSys-with-Merlin-Systems", "examples/Building-and-deploying-multi-stage-RecSys/index", "examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction", "examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction-with-pretrained-embeddings", "examples/getting-started-movielens/01-Download-Convert", "examples/getting-started-movielens/02-ETL-with-NVTabular", "examples/getting-started-movielens/03-Training-with-HugeCTR", "examples/getting-started-movielens/03-Training-with-PyTorch", "examples/getting-started-movielens/03-Training-with-TF", "examples/getting-started-movielens/04-Triton-Inference-with-HugeCTR", "examples/getting-started-movielens/04-Triton-Inference-with-TF", "examples/getting-started-movielens/index", "examples/index", "examples/quick_start/index", "examples/quick_start/ranking", "examples/quick_start/scripts/inference/index", "examples/quick_start/scripts/inference/inference", "examples/quick_start/scripts/preproc/index", "examples/quick_start/scripts/ranking/hypertuning/index", "examples/quick_start/scripts/ranking/hypertuning/tutorial_with_wb_sweeps", "examples/quick_start/scripts/ranking/index", "examples/ranking/index", "examples/ranking/tf/Training-and-Deploying-DLRM-model-with-Models-and-Systems", "examples/sagemaker-tensorflow/index", "examples/sagemaker-tensorflow/sagemaker-merlin-tensorflow", "examples/scaling-criteo/01-Download-Convert", "examples/scaling-criteo/02-ETL-with-NVTabular", "examples/scaling-criteo/03-Training-with-HugeCTR", "examples/scaling-criteo/03-Training-with-Merlin-Models-TensorFlow", "examples/scaling-criteo/04-Triton-Inference-with-HugeCTR", "examples/scaling-criteo/04-Triton-Inference-with-Merlin-Models-TensorFlow", "examples/scaling-criteo/index", "examples/traditional-ml/Serving-An-Implicit-Model-With-Merlin-Systems", "examples/traditional-ml/Serving-An-XGboost-Model-With-Merlin-Systems", "generated/nvcr.io-nvidia-merlin-merlin-hugectr", "generated/nvcr.io-nvidia-merlin-merlin-pytorch", "generated/nvcr.io-nvidia-merlin-merlin-tensorflow", "guide/recommender_models", "guide/recommender_system_guide", "index", "support_matrix/index", "support_matrix/support_matrix_merlin_hugectr", "support_matrix/support_matrix_merlin_pytorch", "support_matrix/support_matrix_merlin_tensorflow"], "filenames": ["README.md", "containers.rst", "examples/Building-and-deploying-multi-stage-RecSys/01-Building-Recommender-Systems-with-Merlin.ipynb", "examples/Building-and-deploying-multi-stage-RecSys/02-Deploying-multi-stage-RecSys-with-Merlin-Systems.ipynb", "examples/Building-and-deploying-multi-stage-RecSys/index.md", "examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction.ipynb", "examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction-with-pretrained-embeddings.ipynb", "examples/getting-started-movielens/01-Download-Convert.ipynb", "examples/getting-started-movielens/02-ETL-with-NVTabular.ipynb", "examples/getting-started-movielens/03-Training-with-HugeCTR.ipynb", "examples/getting-started-movielens/03-Training-with-PyTorch.ipynb", "examples/getting-started-movielens/03-Training-with-TF.ipynb", "examples/getting-started-movielens/04-Triton-Inference-with-HugeCTR.ipynb", "examples/getting-started-movielens/04-Triton-Inference-with-TF.ipynb", "examples/getting-started-movielens/index.md", "examples/index.md", "examples/quick_start/index.md", "examples/quick_start/ranking.md", "examples/quick_start/scripts/inference/index.md", "examples/quick_start/scripts/inference/inference.ipynb", "examples/quick_start/scripts/preproc/index.md", "examples/quick_start/scripts/ranking/hypertuning/index.md", "examples/quick_start/scripts/ranking/hypertuning/tutorial_with_wb_sweeps.md", "examples/quick_start/scripts/ranking/index.md", "examples/ranking/index.md", "examples/ranking/tf/Training-and-Deploying-DLRM-model-with-Models-and-Systems.ipynb", "examples/sagemaker-tensorflow/index.md", "examples/sagemaker-tensorflow/sagemaker-merlin-tensorflow.ipynb", "examples/scaling-criteo/01-Download-Convert.ipynb", "examples/scaling-criteo/02-ETL-with-NVTabular.ipynb", "examples/scaling-criteo/03-Training-with-HugeCTR.ipynb", "examples/scaling-criteo/03-Training-with-Merlin-Models-TensorFlow.ipynb", "examples/scaling-criteo/04-Triton-Inference-with-HugeCTR.ipynb", "examples/scaling-criteo/04-Triton-Inference-with-Merlin-Models-TensorFlow.ipynb", "examples/scaling-criteo/index.md", "examples/traditional-ml/Serving-An-Implicit-Model-With-Merlin-Systems.ipynb", "examples/traditional-ml/Serving-An-XGboost-Model-With-Merlin-Systems.ipynb", "generated/nvcr.io-nvidia-merlin-merlin-hugectr.rst", "generated/nvcr.io-nvidia-merlin-merlin-pytorch.rst", "generated/nvcr.io-nvidia-merlin-merlin-tensorflow.rst", "guide/recommender_models.rst", "guide/recommender_system_guide.rst", "index.rst", "support_matrix/index.rst", "support_matrix/support_matrix_merlin_hugectr.rst", "support_matrix/support_matrix_merlin_pytorch.rst", "support_matrix/support_matrix_merlin_tensorflow.rst"], "titles": ["NVIDIA Merlin", "Merlin Containers", "Building Intelligent Recommender Systems with Merlin", "Deploying a Multi-Stage RecSys into Production with Merlin Systems and Triton Inference Server", "Deploying a Multi-Stage Recommender System", "Transformer-based architecture for next-item prediction task", "Transformer-based architecture for next-item prediction task with pretrained embeddings", "Getting Started MovieLens: Download and Convert", "Getting Started MovieLens: ETL with NVTabular", "Getting Started MovieLens: Training with HugeCTR", "Getting Started MovieLens: Training with PyTorch", "Getting Started MovieLens: Training with TensorFlow", "Serve Recommendations from the HugeCTR Model", "Serve Recommendations from the TensorFlow Model", "Getting Started with Merlin and the MovieLens Dataset", "NVIDIA Merlin Example Notebooks", "Quick-start for Merlin", "Quick-start for ranking models with Merlin", "Deploying a Ranking model on Triton Inference Server", "Deploying Ranking Models with Merlin Systems", "Preprocessing script", "Benchmark of ranking models", "Hyperparameter tuning with Quick-start and Weights&Biases Sweeps", "Ranking script", "Training and Deploying Ranking models with Merlin", "Exporting Ranking Models", "Training and Serving Merlin on AWS SageMaker", "Training and Serving Merlin on AWS SageMaker", "Scaling Criteo: Download and Convert", "Scaling Criteo: ETL with NVTabular", "Scaling Criteo: Training with HugeCTR", "Scaling Criteo: Training with Merlin Models TensorFlow", "Scaling Criteo: Triton Inference with HugeCTR", "Scaling Criteo: Triton Inference with Merlin Models TensorFlow", "Scaling Large Datasets with Criteo", "Serving an Implicit Model with Merlin Systems", "Serving an XGBoost Model with Merlin Systems", "23.xx Container Images", "23.xx Container Images", "23.xx Container Images", "Recommender Models", "Recommender Systems Guide", "Merlin", "Merlin Support Matrix", "Merlin HugeCTR Support Matrix", "Merlin PyTorch Support Matrix", "Merlin TensorFlow Support Matrix"], "terms": {"an": [0, 2, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 18, 19, 20, 22, 23, 25, 27, 28, 29, 30, 31, 32, 33, 40, 42], "open": [0, 2, 4, 5, 12, 15, 18, 19, 24, 30, 32, 35, 36, 42], "sourc": [0, 2, 3, 5, 9, 11, 18, 19, 27, 30, 35, 36, 42], "librari": [0, 1, 4, 5, 6, 7, 8, 11, 12, 13, 15, 16, 17, 18, 19, 20, 21, 23, 27, 28, 29, 30, 31, 33, 35, 36, 40, 42], "acceler": [0, 8, 10, 11, 14, 15, 17, 19, 20, 26, 27, 29, 30, 35, 36, 42], "recommend": [0, 5, 7, 8, 9, 10, 11, 14, 15, 16, 17, 18, 19, 20, 23, 24, 26, 28, 29, 30, 31, 32, 33, 34, 42], "system": [0, 5, 6, 7, 8, 9, 10, 11, 13, 14, 15, 16, 18, 20, 23, 24, 26, 27, 28, 29, 30, 31, 32, 33, 34, 37, 38, 39, 40, 42, 44, 45, 46], "gpu": [0, 1, 2, 3, 4, 5, 6, 8, 9, 11, 12, 14, 15, 17, 20, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 34, 37, 38, 39, 40, 42, 44, 45, 46], "The": [0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 11, 12, 13, 14, 15, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46], "enabl": [0, 2, 3, 5, 6, 11, 19, 20, 22, 23, 25, 27, 30, 31, 44, 45, 46], "data": [0, 2, 3, 4, 5, 7, 8, 9, 10, 12, 13, 14, 15, 16, 18, 19, 23, 25, 27, 28, 30, 31, 32, 33, 37, 38, 39, 40, 44, 45, 46], "scientist": 0, "machin": [0, 2, 22, 23, 25, 29, 40], "learn": [0, 9, 12, 14, 15, 20, 22, 24, 25, 26, 27, 40], "engin": [0, 4, 6, 8, 11, 14, 15, 16, 17, 26, 27, 28, 29, 32, 33, 34, 40], "research": [0, 40], "build": [0, 3, 4, 17, 18, 23, 24, 40, 42], "high": [0, 8, 14, 15, 27, 29, 40, 42], "perform": [0, 1, 2, 3, 4, 5, 6, 11, 15, 17, 19, 20, 21, 23, 24, 25, 27, 31, 35, 36, 42], "scale": [0, 5, 6, 8, 19, 20, 23, 25, 35, 36, 40, 42], "includ": [0, 1, 5, 6, 15, 17, 18, 19, 20, 21, 22, 23, 27, 28, 32, 35, 36], "tool": [0, 2, 5, 6, 21, 25], "address": [0, 13], "common": [0, 17, 20, 23, 24, 27, 29, 40], "featur": [0, 4, 5, 7, 8, 10, 11, 14, 15, 16, 17, 21, 26, 27, 28, 29, 31, 32, 33, 34, 35, 36, 40], "train": [0, 1, 3, 4, 8, 12, 13, 14, 16, 18, 19, 21, 22, 28, 29, 32, 33, 34, 40, 43, 44, 45, 46], "infer": [0, 1, 2, 4, 6, 8, 9, 12, 14, 15, 23, 24, 26, 28, 30, 34, 37, 38, 39, 43, 44, 45, 46], "challeng": [0, 2, 6, 8], "each": [0, 1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 17, 18, 19, 20, 21, 22, 23, 25, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 40, 41, 43], "stage": [0, 2, 13, 16, 18, 21, 26, 27, 32, 40], "pipelin": [0, 2, 3, 4, 5, 6, 9, 10, 15, 17, 18, 19, 21, 23, 25, 27, 30, 32, 33, 35, 36], "optim": [0, 2, 3, 5, 6, 9, 10, 11, 19, 21, 22, 23, 25, 27, 30, 31], "hundr": [0, 20, 23, 40], "terabyt": [0, 8, 29], "which": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 15, 16, 17, 18, 19, 20, 21, 22, 23, 25, 27, 29, 30, 31, 33, 40], "all": [0, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 17, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 44, 45, 46], "access": [0, 1, 6, 12, 15, 19, 20, 30, 40], "through": [0, 3, 5, 6, 8, 9, 10, 11, 12, 16, 20, 23, 25, 27, 29, 33, 35, 36, 40], "easi": [0, 9, 17, 18, 19, 22, 23, 27, 30, 35, 36], "us": [0, 1, 2, 3, 4, 6, 7, 8, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 44, 45, 46], "api": [0, 5, 8, 9, 14, 15, 17, 18, 22, 23, 27, 29, 30, 31, 32, 33, 35, 36], "For": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 17, 18, 19, 20, 21, 22, 23, 25, 26, 28, 29, 32, 33, 35, 36, 40, 42], "more": [0, 1, 2, 3, 4, 5, 6, 8, 9, 10, 11, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 29, 30, 31, 32, 33, 35, 36, 40, 41, 42], "inform": [0, 1, 2, 4, 5, 6, 7, 8, 9, 10, 11, 13, 17, 18, 19, 20, 22, 23, 24, 25, 29, 30, 32, 33, 35, 36, 40, 42], "see": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 18, 19, 20, 21, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40, 42, 43], "develop": [0, 2, 5, 9, 19, 24, 27, 30, 35, 36, 41, 42], "web": 0, "site": 0, "scalabl": [0, 5, 18, 19, 23, 40], "solut": [0, 9, 20, 22, 23, 25, 30], "make": [0, 3, 5, 6, 8, 15, 17, 19, 20, 21, 22, 23, 25, 26, 27, 29, 35, 36], "from": [0, 1, 2, 4, 5, 6, 7, 8, 9, 10, 11, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 28, 29, 30, 31, 40, 41], "end": [0, 2, 5, 15, 19, 22, 23, 30, 33, 35, 36, 40], "With": [0, 1, 3, 5, 21, 26, 27, 32, 33, 35, 36], "you": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 44, 45, 46], "can": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 13, 15, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 40, 43, 44, 45, 46], "transform": [0, 2, 8, 13, 18, 19, 25, 29, 32, 35, 40], "etl": [0, 4, 9, 10, 15, 25, 28, 32, 36], "preprocess": [0, 2, 4, 6, 9, 12, 13, 14, 15, 16, 18, 19, 22, 23, 25, 26, 27, 32, 35, 36], "your": [0, 2, 3, 4, 5, 6, 11, 13, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 32, 33, 35, 36], "exist": [0, 2, 5, 6, 8, 9, 10, 22, 23, 27, 28, 29], "tensorflow": [0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 14, 15, 17, 19, 22, 23, 24, 25, 26, 27, 28, 29, 32, 34, 35, 36, 39, 40, 43], "pytorch": [0, 1, 2, 3, 5, 6, 7, 8, 14, 15, 19, 24, 25, 28, 29, 35, 36, 38, 40, 43], "fastai": 0, "leverag": [0, 6, 8, 10, 11, 15, 17, 23, 29, 30, 35, 36, 40], "custom": [0, 2, 5, 6, 8, 17, 20, 23, 25, 27, 35, 36, 40], "loader": [0, 3, 6, 9, 10], "larg": [0, 8, 9, 17, 20, 23, 30, 31, 40], "deep": [0, 2, 3, 5, 6, 8, 9, 11, 12, 17, 19, 21, 25, 27, 29, 30, 31, 32, 35, 36, 40], "model": [0, 1, 3, 4, 8, 10, 14, 15, 16, 20, 22, 26, 29, 30, 32, 34, 37, 38, 39, 41, 43, 44, 45, 46], "distribut": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 20, 22, 23, 25, 27, 28, 30, 31, 32, 33, 35, 36, 37, 38, 39, 44, 45, 46], "embed": [0, 1, 3, 4, 5, 8, 9, 10, 11, 17, 20, 21, 29, 30, 37, 38, 39, 40, 44, 45, 46], "tabl": [0, 1, 3, 9, 20, 21, 23, 30], "exce": 0, "avail": [0, 1, 4, 6, 9, 13, 15, 16, 17, 20, 21, 22, 23, 24, 25, 27, 28, 30, 31, 32, 34, 35, 36, 40, 42, 43], "cpu": [0, 2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 17, 19, 20, 23, 25, 27, 28, 29, 31], "memori": [0, 2, 3, 5, 6, 8, 9, 11, 17, 20, 23, 25, 27, 29, 30, 31], "deploi": [0, 2, 12, 13, 17, 20, 21, 22, 23, 26, 27, 31, 34, 35, 36, 40, 43, 44, 45, 46], "product": [0, 2, 4, 5, 6, 8, 9, 13, 18, 19, 20, 21, 25, 31, 32, 33, 35, 36, 40, 44, 45, 46], "onli": [0, 2, 5, 7, 8, 9, 10, 11, 12, 17, 18, 20, 21, 23, 25, 27, 31, 32, 35, 36, 40], "few": [0, 11, 23], "line": [0, 2, 3, 4, 6, 11, 15, 17, 21, 22, 24, 25, 27], "code": [0, 2, 3, 5, 6, 8, 11, 18, 19, 20, 25, 27, 29, 30, 35, 36, 40], "consist": [0, 3, 5, 6, 8, 18, 19, 35, 36], "follow": [0, 1, 2, 3, 4, 5, 6, 9, 11, 13, 14, 15, 17, 18, 19, 20, 21, 23, 24, 25, 26, 27, 29, 31, 32, 33, 34, 35, 36, 37, 38, 39, 43, 44, 45, 46], "nvtabular": [0, 3, 4, 6, 7, 9, 10, 11, 12, 14, 15, 17, 18, 19, 20, 26, 27, 28, 31, 32, 34, 35, 36, 37, 38, 39, 44, 45, 46], "tabular": [0, 5, 6, 8, 20, 23, 29], "quickli": [0, 8, 16, 17, 19, 29, 35, 36], "easili": [0, 2, 3, 5, 8, 17, 18, 23, 25, 29], "manipul": [0, 8, 29], "size": [0, 5, 6, 8, 10, 15, 17, 21, 27, 28, 29, 30, 34, 37, 38, 39, 44, 45, 46], "dataset": [0, 2, 3, 8, 9, 10, 11, 12, 13, 16, 18, 19, 21, 22, 23, 24, 25, 29, 30, 31, 32, 33, 40, 41], "ar": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40, 41], "base": [0, 2, 3, 4, 7, 8, 9, 10, 11, 13, 15, 16, 17, 19, 20, 24, 25, 27, 28, 29, 32, 33, 35, 36, 37, 38, 39, 41, 44, 45, 46], "offer": [0, 5, 9, 27, 30], "level": [0, 8, 14, 15, 17, 21, 23, 27, 29], "defin": [0, 2, 3, 5, 6, 7, 11, 13, 17, 18, 19, 22, 25, 27, 28, 30, 31, 33, 35, 36, 40], "complex": [0, 2, 8, 23, 41], "workflow": [0, 2, 3, 5, 8, 9, 11, 12, 15, 17, 18, 19, 20, 25, 27, 29, 31, 32, 35, 36, 44, 45, 46], "prepar": [0, 15, 19, 20, 22, 25, 27, 29], "experiment": 0, "so": [0, 2, 5, 6, 7, 8, 10, 11, 15, 17, 18, 19, 20, 21, 22, 23, 25, 27, 29, 32, 35, 36], "process": [0, 2, 5, 6, 11, 15, 16, 17, 20, 23, 25, 29, 30, 31, 32, 33, 35, 36, 40, 42], "without": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 16, 17, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40], "have": [0, 2, 3, 5, 6, 8, 9, 10, 12, 13, 15, 17, 18, 19, 20, 21, 22, 23, 25, 27, 29, 32, 35, 36, 40], "worri": 0, "about": [0, 1, 2, 5, 6, 8, 10, 11, 15, 17, 18, 22, 24, 25, 26, 27, 29, 30, 35, 36, 40, 42, 43], "focu": [0, 15, 19, 35, 36, 40], "what": [0, 2, 6, 9, 11, 13, 17, 29, 35, 36], "do": [0, 3, 4, 5, 6, 9, 12, 13, 17, 18, 19, 20, 25, 27, 32, 33, 35, 36], "how": [0, 1, 2, 4, 5, 6, 7, 8, 10, 11, 12, 15, 17, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 32, 33, 34, 35, 36, 40], "abstract": [0, 8, 25, 29], "oper": [0, 2, 3, 4, 5, 6, 8, 11, 19, 20, 23, 25, 27, 31, 35, 36, 37, 38, 39, 44, 45, 46], "hugectr": [0, 1, 7, 8, 14, 15, 28, 29, 34, 37, 38, 39, 40, 43, 45, 46], "framework": [0, 3, 9, 11, 15, 19, 24, 25, 29, 30, 31, 35, 36, 40], "across": [0, 8, 9, 20, 22, 29], "multipl": [0, 2, 3, 5, 6, 8, 9, 10, 15, 17, 18, 20, 21, 22, 23, 29, 30, 34, 40], "node": [0, 3, 5, 6, 9, 11, 15, 25, 27, 29, 30, 34, 35, 36], "contain": [0, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 16, 19, 20, 21, 22, 23, 24, 25, 26, 28, 29, 30, 31, 32, 33, 34, 35, 36, 40, 43], "provid": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 16, 17, 18, 20, 21, 22, 23, 24, 25, 27, 28, 29, 31, 33, 34, 40, 41, 43], "strategi": [0, 8, 17, 20, 40], "beyond": [0, 2], "over": [0, 2, 8, 9, 10, 16, 17, 22, 28, 29, 30, 31], "load": [0, 2, 5, 6, 7, 8, 9, 10, 11, 18, 19, 23, 25, 27, 29, 31, 32, 33, 35, 36], "subset": [0, 28], "coars": 0, "grain": 0, "demand": [0, 8], "manner": [0, 29], "dure": [0, 2, 5, 8, 11, 17, 18, 19, 21, 22, 23, 25, 32, 33, 35, 36], "standard": [0, 5, 6, 8, 18, 19, 20, 27, 29], "aim": [0, 23], "qualiti": [0, 40], "implement": [0, 5, 8, 10, 11, 20, 23, 27, 40], "rang": [0, 10, 12, 20, 21, 22, 23, 28, 29, 32], "classic": [0, 40], "highli": [0, 9, 10, 11, 13, 30], "advanc": [0, 5, 6, 20, 21, 23], "rank": [0, 3, 4, 16, 20, 22, 26, 27, 33, 36, 41], "up": [0, 4, 6, 8, 9, 10, 17, 19, 22, 23, 26, 29, 30, 35, 36], "10x": [0, 9, 30], "iter": [0, 8, 10, 16, 17, 28, 29, 32], "rapidli": 0, "explor": [0, 2, 14, 17, 21, 22, 25, 34, 40], "map": [0, 2, 3, 5, 6, 8, 10, 19, 20, 22, 23, 25, 29, 30, 32], "creat": [0, 4, 5, 7, 8, 9, 10, 11, 12, 13, 15, 17, 19, 20, 23, 28, 29, 30, 31, 32, 33], "input": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 17, 19, 22, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40], "layer": [0, 5, 6, 8, 10, 11, 17, 21, 23, 29, 40], "automat": [0, 5, 10, 18, 20, 23, 28, 32], "chang": [0, 3, 5, 6, 8, 20], "either": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 20, 21, 22, 23, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "impact": 0, "other": [0, 2, 3, 5, 6, 8, 11, 17, 19, 20, 21, 22, 23, 25, 27, 31], "assembl": 0, "connect": [0, 13, 18, 19, 25, 28, 29, 40], "block": [0, 5, 17, 23, 25, 40], "recsi": [0, 2, 4, 16, 17, 20], "architectur": [0, 2, 7, 8, 9, 11, 17, 21, 25, 29, 30, 31, 33, 37, 38, 39, 40, 41, 44, 45, 46], "new": [0, 2, 3, 8, 20, 23, 27, 29, 36, 40], "transformers4rec": [0, 37, 38, 39, 40, 44, 45, 46], "sequenti": [0, 3, 10, 18, 22, 41], "session": [0, 5, 6, 16, 27, 41], "modular": 0, "compat": [0, 2, 17, 20, 27, 29, 35, 36, 37, 38, 39, 44, 45, 46], "modul": [0, 2, 3, 5, 6, 8, 10, 11, 19, 23, 25, 28, 29, 36], "design": [0, 5, 8, 9, 11, 16, 17, 18, 20, 21, 24, 29], "tower": [0, 17, 21, 23, 40], "head": [0, 2, 5, 6, 7, 8, 12, 17, 20, 21, 23, 25, 36], "task": [0, 2, 3, 11, 19, 22, 25, 27, 31, 36, 40], "loss": [0, 2, 5, 6, 9, 10, 11, 17, 21, 25, 27, 30, 31], "ani": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 19, 20, 23, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 44, 45, 46], "take": [0, 2, 3, 5, 6, 7, 8, 10, 11, 19, 21, 23, 25, 28, 29, 32, 36], "advantag": [0, 9, 30, 40], "integr": [0, 2, 9, 10, 25], "seamless": [0, 25], "next": [0, 3, 7, 10, 16, 17, 18, 21, 25, 29, 30, 40], "item": [0, 3, 4, 7, 8, 10, 11, 17, 19, 20, 23, 25, 27, 35, 36, 40], "predict": [0, 2, 3, 8, 11, 12, 15, 17, 19, 20, 21, 23, 25, 27, 31, 35, 36, 40], "well": [0, 1, 5, 7, 8, 10, 11, 15, 17, 28, 29, 32, 34, 40], "binari": [0, 2, 3, 5, 6, 8, 9, 11, 15, 17, 19, 20, 21, 23, 25, 27, 31], "classif": [0, 15, 17, 20, 23], "regress": [0, 17, 20, 23, 36], "combin": [0, 3, 5, 6, 9, 10, 11, 20, 23, 30], "element": 0, "like": [0, 2, 4, 5, 6, 8, 11, 15, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 28, 29, 32, 35, 36], "store": [0, 4, 6, 10, 12, 15, 17, 20, 23, 27, 29, 32, 33, 40], "nearest": [0, 2, 3, 33, 40], "neighbor": [0, 3, 33, 40], "search": [0, 2, 3, 4, 15, 17, 22], "serv": [0, 2, 3, 4, 11, 14, 18, 19, 21, 25, 32, 33, 40], "triton": [0, 2, 4, 6, 9, 11, 14, 15, 16, 24, 26, 34, 37, 38, 39, 40, 44, 45, 46], "server": [0, 2, 4, 6, 9, 11, 14, 15, 22, 24, 26, 30, 34, 37, 38, 39, 44, 45, 46], "start": [0, 2, 4, 5, 6, 18, 20, 21, 23, 24, 27, 28, 29, 30, 31, 33, 40], "platform": [0, 2, 3, 5, 6, 11, 19, 20, 25, 27, 31], "graph": [0, 5, 8, 27, 30, 33], "gener": [0, 2, 3, 8, 9, 10, 12, 17, 18, 19, 20, 21, 23, 25, 32, 33, 35, 36, 40], "popular": [0, 2, 7, 9, 10, 11, 14, 15, 17, 20, 23, 30, 31, 40], "commonli": 0, "found": [0, 2, 3, 5, 6, 8, 9, 10, 11, 12, 21, 25, 27, 29, 31, 35, 36], "core": [0, 2, 3, 5, 6, 7, 8, 11, 12, 13, 15, 17, 19, 23, 25, 27, 28, 29, 31, 32, 33, 35, 36, 37, 38, 39, 44, 45, 46], "function": [0, 5, 6, 8, 10, 11, 18, 20, 23, 25, 27, 29, 31, 40], "throughout": [0, 29, 35, 36], "ecosystem": 0, "schema": [0, 2, 3, 5, 6, 8, 10, 11, 13, 17, 19, 20, 23, 25, 27, 29, 31, 32, 33, 35, 36], "identifi": [0, 1, 4, 5, 15, 24, 25, 29], "kei": [0, 1, 5, 6, 9, 12, 20, 22, 23, 32], "autom": [0, 17], "routin": 0, "simplifi": [0, 8, 18, 25, 29, 42], "share": [0, 17, 20, 21, 22, 23, 40], "construct": [0, 8, 35, 36], "simplest": 0, "wai": [0, 5, 6, 8, 11, 18, 19, 20, 22, 35, 36, 40], "run": [0, 1, 2, 3, 5, 6, 9, 12, 17, 18, 19, 20, 22, 23, 25, 27, 29, 32, 33, 35, 36, 37, 38, 39, 44, 45, 46], "docker": [0, 1, 4, 12, 13, 15, 20, 22, 23, 24, 26], "cloud": [0, 1, 15, 22, 24, 25], "ngc": [0, 1, 2, 4, 15, 27], "depend": [0, 6, 7, 8, 9, 10, 11, 13, 15, 17, 23, 25, 29, 30], "receiv": [0, 10, 13, 17, 18, 25, 32, 33, 35, 36, 40], "unit": 0, "test": [0, 2, 5, 6, 7, 17, 18, 25, 27, 28, 35], "page": [0, 2, 4, 17, 24, 43], "To": [0, 1, 2, 3, 4, 5, 6, 10, 11, 15, 17, 19, 24, 25, 26, 27, 31, 32, 43], "contribut": 0, "review": [0, 40], "document": [0, 5, 6, 8, 12, 16, 17, 19, 20, 21, 23, 24, 25, 26, 29, 32, 33, 35, 36, 43], "environ": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 17, 19, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36], "set": [0, 4, 5, 6, 7, 8, 11, 12, 17, 18, 19, 21, 22, 25, 26, 27, 28, 29, 32, 35, 36, 40, 41], "conda": 0, "pip": [0, 2, 3, 17, 22, 25, 27], "step": [0, 3, 5, 6, 10, 11, 12, 13, 15, 17, 18, 19, 23, 24, 25, 26, 27, 35, 36, 40], "http": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 19, 22, 23, 24, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 42], "github": [0, 3, 5, 8, 30, 42], "io": [0, 2, 4, 5, 8, 9, 10, 11, 12, 17, 22, 24, 25, 27, 29, 31, 35], "master": 0, "hugectr_contributor_guid": 0, "html": [0, 2, 5, 8, 9, 10, 11, 12, 25, 35], "A": [0, 5, 8, 9, 12, 17, 20, 23, 28, 40], "collect": [0, 8, 10, 15, 18, 27, 29], "form": [0, 5, 20, 23, 29], "jupyt": [0, 2, 4, 5, 8, 9, 10, 11, 12, 15, 19, 24, 25, 35, 36], "demonstr": [0, 3, 4, 15, 17, 18, 19, 22, 23, 24, 25, 34, 35, 36, 40], "download": [0, 14, 25, 27, 29, 34, 35, 36], "These": [0, 3, 5, 6, 11, 15, 18, 20, 25, 27, 31], "differ": [0, 2, 4, 7, 10, 15, 17, 20, 21, 23, 25, 27, 29, 31, 40, 41], "wide": [0, 9, 17, 21, 30, 40], "real": [0, 2, 25, 27], "world": [0, 2, 27], "case": [0, 3, 5, 6, 16, 17, 20, 21, 22, 23, 24, 27, 40, 41], "rapid": [0, 8, 29], "cudf": [0, 2, 3, 5, 6, 7, 8, 12, 13, 17, 19, 20, 25, 29, 35, 36, 37, 38, 39, 44, 45, 46], "reli": 0, "datafram": [0, 2, 5, 7, 8, 12, 13, 17, 20, 25, 27, 32], "dask": [0, 8, 17, 20, 28, 36], "within": [0, 5, 9, 27], "dataload": [0, 3, 6, 11, 14, 15, 17, 23, 31, 37, 38, 39, 44, 45, 46], "report": [0, 21, 27], "bug": [0, 35, 36], "get": [0, 2, 3, 5, 6, 13, 17, 19, 22, 23, 25, 27, 28, 29, 30, 31, 35, 36, 40], "help": [0, 3, 5, 6, 8, 15, 23, 25], "pleas": [0, 2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 18, 19, 22, 24, 25, 27, 29, 35, 36], "issu": [0, 2, 17, 20, 29], "compon": [1, 6, 15, 19, 23, 25, 35, 36, 42], "nvidia": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 18, 19, 22, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 40, 41, 42, 44, 45, 46], "ncg": 1, "catalog": [1, 2, 3, 4, 15, 23, 40], "com": [1, 3, 5, 15, 25, 27, 28, 42], "name": [1, 2, 3, 5, 6, 8, 10, 11, 12, 13, 17, 19, 21, 22, 23, 25, 27, 28, 30, 31, 32, 35, 36], "url": [1, 4, 6, 12, 13, 15, 19, 22, 24, 25, 27], "org": [1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 23, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "team": [1, 5, 22, 23], "plugin": 1, "must": [1, 5, 6, 11, 18, 23, 25, 27, 35, 36], "instal": [1, 2, 3, 5, 6, 17, 19, 22, 25, 27, 32, 33, 35, 36], "toolkit": [1, 27], "support": [1, 3, 8, 9, 10, 17, 20, 24, 25, 27, 28, 30, 31, 32, 33, 37, 38, 39], "link": [1, 5, 6, 17, 22, 41], "referenc": [1, 6, 13, 33], "preced": [1, 2], "launch": [1, 3, 4, 22, 25, 27], "befor": [1, 2, 3, 6, 13, 20, 23, 25, 27, 32, 40, 43], "one": [1, 4, 5, 6, 8, 10, 11, 15, 17, 18, 19, 20, 22, 23, 24, 25, 27, 28, 31, 32, 33, 35, 36, 40], "separ": [1, 8, 20, 21, 22, 23, 40, 43], "copyright": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "2023": [2, 3, 5, 6, 11, 19, 25], "corpor": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "right": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "reserv": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "licens": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "under": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 24, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "apach": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "version": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 24, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 43, 44, 45, 46], "2": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 20, 21, 23, 25, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 44, 45, 46], "0": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 40, 44, 45, 46], "mai": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "thi": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40, 41, 44, 45, 46], "file": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 18, 19, 20, 23, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "except": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 21, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "complianc": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "obtain": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 22, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "copi": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 19, 20, 24, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "www": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "unless": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "applic": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 44, 45, 46], "law": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "agre": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "write": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "softwar": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 18, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 43], "i": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 44, 45, 46], "AS": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "basi": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 22, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "warranti": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "OR": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "condit": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 20, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "OF": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "kind": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "express": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 20, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "impli": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "specif": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 20, 23, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40], "languag": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40], "govern": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 20, 21, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "permiss": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "limit": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "respons": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 18, 19, 23, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "check": [2, 3, 5, 6, 7, 9, 10, 11, 12, 13, 17, 19, 20, 21, 22, 23, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "content": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40, 41], "determin": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "suitabl": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 20, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "intend": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 18, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "notebook": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 17, 18, 19, 25, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36], "latest": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 22, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 40], "stabl": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36], "modern": 2, "internet": 2, "catalyst": 2, "human": 2, "decis": 2, "becaus": [2, 5, 6, 8, 17, 21, 23, 25, 27], "offlin": 2, "filter": [2, 5, 6, 15, 17, 23], "order": [2, 3, 5, 6, 8, 10, 15, 17, 18, 20, 23, 25, 27, 29, 35, 36, 40], "etc": [2, 20, 23, 40], "work": [2, 5, 6, 8, 16, 22, 25, 27], "togeth": [2, 20, 23, 42], "seamlessli": 2, "effici": [2, 3, 9, 10, 30], "biggest": 2, "practition": 2, "lack": 2, "understand": [2, 21], "around": 2, "look": [2, 3, 4, 5, 6, 7, 8, 10, 11, 15, 17, 24, 25, 27, 29, 32, 40], "gap": 2, "between": [2, 3, 6, 8, 20, 21, 23, 25, 40], "exampl": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 16, 17, 18, 19, 20, 21, 22, 23, 25, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 40, 41], "simpl": [2, 7, 10, 23, 25, 40], "readi": [2, 3, 5, 6, 11, 13, 17, 19, 22, 25, 29, 32, 33, 35, 36], "figur": [2, 3, 11, 21, 28, 40], "below": [2, 3, 5, 6, 9, 11, 12, 13, 18, 19, 20, 21, 22, 25, 27, 32, 35, 36], "repres": [2, 3, 5, 8, 10, 18, 20, 23, 25, 29], "four": [2, 5, 21, 28, 29, 31, 36], "than": [2, 5, 6, 10, 17, 20, 21, 23, 27, 29, 30, 40], "singl": [2, 5, 6, 8, 9, 10, 14, 15, 17, 18, 19, 22, 23, 29, 40], "much": [2, 16, 20, 21, 27], "realist": [2, 20], "closer": 2, "": [2, 3, 5, 6, 7, 8, 10, 11, 12, 13, 15, 17, 18, 19, 25, 27, 28, 29, 30, 31, 32, 35, 36, 40], "happen": [2, 23], "In": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 17, 18, 19, 20, 21, 22, 23, 25, 27, 29, 30, 32, 33, 35, 36, 40, 41], "seri": [2, 3, 19], "we": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 40, 41], "go": [2, 3, 8, 16, 17, 20, 22, 33, 35, 36], "showcas": [2, 21], "let": [2, 3, 5, 6, 8, 10, 11, 12, 13, 22, 25, 29, 31, 32, 35, 36], "concept": [2, 11, 14, 15, 26, 27], "briefli": 2, "narrow": 2, "down": [2, 8], "million": [2, 17, 20, 23], "thousand": [2, 23], "candid": [2, 3, 23, 33, 40], "relev": [2, 3, 17, 20, 23, 40], "top": [2, 3, 17, 20, 21, 23], "k": [2, 3, 20, 27], "exclud": 2, "alreadi": [2, 3, 6, 8, 17, 18, 20, 22, 23, 25, 27, 28, 29, 36], "interact": [2, 5, 11, 15, 16, 17, 20, 23, 30, 40], "undesir": 2, "appli": [2, 3, 6, 8, 12, 25, 29, 32, 35, 36, 40], "busi": [2, 25], "logic": [2, 25], "rule": [2, 40], "although": [2, 7, 10, 27], "skip": [2, 5, 6, 28, 35], "score": [2, 3, 13, 15, 19, 20, 23, 32, 33, 40], "also": [2, 5, 8, 17, 19, 20, 21, 22, 23, 25, 27], "known": [2, 25], "here": [2, 3, 5, 6, 8, 10, 11, 16, 17, 20, 21, 22, 23, 24, 27, 29], "being": [2, 5, 6, 20, 24], "abl": [2, 3, 5, 6, 9, 10, 12, 17, 18, 23, 27, 29, 30, 35, 36], "our": [2, 5, 6, 7, 10, 12, 13, 17, 18, 19, 20, 21, 22, 23, 25, 27, 31, 32, 33, 34, 35, 36, 40, 42], "At": [2, 3, 13, 18, 29, 40], "final": [2, 3, 5, 19, 20, 22, 25, 27, 31, 32, 33, 40], "want": [2, 6, 7, 8, 9, 15, 17, 20, 21, 22, 23, 25, 27, 29, 32, 33, 35, 36], "re": [2, 5, 6, 27, 29, 37, 38, 39, 44, 45, 46], "align": 2, "output": [2, 3, 4, 5, 6, 9, 10, 11, 12, 13, 17, 18, 19, 22, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40], "need": [2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 17, 19, 20, 22, 23, 25, 26, 27, 29, 30, 32, 33, 35], "constraint": 2, "criteria": 2, "listen": 2, "even": [2, 5, 6, 33], "oldridg": 2, "move": [2, 3, 5, 6, 9, 11, 12, 25, 32], "talk": [2, 40], "kdd": 2, "21": [2, 5, 8, 9, 10, 25, 27, 37, 38, 39, 44, 45, 46], "read": [2, 3, 5, 6, 8, 11, 19, 25, 27, 29, 30, 32, 33, 35, 36], "blog": [2, 29, 40], "post": [2, 5, 12, 13, 19, 23, 25, 40], "approxim": [2, 3, 33, 40], "neighbour": 2, "ann": [2, 3, 15, 40], "addit": [2, 7, 8, 15, 20, 23, 33, 40], "client": [2, 5, 6, 12, 13, 19, 25, 27, 28, 29, 32, 33, 35, 36], "extern": [2, 7, 8, 9, 10, 11, 13, 20, 29], "faiss": [2, 4, 15, 27], "similar": [2, 4, 8, 9, 15, 20, 21, 24, 25, 29, 30, 40], "cluster": [2, 3, 28, 36], "dens": [2, 9, 10, 11, 23, 30], "vector": [2, 3, 6, 11, 23, 40], "find": [2, 3, 4, 17, 21, 22, 23, 24, 30], "tag": [2, 4, 5, 6, 8, 10, 11, 17, 19, 20, 23, 24, 25, 27, 29, 31, 33, 35, 36], "most": [2, 6, 8, 14, 17, 20, 22, 23, 24, 29, 40], "recent": [2, 23, 40], "releas": [2, 4, 8, 16, 24, 27, 37, 38, 39, 43, 44, 45, 46], "refer": [2, 3, 8, 20, 21, 22, 26, 29], "31": [2, 3, 5, 12, 13, 27, 29, 37, 38, 39, 44, 45, 46], "uncom": [2, 27], "uninstal": [2, 22, 27], "o": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36], "nvt": [2, 6, 7, 8, 9, 10, 11, 12, 13, 25, 27, 28, 29, 32, 35, 36], "op": [2, 3, 5, 6, 8, 9, 11, 20, 25, 27, 29, 33, 35, 36], "renam": [2, 5, 11, 31, 35], "dropna": 2, "lambdaop": [2, 5, 8], "categorifi": [2, 5, 6, 8, 20, 25, 27, 29, 35, 36], "tagasuserfeatur": [2, 25, 27], "tagasuserid": [2, 8, 25, 27, 35], "tagasitemfeatur": [2, 25, 27], "tagasitemid": [2, 6, 8, 25, 27, 35], "addmetadata": [2, 25, 27, 29], "dag": [2, 3, 5, 6, 8, 11, 25, 27, 33, 35, 36], "subgraph": 2, "tf": [2, 3, 5, 6, 8, 10, 11, 25, 27, 31, 33], "mm": [2, 5, 6, 11, 25, 27, 31, 33], "ecommerc": [2, 5], "transform_aliccp": 2, "comment": [2, 20, 23], "out": [2, 3, 5, 6, 8, 10, 11, 12, 15, 17, 19, 20, 22, 23, 24, 25, 29, 31, 32, 33], "tf_gpu_alloc": [2, 5, 11, 17, 18, 19, 22, 23, 25, 31, 33], "cuda_malloc_async": [2, 5, 11, 17, 18, 19, 22, 23, 25, 31, 33], "06": [2, 3, 5, 6, 8, 25, 28, 29, 30, 31, 37, 38, 39, 43, 44, 45, 46], "29": [2, 3, 5, 6, 18, 19, 25, 27, 31, 35, 36, 37, 38, 39, 44, 45, 46], "19": [2, 3, 5, 27, 29, 37, 38, 39, 44, 45, 46], "49": [2, 37, 39, 44, 46], "32": [2, 3, 10, 11, 17, 22, 23, 25, 27, 28, 29, 31, 35], "836544": 2, "cpu_feature_guard": [2, 3, 5, 6, 11, 19, 25, 27, 31], "cc": [2, 3, 5, 6, 11, 18, 19, 25, 27, 31, 35, 36], "194": [2, 3, 5, 6, 11, 19, 27, 31], "oneapi": [2, 3, 5, 6, 11, 19, 27, 31], "neural": [2, 3, 5, 6, 7, 8, 9, 11, 19, 20, 23, 27, 29, 31, 40], "network": [2, 3, 5, 6, 7, 8, 9, 11, 17, 19, 20, 23, 27, 29, 31, 40], "onednn": [2, 3, 5, 6, 11, 19, 27, 31], "instruct": [2, 3, 5, 6, 11, 13, 18, 19, 25, 27, 31, 35, 36], "critic": [2, 3, 5, 6, 11, 19, 25, 27, 31], "sse3": [2, 3, 5, 6, 11, 19, 25, 27, 31], "sse4": [2, 3, 5, 6, 11, 19, 25, 27, 31], "1": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 17, 19, 20, 21, 23, 24, 25, 26, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 40, 43, 44, 45, 46], "avx": [2, 3, 5, 6, 11, 19, 25, 27, 31], "them": [2, 3, 5, 6, 8, 10, 11, 12, 17, 18, 19, 20, 23, 25, 27, 29, 31, 32, 33, 40, 41], "rebuild": [2, 3, 5, 6, 11, 19, 25, 27, 31], "appropri": [2, 3, 5, 6, 11, 19, 25, 27, 31], "compil": [2, 3, 5, 6, 9, 11, 19, 25, 27, 30, 31], "flag": [2, 3, 4, 5, 6, 11, 12, 13, 19, 25, 27, 31], "usr": [2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 31, 32, 35, 36], "local": [2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 15, 19, 22, 25, 27, 28, 29, 35, 36], "lib": [2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 31, 32, 35, 36], "python3": [2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 31, 32, 35, 36], "8": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 22, 25, 27, 28, 29, 30, 31, 32, 35, 36, 37, 38, 39, 44, 45, 46], "dist": [2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 31, 32, 35, 36], "packag": [2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 19, 22, 25, 26, 27, 28, 31, 32, 35, 36, 37, 38, 39, 44, 45, 46], "dtype": [2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 19, 25, 28, 29, 31, 36], "torch": [2, 3, 5, 6, 10, 19, 25], "py": [2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 17, 18, 19, 20, 22, 23, 25, 27, 28, 30, 31, 32, 35, 36], "43": [2, 3, 5, 6, 19, 25, 36, 39, 46], "userwarn": [2, 3, 5, 6, 8, 10, 19, 25, 27, 35, 36], "did": [2, 3, 5, 6, 8, 10, 19, 25, 35, 36], "successfulli": [2, 3, 5, 6, 8, 10, 19, 25, 27], "due": [2, 3, 5, 6, 8, 10, 19, 21, 25, 27], "error": [2, 3, 5, 6, 8, 10, 17, 19, 20, 25, 27], "No": [2, 3, 5, 6, 8, 10, 11, 19, 25, 27], "warn": [2, 3, 5, 6, 8, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 35, 36], "f": [2, 3, 5, 6, 8, 10, 11, 12, 18, 19, 25, 27, 29, 30, 32], "exc": [2, 3, 5, 6, 8, 10, 19, 25], "msg": [2, 3, 5, 6, 8, 10, 19, 25], "fix": [2, 5, 6, 11, 23, 25, 29], "python": [2, 3, 5, 6, 9, 11, 15, 17, 18, 20, 22, 23, 24, 25, 28, 30, 31], "track": [2, 5, 6, 11, 22, 25], "data_structur": [2, 5, 6, 11, 25], "ha": [2, 3, 5, 6, 7, 8, 10, 11, 17, 20, 21, 23, 25, 28, 35, 36, 40], "been": [2, 5, 6, 8, 11, 12, 13, 20, 23, 25, 27], "trackabl": [2, 5, 6, 11, 25], "old": [2, 5, 6, 7, 11, 25], "delet": [2, 5, 6, 11, 25, 28], "11": [2, 5, 6, 11, 25, 27, 30, 31, 32, 37, 38, 39, 44, 45, 46], "info": [2, 3, 5, 6, 11, 23, 25, 27, 28, 29, 30, 31, 36], "sparse_operation_kit": [2, 5, 6, 25], "sok": [2, 5, 6, 25], "merlin_sok": [2, 5, 6, 25, 27], "4": [2, 5, 6, 7, 8, 10, 12, 15, 17, 19, 23, 25, 27, 29, 30, 31, 32, 34, 35, 36, 37, 38, 39, 40, 44, 45, 46], "py3": [2, 5, 6, 25, 27], "linux": [2, 5, 6, 25, 27], "x86_64": [2, 5, 6, 25, 27], "egg": [2, 5, 6, 25, 27], "libsok_experi": [2, 5, 6, 25], "37": [2, 3, 27, 30, 32], "094972": 2, "38": [2, 6, 27, 29, 32, 37, 44], "134481": 2, "w": [2, 5, 6, 12, 21, 22, 25, 30, 32], "common_runtim": [2, 3, 5, 6, 11, 25, 27, 31], "gpu_bfc_alloc": [2, 5, 6, 25], "42": [2, 5, 6, 23, 35, 37, 39, 44, 46], "overrid": [2, 5, 6, 20, 25], "orig_valu": [2, 5, 6, 25], "tf_force_gpu_allow_growth": [2, 5, 6, 25], "variabl": [2, 3, 5, 6, 17, 22, 25, 27, 28, 29], "origin": [2, 5, 6, 7, 10, 11, 20, 23, 25, 28], "config": [2, 3, 5, 6, 11, 12, 13, 18, 25, 27, 32, 33], "valu": [2, 3, 5, 6, 10, 11, 12, 19, 20, 22, 23, 25, 27, 28, 29, 32, 33, 35, 36], "wa": [2, 5, 6, 8, 11, 17, 20, 21, 25, 27, 28, 33, 35, 36, 40], "134526": 2, "gpu_devic": [2, 3, 5, 6, 11, 25, 27, 31], "1621": [2, 3, 5, 6], "devic": [2, 3, 5, 6, 8, 10, 11, 20, 23, 25, 27, 28, 29, 30, 31], "job": [2, 3, 5, 6, 11, 21, 22, 25, 27, 31], "localhost": [2, 3, 5, 6, 11, 12, 13, 19, 25, 27, 31, 32, 33, 35, 36], "replica": [2, 3, 5, 6, 11, 25, 27, 31], "24576": [2, 3, 5, 6, 11], "mb": [2, 3, 5, 6, 11, 25, 27, 31], "quadro": [2, 3, 5, 6, 11, 25], "rtx": [2, 3, 5, 6, 11], "8000": [2, 3, 4, 5, 6, 11, 12, 13, 18, 19, 25, 35, 36], "pci": [2, 3, 5, 6, 11, 25, 27, 31], "bu": [2, 3, 5, 6, 11, 25, 27, 31], "id": [2, 3, 5, 6, 8, 10, 11, 12, 13, 17, 20, 22, 23, 25, 27, 28, 31, 40], "0000": [2, 3, 5, 6, 11, 25, 27, 31], "15": [2, 3, 8, 17, 25, 29, 30, 35, 37, 38, 39, 44, 45, 46], "00": [2, 3, 5, 6, 7, 10, 11, 25, 27, 29, 31, 35, 36, 37, 38, 39, 44, 45, 46], "comput": [2, 3, 5, 6, 8, 11, 20, 21, 22, 23, 25, 27, 29, 31, 35, 36, 40], "capabl": [2, 3, 5, 6, 11, 15, 25, 27, 31], "7": [2, 3, 5, 6, 8, 10, 11, 17, 19, 20, 25, 27, 28, 29, 31, 35, 36, 37, 38, 39, 43, 44, 45, 46], "5": [2, 3, 5, 6, 7, 8, 11, 12, 17, 19, 20, 21, 23, 25, 27, 29, 31, 32, 35, 36, 37, 38, 39, 44, 45, 46], "135533": 2, "135562": 2, "2d": [2, 3, 25], "tqdm": [2, 5, 8, 9, 10, 11, 12, 25, 35], "auto": [2, 5, 8, 9, 10, 11, 12, 25, 35], "tqdmwarn": [2, 5, 8, 9, 10, 11, 12, 25, 35], "iprogress": [2, 5, 8, 9, 10, 11, 12, 25, 35], "updat": [2, 3, 5, 6, 8, 9, 10, 11, 12, 25, 27, 33, 35, 40], "ipywidget": [2, 5, 8, 9, 10, 11, 12, 25, 35], "readthedoc": [2, 5, 8, 9, 10, 11, 12, 25, 35], "en": [2, 5, 8, 9, 10, 11, 12, 25, 35], "user_instal": [2, 5, 8, 9, 10, 11, 12, 25, 35], "autonotebook": [2, 5, 8, 9, 10, 11, 12, 25, 35], "notebook_tqdm": [2, 5, 8, 9, 10, 11, 12, 25, 35], "initi": [2, 3, 5, 6, 8, 12, 17, 20, 23, 25, 27, 28, 29, 31, 36], "finish": [2, 3, 5, 6, 18, 22, 25], "commun": [2, 3, 5, 6, 25, 27, 31, 40], "horovod": [2, 5, 6, 25], "disabl": [2, 12], "debug": [2, 23, 25, 27], "log": [2, 4, 15, 17, 20, 22, 24, 27, 29, 31, 34], "everywher": 2, "synthet": [2, 23, 25, 27, 40], "mimick": [2, 25, 27], "ali": [2, 25, 27], "ccp": [2, 25, 27], "alibaba": [2, 25, 27], "click": [2, 3, 9, 15, 17, 19, 20, 21, 22, 23, 25, 27, 29, 31, 32, 34, 40], "convers": [2, 15, 25, 27], "first": [2, 3, 5, 6, 7, 8, 10, 12, 13, 17, 19, 20, 21, 22, 23, 25, 28, 29, 30, 32, 33], "path": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 18, 19, 20, 22, 23, 25, 27, 28, 29, 30, 31, 32, 33], "repo": [2, 3], "data_fold": [2, 3, 25, 27], "workspac": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 25, 27, 36], "dir": [2, 3, 28], "base_dir": [2, 3, 28, 29, 30, 32, 33], "multi": [2, 7, 8, 10, 14, 16, 19, 20, 26, 27, 28, 29, 40], "Then": [2, 22, 23], "generate_data": [2, 6, 25, 27], "util": [2, 3, 5, 6, 7, 10, 11, 12, 19, 25, 27, 28, 36], "num_row": [2, 25, 27], "100_000": [2, 6], "train_raw": 2, "valid_raw": 2, "aliccp": [2, 25, 27], "raw": [2, 3, 5, 8, 17, 18, 19, 20, 25, 32, 33], "int": [2, 3, 5, 6, 7, 10, 25, 27, 29, 31], "set_siz": [2, 25, 27], "3": [2, 5, 6, 7, 8, 9, 10, 12, 13, 17, 19, 23, 25, 28, 29, 31, 32, 33, 35, 36, 37, 38, 39, 40, 44, 45, 46], "If": [2, 5, 6, 17, 19, 20, 22, 23, 25, 27, 28, 30, 31, 32, 33, 35, 36, 40], "would": [2, 5, 6, 20, 23, 25, 27, 35, 36, 40], "get_aliccp": [2, 25, 27], "instead": [2, 4, 12, 13, 21, 23, 25, 27, 36], "csv": [2, 5, 6, 7, 8, 17, 19, 20, 23, 25, 27], "parquet": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 20, 23, 25, 27, 29, 30, 31, 32, 33], "directli": [2, 3, 5, 6, 11, 25, 27, 31], "fed": [2, 10, 20, 23, 40], "abov": [2, 3, 5, 6, 9, 12, 13, 22, 25, 40], "onto": [2, 11, 27], "repositori": [2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 18, 19, 25, 26, 27, 30, 32, 33, 35, 36, 42], "customiz": 2, "infrastructur": [2, 40], "manag": [2, 17, 21, 22, 35, 36], "time": [2, 3, 5, 6, 8, 10, 12, 13, 17, 20, 21, 22, 23, 25, 27, 28, 29, 30, 31, 32], "current": [2, 7, 8, 16, 20, 23, 24, 25, 27, 32, 33, 40], "directori": [2, 4, 6, 7, 8, 9, 10, 11, 12, 13, 15, 24, 25, 27, 28, 29, 30, 31, 32, 33, 36], "u": [2, 5, 6, 11, 12, 25, 27, 29, 35, 36], "rm": [2, 9, 15, 17, 22, 24, 26, 30, 32], "rf": [2, 9, 30, 32], "feast_repo": [2, 3], "cd": [2, 3, 17, 18, 20, 22, 23, 27], "init": [2, 6, 22, 30], "1m": [2, 3], "32m": 2, "raid": [2, 3, 28, 29, 30, 31, 32, 33], "workshar": [2, 3], "0m": [2, 3, 27], "should": [2, 3, 4, 6, 8, 9, 13, 15, 20, 21, 22, 23, 24, 28, 32, 40], "messag": [2, 3, 4, 5, 6, 15, 18, 19, 22, 24, 25, 35, 36], "print": [2, 3, 5, 10, 12, 13, 19, 22, 25, 27, 29, 30, 32, 33], "now": [2, 3, 5, 6, 8, 9, 11, 13, 18, 19, 22, 25, 29, 32, 35, 36], "navig": [2, 4, 15, 24], "feature_repo": [2, 3], "folder": [2, 3, 5, 9, 16, 17, 18, 19, 22, 23, 24, 25, 28, 32], "remov": [2, 3, 4, 5, 8, 12, 13, 17, 19, 20, 25, 27, 28, 32, 33], "demo": [2, 27], "default": [2, 5, 6, 20, 22, 23, 27, 28], "feature_repo_path": 2, "join": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 19, 25, 27, 28, 29, 30, 31, 32, 33], "example_repo": 2, "driver_stat": 2, "unique_rows_by_featur": 2, "user_featur": [2, 3, 20, 25, 27], "user_id": [2, 3, 5, 8, 17, 19, 20, 23, 25, 27, 35, 40], "reset_index": [2, 19, 35], "drop": [2, 6, 7, 19, 32, 35, 36], "true": [2, 3, 5, 6, 8, 9, 10, 12, 13, 17, 19, 23, 25, 27, 28, 29, 30, 31, 32, 35], "artifici": 2, "add": [2, 3, 5, 8, 9, 12, 13, 16, 20, 22, 23, 25, 30], "datetim": 2, "timestamp": [2, 7, 20, 40], "column": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 20, 23, 25, 27, 28, 29, 32, 33, 35, 36, 40], "creation": [2, 12, 13, 19, 25, 27], "when": [2, 3, 4, 5, 6, 8, 9, 15, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 29, 30, 35, 36, 40], "astyp": [2, 3, 5, 6, 8], "datetime64": 2, "n": [2, 8, 12, 13, 17, 20, 23, 27, 29], "user_shop": [2, 3, 25, 27], "user_profil": [2, 3, 25, 27], "user_group": [2, 3, 25, 27], "user_gend": [2, 3, 23, 25, 27], "user_ag": [2, 3, 25, 27, 40], "user_consumption_1": 2, "user_consumption_2": [2, 3, 25, 27], "user_is_occupi": [2, 3, 25, 27], "user_geographi": [2, 3, 25, 27], "user_intent": [2, 3, 25, 27], "user_brand": [2, 3, 25, 27], "user_categori": [2, 3, 25, 27], "6": [2, 5, 6, 8, 10, 17, 19, 22, 23, 25, 27, 29, 35, 37, 38, 39, 44, 45, 46], "590": 2, "171": 2, "293": 2, "50": [2, 3, 5, 6, 18, 19, 23, 25, 29, 35, 36, 37, 38, 39, 44, 45, 46], "300270": 2, "303330": 2, "to_parquet": [2, 5, 6, 7, 8, 25, 27, 28, 29], "item_featur": [2, 3, 20, 23, 25, 27], "item_id": [2, 3, 5, 8, 17, 19, 20, 23, 25, 27], "item_categori": [2, 3, 25, 27], "item_shop": [2, 3, 25, 27], "item_brand": [2, 3, 25, 27], "item_intent": 2, "410715": 2, "412307": 2, "412": 2, "142": 2, "66": [2, 25], "12": [2, 5, 12, 25, 27, 28, 29, 31, 32, 37, 38, 39, 44, 45, 46], "824": 2, "284": 2, "132": [2, 27], "18": [2, 3, 5, 6, 18, 19, 25, 27, 29, 35, 36, 38, 39, 45, 46], "1236": 2, "426": 2, "197": [2, 36], "24": [2, 5, 27, 28, 29, 30, 31, 36], "1648": 2, "568": 2, "263": 2, "disk": [2, 5, 6, 7, 8, 11, 17, 28, 29, 31, 32, 33, 35, 36], "output_path": [2, 8, 17, 18, 20, 22, 23, 25, 27, 28, 29, 32], "processed_nvt": [2, 3], "notic": [2, 17, 21, 22], "row": [2, 5, 6, 8, 13, 17, 19, 21, 23, 32, 33, 40], "miss": [2, 6, 20, 29], "after": [2, 3, 4, 5, 6, 9, 10, 11, 15, 17, 18, 19, 21, 22, 23, 24, 25, 26, 27, 31, 32, 33, 35, 36], "doe": [2, 5, 19, 21, 25, 37, 38, 39, 44, 45, 46], "null": [2, 20, 27], "entri": 2, "might": [2, 6, 15, 17, 18, 20, 21, 22, 23, 27], "own": [2, 16, 17, 21, 22, 26, 27], "therefor": [2, 25, 32], "while": [2, 3, 5, 6, 8, 11, 21, 22, 23, 25, 27, 31], "regist": [2, 6, 17], "user_id_raw": [2, 3], "item_id_raw": 2, "avoid": [2, 5, 6, 17, 18, 20, 23], "potenti": 2, "occur": [2, 23, 29], "postfix": 2, "_raw": 2, "lambda": [2, 5, 6, 8, 12, 20], "col": [2, 5, 8, 13, 25, 28, 29, 32, 33, 35, 36], "int32": [2, 3, 10, 12, 19, 20, 25, 27, 28, 29, 32, 33, 36], "item_cat": 2, "subgraph_item": 2, "items_cat": 2, "subgraph_us": 2, "target": [2, 3, 5, 6, 8, 11, 17, 19, 21, 23, 25, 27, 29, 31, 32, 33, 36, 40], "binary_classif": [2, 25, 27, 29], "nvt_wkflow": 2, "call": [2, 3, 5, 6, 8, 18, 21, 27, 32, 40], "fit": [2, 5, 6, 8, 9, 11, 23, 25, 27, 29, 30, 31, 35, 36], "nvt_workflow": [2, 3], "workflow_nam": 2, "visit": [2, 5, 6, 24, 25], "05": [2, 5, 8, 10, 17, 19, 22, 23, 35, 36, 37, 38, 39, 44, 45, 46], "ipynb": [2, 3, 18, 25], "categor": [2, 5, 6, 7, 8, 10, 12, 14, 15, 17, 20, 22, 25, 28, 29, 32], "encod": [2, 7, 10, 12, 23, 30, 32], "output_path2": 2, "train_tt": 2, "valid_tt": 2, "valid": [2, 5, 8, 9, 10, 11, 12, 13, 19, 23, 25, 27, 28, 29, 30, 31], "select": [2, 5, 17, 23, 29], "posit": [2, 5, 17, 20, 21, 23, 40], "where": [2, 3, 5, 17, 20, 21, 22, 23, 27, 29, 32, 40], "column_nam": [2, 3, 5, 6, 8, 11, 19, 25, 27, 31, 32, 33, 35, 36], "df": [2, 5, 6, 8, 12], "pbtxt": [2, 3, 12, 17, 25, 27, 32], "protobuf": [2, 27], "text": [2, 8, 12, 13, 19, 25, 27, 28], "02": [2, 9, 19, 25, 32, 37, 38, 39, 44, 45, 46], "part_siz": [2, 6, 8, 29], "500mb": [2, 6], "select_by_tag": [2, 5, 6, 11, 25, 27, 31, 33], "model_tt": 2, "twotowermodel": 2, "query_tow": [2, 3], "mlpblock": [2, 5, 6, 11, 23, 25, 27, 31], "128": [2, 5, 6, 9, 10, 17, 22, 23, 25, 27, 30, 31, 32], "64": [2, 5, 6, 11, 12, 13, 17, 22, 23, 25, 27, 29, 30, 31, 32], "no_activation_last_lay": [2, 5, 6], "sampler": 2, "inbatchsampl": 2, "embedding_opt": 2, "embeddingopt": 2, "infer_embedding_s": 2, "adam": [2, 5, 6, 9, 10, 23, 25, 27], "run_eagerli": [2, 5, 6, 11, 25, 27, 31], "fals": [2, 5, 6, 7, 8, 9, 10, 11, 19, 20, 25, 27, 30, 31], "categorical_crossentropi": [2, 5, 6], "metric": [2, 5, 6, 11, 18, 19, 21, 22, 23, 25, 27, 35, 36], "recallat": 2, "10": [2, 3, 9, 12, 17, 19, 20, 22, 23, 25, 27, 28, 29, 32, 35, 36, 37, 38, 39, 44, 45, 46], "ndcgat": 2, "validation_data": [2, 5, 11, 25, 27, 31], "batch_siz": [2, 5, 6, 10, 11, 25, 27, 31], "1024": [2, 10, 11, 23, 27, 30, 31], "epoch": [2, 5, 6, 10, 17, 22, 23, 27, 31], "kera": [2, 5, 6, 9, 11, 15, 23, 25, 27, 30, 31, 33], "initializers_v2": [2, 5, 6], "120": [2, 5, 6], "truncatednorm": [2, 5, 6], "unseed": [2, 5, 6], "return": [2, 3, 5, 6, 8, 10, 11, 12, 13, 20, 25, 27, 35, 36], "ident": [2, 5, 6, 27], "seed": [2, 5, 6, 23, 30], "same": [2, 3, 5, 6, 8, 17, 18, 19, 20, 22, 23, 25, 28, 29, 32, 33, 35, 36], "instanc": [2, 5, 6, 22, 25, 27], "onc": [2, 3, 5, 6, 8, 18, 27, 29, 35, 36], "9": [2, 6, 11, 12, 19, 25, 27, 29, 32, 35, 37, 38, 39, 44, 45, 46], "275m": 2, "9538": 2, "recall_at_10": [2, 5, 6], "0101": 2, "ndcg_at_10": [2, 5, 6], "0067": 2, "regularization_loss": [2, 5, 6, 11, 25, 27, 31], "0000e": [2, 5, 6, 11, 25, 27, 31], "loss_batch": [2, 5, 6, 11, 25, 31], "8711": 2, "val_loss": [2, 11, 25, 27, 31], "9179": 2, "val_recall_at_10": 2, "0212": 2, "val_ndcg_at_10": 2, "0155": 2, "val_regularization_loss": [2, 11, 25, 27, 31], "val_loss_batch": [2, 11, 25, 31], "5806": [2, 5], "callback": [2, 5, 6, 11, 25, 31], "histori": [2, 5, 6, 11, 25, 31, 40], "0x7fd4b04139d0": 2, "later": [2, 5, 6, 8, 20, 23, 27, 32, 37, 38, 39, 44, 45, 46], "deploy": [2, 5, 9, 16, 18, 19, 21, 25, 34, 40], "retrieval_block": 2, "query_block": 2, "target_column": [2, 11, 25, 27], "propos": [2, 11, 23], "facebook": [2, 9, 27, 30, 31, 40], "2019": 2, "introduc": [2, 3, 20, 23], "person": [2, 23, 40], "spars": [2, 23, 29], "multilay": 2, "perceptron": [2, 23, 40], "mlp": [2, 21, 22, 23, 40], "explicitli": [2, 23], "statist": [2, 8, 17, 20, 29], "techniqu": [2, 5, 6, 20, 23, 40], "architetctur": 2, "gh": 2, "dlrmmodel": [2, 15, 23, 25, 27, 31], "embedding_dim": [2, 23, 25, 27, 31], "bottom_block": [2, 25, 27, 31], "top_block": [2, 25, 27, 31], "prediction_task": [2, 11, 25, 27, 31], "binaryclassificationtask": [2, 27, 31], "auc": [2, 11, 21, 22, 25, 27, 30, 31], "16": [2, 5, 6, 8, 9, 10, 11, 12, 19, 25, 27, 28, 32, 37, 38, 39, 44, 45, 46], "305m": 2, "6932": [2, 25, 27], "5005": 2, "6931": [2, 25, 27], "val_auc": [2, 11, 25, 27, 31], "5029": 2, "0x7fd449398a30": 2, "back": [2, 23, 25, 40], "cell": [2, 3, 5, 18, 25, 27], "predicttensorflow": [2, 3, 5, 6, 11, 25, 27, 33], "transformworkflow": [2, 3, 5, 6, 11, 25, 27, 33, 35, 36], "get_subworkflow": [2, 3], "item_block": 2, "item_embed": [2, 3], "fit_transform": [2, 5, 6, 35, 36], "to_ddf": [2, 25, 27], "tail": [2, 20], "output_1": [2, 3], "453": 2, "945": 2, "012117806822061539": 2, "02241620607674122": 2, "454": 2, "948": 2, "455": 2, "956": 2, "456": [2, 6], "1437": 2, "457": 2, "1469": 2, "timedelta": 2, "entiti": [2, 3], "field": [2, 27, 35, 36], "featureview": [2, 3], "valuetyp": 2, "type": [2, 8, 12, 13, 17, 19, 23, 25, 27, 29, 40], "infra": 2, "offline_stor": 2, "file_sourc": 2, "filesourc": 2, "timestamp_field": 2, "created_timestamp_column": 2, "value_typ": 2, "join_kei": 2, "user_features_view": 2, "ttl": 2, "onlin": [2, 6, 23, 27, 40], "dict": [2, 5, 6, 10, 27, 32], "format": [2, 7, 8, 9, 20, 23, 27, 28, 29, 32], "close": [2, 21], "item_features_view": 2, "checkout": [2, 5], "structur": [2, 3, 4, 25, 28], "seedir": [2, 3, 25], "satisfi": [2, 25, 27], "natsort": [2, 25], "sd": [2, 3, 25], "style": [2, 3, 25], "itemlimit": [2, 3, 25], "depthlimit": [2, 3, 25], "exclude_fold": [2, 3, 25], "ipynb_checkpoint": [2, 3, 25], "sort": [2, 3, 5, 8, 10, 22, 25, 31, 32, 33, 40], "readm": [2, 3, 19], "md": [2, 3, 13], "__init__": [2, 3, 12, 13, 27, 31, 32], "__pycache__": [2, 3], "cpython": [2, 32], "pyc": [2, 32], "test_workflow": [2, 3], "feature_stor": [2, 3], "yaml": [2, 3, 22], "ti": [2, 3, 5, 6, 11, 25, 33, 35, 36], "ensembl": [2, 4, 5, 6, 11, 12, 13, 15, 19, 27], "result": [2, 7, 8, 10, 12, 17, 19, 20, 27, 35, 36, 40], "given": [2, 3, 5, 6, 7, 10, 11, 13, 19, 20, 21, 23, 35, 36, 40], "request": [2, 3, 12, 13, 15, 17, 19, 25, 31, 32, 33, 35, 36, 40], "user": [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 19, 20, 23, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40], "point": [3, 9, 10, 27, 29], "reach": 3, "expect": [3, 8, 20, 23, 25, 35, 36], "execut": [3, 4, 5, 9, 15, 17, 18, 19, 22, 24, 25, 27, 29, 30, 33, 35, 36], "01": [3, 7, 10, 11, 22, 27, 29, 31, 32, 37, 38, 39, 44, 45, 46], "queri": [3, 17, 19, 20, 27, 35, 36, 40], "describ": [3, 18, 20, 21, 23, 27, 29], "built": [3, 5, 6, 17, 21, 24, 27], "usag": [3, 16], "assum": [3, 22, 23, 26, 27, 31], "feast": [3, 4, 15], "script": [3, 9, 12, 15, 16, 17, 18, 19, 21, 22, 32], "numpi": [3, 5, 6, 8, 11, 12, 19, 25, 27, 28, 29, 32, 35, 36], "np": [3, 5, 6, 8, 11, 12, 19, 25, 27, 28, 29, 32, 35, 36], "panda": [3, 6, 7, 8, 12, 13, 19, 25, 27, 35, 36], "pd": [3, 6, 12, 19], "columnschema": [3, 6, 13, 25], "softmax_sampl": 3, "softmaxsampl": 3, "unroll_featur": 3, "unrollfeatur": 3, "send_triton_request": [3, 19], "metadata": [3, 7, 8, 10, 11, 17, 25, 27, 32, 35, 36], "orc_column_statistics_pb2": 3, "deprecationwarn": [3, 12, 13], "deprec": [3, 5, 8, 12, 13, 27, 28], "filedescriptor": 3, "note": [3, 5, 9, 10, 12, 13, 17, 18, 19, 27, 28, 35, 36, 37, 38, 39, 44, 45, 46], "unlink": 3, "descriptor": 3, "awai": [3, 8, 25], "descriptor_pool": 3, "_descriptor": 3, "fielddescriptor": 3, "30": [3, 5, 12, 17, 20, 25, 27], "_integerstatist": 3, "56": [3, 30], "885234": 3, "tensor_shape_pb2": 3, "36": [3, 6, 25, 32], "_tensorshapeproto_dim": 3, "types_pb2": 3, "33": [3, 37, 38, 39, 44, 45, 46], "enumvaluedescriptor": 3, "27": [3, 27, 28, 29, 32, 36, 37, 38, 39, 44, 45, 46], "enumdescriptor": 3, "_datatyp": 3, "287": 3, "280": 3, "_serializeddtyp": 3, "resource_handle_pb2": 3, "20": [3, 6, 8, 11, 12, 27, 28, 31, 36, 37, 38, 39, 44, 45, 46], "39": [3, 6, 11, 32, 36, 38, 45], "futur": [3, 5, 8, 12, 13, 20, 27, 28, 33], "merlinload": 3, "registri": 3, "central": 3, "definit": [3, 25], "relat": [3, 21], "organ": [3, 5, 6, 23], "those": [3, 16, 17, 20, 23, 25, 35, 36], "both": [3, 6, 8, 20, 25, 40], "tell": 3, "command": [3, 4, 5, 6, 9, 12, 13, 15, 17, 19, 22, 24, 25, 26, 27, 31, 32, 33, 35, 36], "direct": [3, 8], "view": [3, 4, 12, 15, 22, 24, 25, 40, 43], "sqlite": 3, "online_stor": 3, "db": 3, "feast_repo_path": 3, "32muser_id": 3, "32mitem_id": 3, "32muser_featur": 3, "32mitem_featur": 3, "32mfeast_repo_item_featur": 3, "32mfeast_repo_user_featur": 3, "materi": [3, 8], "done": [3, 19, 25, 30], "keep": [3, 17, 20, 21, 23, 29], "date": [3, 20, 27], "As": [3, 8, 9, 10, 17, 21, 22, 30, 32, 33, 40], "continu": [3, 5, 8, 10, 12, 20, 23, 25, 27, 29, 32], "1995": [3, 7, 36], "2025": 3, "some": [3, 6, 17, 19, 20, 21, 22, 23, 25, 26, 27, 29, 31, 35, 36, 40], "01t01": 3, "32m2": 3, "32m1995": 3, "32m2025": 3, "32msqlite": 3, "100": [3, 5, 7, 9, 17, 22, 23, 29, 30, 35, 36], "460": [3, 37, 38, 39, 44, 45, 46], "2521": 3, "27it": 3, "458": 3, "3335": 3, "12it": 3, "again": [3, 27], "ran": [3, 19, 21], "isdir": [3, 29], "faiss_index": 3, "makedir": [3, 11, 28, 29], "faiss_index_path": 3, "retrieval_model_path": 3, "ranking_model_path": 3, "dlrm": [3, 9, 17, 21, 22, 24, 27, 30, 33, 40], "queryfaiss": 3, "interfac": [3, 22], "nearbi": 3, "setup_faiss": 3, "l2": [3, 6, 17, 23], "distanc": [3, 20], "read_parquet": [3, 8, 12, 13, 19, 25, 27, 32, 33], "embedding_column": [3, 12], "centroid": 3, "least": [3, 5, 6, 11, 20, 27, 35, 36], "1248": 3, "featurestor": 3, "fetch": 3, "queryfeast": 3, "ensur": [3, 5, 6, 8, 11, 19, 20, 22, 25, 32, 33, 35, 36], "correctli": [3, 22, 25, 27], "tritonserv": [3, 5, 6, 12, 13, 18, 19, 25, 27, 32, 33, 35, 36], "user_attribut": 3, "from_feature_view": 3, "include_id": 3, "32m1": 3, "32m2023": 3, "51": [3, 5, 25, 27, 37, 38, 39, 44, 45, 46], "0it": 3, "user_subgraph": 3, "backend": [3, 12, 13, 32, 33, 40], "prevent": 3, "claim": [3, 27], "tf_util": [3, 5, 6], "configure_tensorflow": 3, "dlpack": 3, "from_dlpack": 3, "dlcapsul": 3, "topk_retriev": 3, "topk": [3, 40], "07": [3, 25, 35, 37, 38, 39, 44, 45, 46], "269579": 3, "430459": 3, "431356": 3, "absl": [3, 5, 6, 11, 25, 27, 31], "untrac": [3, 5, 6, 11, 25, 27, 31], "restored_function_bodi": 3, "save": [3, 4, 5, 6, 7, 8, 9, 12, 13, 17, 18, 19, 20, 21, 22, 23, 27, 29, 32], "show": [3, 4, 5, 6, 7, 11, 12, 15, 18, 19, 22, 23, 24, 25, 26, 27, 29, 31, 35, 36], "52": [3, 5, 8, 10, 27, 28, 29, 37, 39, 44, 46], "callabl": [3, 5, 6, 11, 25, 27, 31], "asset": [3, 5, 6, 11, 23, 25, 27, 31], "written": [3, 5, 6, 9, 11, 25, 30, 31, 35, 36], "tmp": [3, 5, 6, 11, 22, 25, 27, 32], "tmpdalflmaz": 3, "item_attribut": 3, "candidate_id": 3, "output_prefix": 3, "14": [3, 5, 25, 27, 28, 29, 32, 37, 38, 39, 44, 45, 46], "item_subgraph": 3, "merg": [3, 5, 6], "were": [3, 5, 17, 19, 20, 21, 23, 43], "unrol": 3, "broadcast": 3, "user_features_to_unrol": 3, "combined_featur": 3, "feed": [3, 6, 8, 10, 11, 12, 25], "98": [3, 22, 25], "tmpqdd_jn5e": 3, "descend": 3, "random": [3, 5, 7, 17, 20, 23, 29], "sampl": [3, 4, 5, 7, 10, 15, 20, 25, 28, 29, 31, 35, 40], "softmax": [3, 17, 23], "top_k": 3, "relevance_col": 3, "binary_classification_task": [3, 27], "temperatur": [3, 5, 6, 23], "00000001": 3, "last": [3, 5, 13, 18, 20, 25, 28, 29, 32, 33, 35, 36, 40], "artifact": [3, 17, 18, 33], "consum": [3, 6, 25], "class": [3, 8, 17, 18, 20, 21, 25], "entir": [3, 23, 25], "It": [3, 5, 6, 8, 9, 10, 11, 16, 17, 20, 21, 22, 23, 25, 27, 28, 29, 30, 31, 33, 35, 36, 40], "interpret": [3, 18, 25], "correct": [3, 6, 13, 18, 25, 32], "suppli": [3, 25], "list": [3, 5, 6, 7, 8, 9, 12, 19, 20, 22, 23, 25, 29, 40], "poc_ensembl": 3, "send": [3, 5, 6, 12, 13, 19, 25, 32, 33, 35, 36], "request_schema": [3, 13], "export_path": [3, 5, 6, 11, 25, 33, 35, 36], "ens_config": 3, "node_config": 3, "output_schema": [3, 5, 6, 8, 25, 27, 33, 35, 36], "ordered_id": 3, "ordered_scor": 3, "0_transformworkflowtriton": [3, 19, 25], "categori": [3, 5, 8, 20, 25, 27, 29, 32], "uniqu": [3, 5, 8, 25, 27, 32, 35], "json": [3, 8, 9, 12, 13, 19, 25, 27, 30, 32], "pkl": [3, 25, 27, 32], "1_predicttensorflowtriton": [3, 5, 6, 19, 25], "savedmodel": [3, 5, 6, 11, 25, 27, 31], "fingerprint": [3, 25], "pb": [3, 25, 27], "keras_metadata": [3, 25, 27], "saved_model": [3, 5, 6, 23, 25, 27], "00000": [3, 25, 27], "00001": [3, 25, 27], "2_transformworkflowtriton": 3, "3_predicttensorflowtriton": 3, "input_schema": [3, 5, 6, 19, 25, 27, 32, 33, 35, 36], "executor_model": [3, 5, 6, 19, 25, 27, 33, 35, 36], "termin": [3, 4, 5, 6, 9, 15, 18, 19, 24, 25, 32, 35, 36], "ensemble_export_path": [3, 18, 19, 35, 36], "argument": [3, 5, 6, 8, 17, 19, 25, 27, 28, 29, 32, 33, 35, 36], "specifi": [3, 4, 5, 6, 8, 18, 19, 20, 22, 25, 27, 28, 29, 30, 32, 33, 35, 36], "previous": [3, 5, 6, 18, 19, 25, 32, 33, 35, 36], "method": [3, 5, 6, 20, 21, 22, 23, 25, 27, 29, 32, 33, 35, 36], "statu": [3, 12, 13, 19, 25, 27, 29], "dispatch": [3, 5, 7, 8, 13, 25, 27, 32, 33], "make_df": 3, "sent": [3, 17, 25, 27], "per": [3, 7, 8, 9, 21, 23, 25, 40], "arrai": [3, 5, 6, 12, 19], "168": 3, "324": 3, "79": [3, 6, 27], "361": 3, "294": 3, "267": [3, 27], "289": 3, "397": 3, "189": [3, 6], "5016385": 3, "50176895": 3, "5017176": 3, "5024097": 3, "5018236": 3, "5018286": 3, "50162375": 3, "5015677": 3, "50175667": 3, "5014358": 3, "float32": [3, 5, 6, 8, 11, 12, 19, 20, 29, 31], "That": [3, 17, 20, 22, 23, 29], "two": [4, 5, 8, 10, 11, 21, 23, 25, 32, 40], "merlin": [4, 5, 6, 7, 8, 9, 11, 12, 13, 18, 20, 21, 22, 23, 28, 29, 30, 32, 34, 37, 38, 39, 40, 41], "retriev": [4, 15, 16, 20, 23, 33, 41], "export": [4, 5, 6, 11, 17, 18, 19, 27, 33, 35, 36], "index": [4, 10, 12, 13, 15, 19, 25, 27, 29, 40], "pull": [4, 15, 17, 22, 24, 27], "nvcr": [4, 17, 22, 24, 27], "nightli": 4, "p": [4, 9, 10, 12, 15, 17, 24, 26, 29, 30, 32], "8001": [4, 5, 6, 13, 18, 19, 25, 32, 33, 35, 36], "8002": [4, 5, 6, 18, 19, 25, 35, 36], "8888": [4, 15, 24, 26], "v": [4, 17, 22, 26, 27], "ipc": [4, 15, 17, 22, 24, 26], "host": [4, 13, 15, 17, 22, 24, 26, 27, 29, 36], "bin": [4, 15, 17, 20, 22, 23, 24, 26, 27], "bash": [4, 15, 17, 22, 24, 26, 27], "shell": [4, 15, 24, 27], "complet": [4, 15, 18, 19, 24, 25, 27], "prompt": [4, 15, 22, 24], "root": [4, 12, 15, 24, 26, 27], "2efa5b50b909": [4, 15, 24], "jupyterlab": [4, 15, 24], "lab": [4, 15, 24], "allow": [4, 8, 11, 15, 16, 17, 20, 21, 22, 23, 24], "ip": [4, 13, 15, 24], "notebookapp": 4, "token": [4, 15, 24], "password": [4, 27], "Or": [4, 15, 24], "past": [4, 15, 20, 22, 24], "9b537d1fda9e4e9cadc673ba2a472e247deee69a6229ff8d": [4, 15, 24], "127": [4, 15, 24, 29, 36], "browser": [4, 15, 22, 24], "2022": [5, 6, 19, 27, 28, 29, 31, 33, 35, 36], "requir": [5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 18, 19, 20, 22, 23, 25, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 44, 45, 46], "anda": [5, 6], "book": 5, "166": 5, "835": 5, "anonym": [5, 40], "hotel": [5, 20], "378": [5, 6], "667": 5, "part": [5, 7, 8, 18, 19, 23, 25], "trip": 5, "utrip_id": 5, "consecut": 5, "reshap": [5, 6, 9], "full": [5, 6, 17, 22, 25, 27, 29, 35, 36], "itinerari": 5, "chronolog": [5, 6], "goal": [5, 6, 22, 25, 40], "city_id": 5, "import": [5, 6, 7, 8, 9, 10, 11, 12, 13, 17, 19, 20, 22, 27, 28, 29, 30, 31, 32, 33, 35, 36, 40], "statement": 5, "get_lib": [5, 7, 8, 13, 25, 27, 32, 33], "get_book": 5, "reset": 5, "alloc": [5, 11, 23, 25, 31], "timeit": 5, "input_data_dir": [5, 7, 8, 9, 10, 11, 12, 13, 28, 29, 31], "output_data_dir": [5, 6, 27, 28, 29, 30, 32, 35, 36], "num_epoch": [5, 6], "25": [5, 6, 25, 27, 28, 36, 37, 38, 39, 44, 45, 46], "697025": 5, "26": [5, 11, 25, 27, 28, 30, 31, 32, 37, 44], "988036": 5, "xla": [5, 6], "stream_executor": [5, 6, 11, 27, 35, 36], "cuda": [5, 6, 9, 10, 11, 25, 27, 29, 30, 31, 35, 36, 37, 38, 39, 44, 45, 46], "cuda_gpu_executor": [5, 6, 11, 27, 35, 36], "998": [5, 6], "success": [5, 6, 11, 27, 35, 36], "numa": [5, 6, 11, 27, 30, 35, 36], "sysf": [5, 6, 11, 27, 35, 36], "had": [5, 6, 11, 21, 27, 35, 36], "neg": [5, 6, 11, 17, 20, 27, 35, 36, 40], "zero": [5, 6, 11, 27, 35, 36], "988386": 5, "988518": 5, "28": [5, 25, 27, 36, 37, 38, 39, 44, 45, 46], "519868": 5, "520815": 5, "520999": 5, "521129": 5, "591345": 5, "591534": 5, "591665": 5, "591770": 5, "591778": 5, "gpu_process_st": [5, 11, 25, 31], "222": [5, 11, 31], "malloc": [5, 11, 25, 31], "async": [5, 11, 25, 31], "591860": 5, "08": [5, 6, 11, 22, 27, 31, 36, 37, 38, 39, 44, 45, 46], "149": 5, "compound": [5, 8, 27], "atom": [5, 8, 27], "session_id": [5, 20, 40], "0x7fe90a7fce80": 5, "0x7fe90a7f7820": 5, "vist": 5, "sever": [5, 6, 19, 35, 36, 42], "destin": 5, "displai": [5, 9, 25, 30], "otherwis": [5, 8, 25, 27, 31], "hidden": 5, "read_csv": [5, 6, 7], "train_set": 5, "parse_d": 5, "checkin": 5, "device_class": 5, "affiliate_id": 5, "1000027": 5, "2016": 5, "13": [5, 6, 25, 27, 28, 29, 30, 31, 36, 37, 38, 39, 44, 45, 46], "8183": 5, "desktop": 5, "7168": 5, "15626": 5, "60902": 5, "30628": 5, "253": 5, "1000033": 5, "04": [5, 7, 12, 22, 25, 28, 31, 37, 38, 39, 44, 45, 46], "09": [5, 22, 25, 27, 28, 35, 37, 38, 39, 44, 45, 46], "38677": 5, "mobil": 5, "359": 5, "booker_countri": 5, "hotel_countri": 5, "elbonia": 5, "gondal": 5, "1000027_1": 5, "cobra": 5, "island": 5, "1000033_1": 5, "sequenc": [5, 21, 40], "attempt": [5, 27], "hop": 5, "journei": [5, 6], "length": [5, 12, 13, 19, 23, 25, 27], "batch": [5, 6, 9, 10, 13, 17, 19, 20, 21, 23, 25, 27, 32, 33], "box": [5, 8, 22], "doesn": [5, 6, 20, 27, 31, 32], "t": [5, 6, 8, 9, 12, 13, 17, 20, 21, 27, 28, 29, 31, 32], "thank": 5, "pad": 5, "trim": 5, "particular": [5, 6, 8, 17, 21, 23], "effect": 5, "mask": [5, 6, 40], "discard": 5, "shorter": 5, "sens": [5, 21, 23], "noth": 5, "could": [5, 20, 21, 23, 29, 40], "present": [5, 21], "begin": [5, 6, 25, 35, 36, 43], "split": [5, 17, 19, 21, 23, 29], "mani": [5, 7, 10, 11, 15, 17, 20, 21, 23, 40], "shuffl": [5, 6, 7, 8, 23, 29, 40], "along": [5, 6, 11, 35, 36], "frac": 5, "number": [5, 7, 8, 10, 17, 20, 21, 22, 23, 25, 28], "len": [5, 6, 7, 27, 29], "217686": 5, "assign": [5, 8, 20, 23], "furthermor": 5, "proper": 5, "citi": [5, 20], "thei": [5, 7, 8, 16, 18, 20, 21, 22, 23, 28, 40], "cannot": [5, 25, 30], "from_panda": 5, "to_panda": 5, "groupbi": [5, 6], "num_exampl": [5, 6], "train_set_utrip_id": 5, "shape": [5, 10, 11, 12, 13, 27, 31, 32, 33], "validation_set_utrip_id": 5, "isin": 5, "sort_valu": [5, 35], "validation_set": 5, "too": [5, 8], "short": 5, "calcul": [5, 8, 10, 29, 40], "total": [5, 8, 10, 12, 13, 25, 28, 29, 31], "schema_select_by_tag": 5, "addtag": [5, 8, 36], "train_set_dataset": 5, "validation_set_dataset": 5, "weekday_checkin": 5, "to_datetim": 5, "dt": 5, "weekdai": 5, "weekday_checkout": 5, "categorical_featur": [5, 17, 20], "groupby_featur": [5, 6], "groupby_col": [5, 6], "agg": [5, 6], "count": [5, 6, 12, 20, 22], "sort_col": [5, 6], "list_featur": 5, "city_id_list": 5, "booker_country_list": 5, "hotel_country_list": 5, "weekday_checkin_list": 5, "weekday_checkout_list": 5, "less": [5, 17, 18, 20, 23, 28, 29, 40], "minimum_session_length": [5, 6], "city_id_count": 5, "filtered_sess": [5, 6], "wf": [5, 6], "train_process": 5, "validation_process": 5, "integ": [5, 8, 25, 29, 32], "8238": 5, "156": [5, 6], "2278": 5, "2097": 5, "63": 5, "1160": 5, "87": [5, 27], "618": 5, "1050": 5, "65": [5, 27], "1032": 5, "757": 5, "140": [5, 17], "3603": 5, "262": [5, 36], "662": 5, "250": 5, "seq_schema": 5, "is_list": [5, 8, 19, 25], "is_rag": [5, 8, 19, 25], "properti": [5, 8, 10, 20, 25], "num_bucket": [5, 8, 25, 29], "freq_threshold": [5, 8, 25, 36], "max_siz": [5, 8, 25, 29], "start_index": [5, 8], "cat_path": [5, 8, 25], "domain": [5, 8, 25, 40, 41], "min": [5, 8, 17, 19, 20, 22, 25], "max": [5, 8, 17, 20, 22, 25], "embedding_s": [5, 8, 10, 25, 31], "cardin": [5, 8, 10, 25], "dimens": [5, 8, 10, 17, 23, 25], "value_count": 5, "int64": [5, 6, 8, 9, 12, 13, 25, 29], "element_typ": [5, 8, 19, 25], "lt": [5, 8, 19, 25, 37, 38, 39, 44, 45, 46], "elementtyp": [5, 8, 19, 25], "none": [5, 6, 11, 12, 13, 19, 20, 25, 27, 28, 29, 31, 32, 33], "37202": 5, "37203": 5, "512": [5, 8, 10, 25, 27, 30], "195": 5, "variou": [5, 32], "hyperparamet": [5, 16, 23, 27, 31], "portion": 5, "xlnet": [5, 40], "masking_prob": [5, 6], "paramet": [5, 6, 9, 10, 12, 20, 22, 23, 27, 29, 30, 35, 36], "timestep": 5, "probabl": [5, 13, 19, 24, 27, 32, 40], "summar": [5, 21, 40], "sequencemaskrandom": [5, 6], "pre": [5, 6, 23, 40], "randomli": [5, 10, 20, 23], "necessari": [5, 17, 18, 20], "configur": [5, 6, 11, 12, 17, 18, 21, 22, 23, 25, 26, 27, 29, 30, 32], "huggingfac": [5, 40], "its": [5, 6, 8, 9, 10, 11, 12, 17, 20, 22, 23, 27, 29, 30], "replac": [5, 6, 7, 12, 13, 20, 23, 28], "dummi": 5, "trainabl": [5, 23], "leakag": 5, "inputblockv2": [5, 6], "xlnetblock": [5, 6], "categoricaloutput": [5, 6], "select_by_nam": [5, 6, 10], "sequencemasklast": 5, "dmodel": [5, 6], "48": [5, 9, 28], "mlp_block": [5, 6], "activ": [5, 6, 10, 20, 22, 23, 35, 36], "relu": [5, 6, 9, 10, 22, 23, 30], "transformer_block": [5, 6], "d_model": [5, 6], "n_head": [5, 6], "n_layer": [5, 6], "sequence_combin": [5, 6], "default_loss": [5, 6], "44": [5, 19, 39, 46], "034041": 5, "cuda_dnn": [5, 6], "428": [5, 6], "cudnn": [5, 6, 30, 37, 38, 39, 44, 45, 46], "8700": [5, 6], "gradient": [5, 6], "mask_emb": [5, 6], "layer_": [5, 6], "_0": [5, 6, 25], "rel_attn": [5, 6], "r_s_bia": [5, 6], "seg_emb": [5, 6], "_1": [5, 6], "minim": [5, 6], "forget": [5, 6, 27], "54": [5, 27, 31], "541024": 5, "grappler": [5, 6], "loop_optim": [5, 6], "907": [5, 6], "loop": [5, 6, 8, 29], "control": [5, 6, 12, 13, 21, 23, 27, 28], "xl_net_block": [5, 6], "sequential_block_5": 5, "replace_masked_embed": [5, 6], "raggedwher": [5, 6], "assert": [5, 6, 36], "assertguard": [5, 6], "branch_execut": [5, 6], "_95": [5, 6], "2720": [5, 30], "81": 5, "25m": [5, 7, 10, 11], "3315": 5, "1973": 5, "mrr_at_10": [5, 6], "0863": 5, "1123": 5, "map_at_10": [5, 6], "precision_at_10": [5, 6], "0197": 5, "3306": 5, "70": [5, 27, 35], "0979": 5, "3633": 5, "1707": 5, "2161": 5, "0363": 5, "0950": 5, "71": [5, 10], "26m": 5, "5827": 5, "4306": 5, "2056": 5, "2588": 5, "0431": 5, "72": 5, "3211": 5, "4627": 5, "2213": 5, "2784": 5, "0463": 5, "3194": 5, "1920": 5, "4787": 5, "2306": 5, "2892": 5, "0479": 5, "1903": 5, "0x7fe67105a7f0": 5, "But": [5, 6, 21, 23], "come": 5, "realiti": 5, "care": [5, 11, 18], "mimic": 5, "scenario": [5, 23], "measur": [5, 8], "return_dict": [5, 31], "968982": 5, "_74": 5, "340": 5, "20m": [5, 27], "7151": 5, "5533": 5, "3083": 5, "3665": 5, "0553": 5, "7149": 5, "715089797973633": 5, "5533444881439209": 5, "30831339955329895": 5, "36654922366142273": 5, "055334459990262985": 5, "635858535766602": 5, "deliv": [5, 18, 19], "fast": [5, 8, 18, 19], "ai": [5, 18, 19, 22, 23, 25], "inf_op": [5, 35, 36], "serial": [5, 6, 11, 28, 31], "tfsharedembed": [5, 6], "_feature_shap": [5, 6], "tensorshap": [5, 6], "_feature_dtyp": [5, 6], "dropout": [5, 6, 10, 17, 21, 22, 23], "model_context_layer_call_fn": [5, 6, 11, 25, 27, 31], "model_context_layer_call_and_return_conditional_loss": [5, 6, 11, 25, 27, 31], "sequence_mask_random_layer_call_fn": [5, 6], "sequence_mask_random_layer_call_and_return_conditional_loss": [5, 6], "sequence_mask_last_layer_call_fn": 5, "108": [5, 29], "tmp1sakw940": 5, "101": [5, 6], "custommaskwarn": [5, 6], "get_config": [5, 6], "pass": [5, 6, 11, 27], "custom_object": [5, 6], "serialize_keras_object": [5, 6], "maybe_valu": [5, 6], "288": [5, 6], "legaci": [5, 6, 11], "layer_seri": [5, 6], "134": [5, 6], "obj": [5, 6], "manual": [5, 6, 11, 25, 27], "produc": [5, 17, 23, 27], "downstream": 5, "wait": [5, 6, 7, 18, 19, 25, 35, 36], "until": [5, 6, 18, 19, 22, 25, 35, 36], "i0414": [5, 6, 18, 19, 25, 35, 36], "741833": [5, 6, 18, 19, 25, 35, 36], "4067": [5, 6, 18, 19, 25, 35, 36], "grpc_server": [5, 6, 18, 19, 25, 35, 36], "4421": [5, 6, 18, 19, 25, 35, 36], "grpcinferenceservic": [5, 6, 18, 19, 25, 35, 36], "742197": [5, 6, 18, 19, 25, 35, 36], "http_server": [5, 6, 18, 19, 25, 35, 36], "3113": [5, 6, 18, 19, 25, 35, 36], "httpservic": [5, 6, 18, 19, 25, 35, 36], "783470": [5, 6, 18, 19, 25, 35, 36], "178": [5, 6, 18, 19, 25, 35, 36], "servic": [5, 6, 18, 19, 20, 23, 25, 35, 36], "correspond": [5, 6, 8, 20, 25, 35, 36], "subsequent": [5, 6], "stop": [5, 10], "convert_df_to_triton_input": [5, 6, 13, 25, 27, 32, 33, 35, 36], "iloc": [5, 6, 12, 19], "tritoncli": [5, 6, 12, 13, 19, 25, 27, 32, 33, 35, 36], "grpc": [5, 6, 12, 13, 19, 25, 32, 33, 35, 36], "grpcclient": [5, 6, 13, 19, 25, 32, 33, 35, 36], "inferenceservercli": [5, 6, 12, 13, 19, 25, 27, 32, 33, 35, 36], "logit": [5, 6], "as_numpi": [5, 6, 12, 13, 25, 27, 33, 35, 36], "categorical_output": [5, 6], "8206294": 5, "3849059": 5, "9042726": 5, "851537": 5, "4237087": 5, "73849726": 5, "anoth": [5, 6, 23, 25, 29, 33, 40], "goe": [5, 6], "deeper": [5, 6], "cover": [5, 6], "weight": [5, 6, 12, 17, 21, 32], "ty": [5, 6], "fashion": [5, 6, 40], "e": [5, 6, 12, 13, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 29, 40], "commerc": [5, 6, 20, 23, 40], "chose": 6, "sigir": 6, "ecom": 6, "2021": [6, 7, 8, 9, 10, 11, 12, 25, 28, 29, 30, 31, 32], "36m": 6, "event": [6, 20, 21, 23], "brows": 6, "action": 6, "taken": [6, 11, 18, 21, 35], "22": [6, 7, 8, 9, 10, 11, 12, 27, 35, 36, 43], "58": [6, 27], "667322": 6, "026020": 6, "026445": 6, "026622": 6, "place": [6, 17, 29], "alongsid": 6, "sigir_dataset": 6, "extract": [6, 10, 20, 23, 28], "By": [6, 8, 20, 23, 28], "boolean": 6, "run_on_synthetic_data": 6, "deal": [6, 17, 21], "nan": [6, 8, 20], "pars": [6, 25, 27], "string": [6, 7, 8, 20, 28], "browsing_train": 6, "sku": 6, "sku_to_cont": 6, "description_vector": 6, "image_vector": 6, "x": [6, 11, 12, 21, 27, 28, 29, 32], "els": [6, 12, 29], "eval": [6, 17, 19, 22, 23, 25, 27, 30], "counterpart": 6, "further": 6, "product_sku_hash": 6, "essenti": 6, "descript": 6, "piec": 6, "sure": [6, 8, 18, 27, 29], "addition": [6, 11, 26], "group": [6, 20, 21, 40], "session_id_hash": 6, "arrang": 6, "industri": 6, "consult": [6, 9, 10, 35, 36], "retain": 6, "cat_op": 6, "event_typ": 6, "product_act": 6, "hashed_url": 6, "server_timestamp_epoch_m": 6, "normalizeminmax": 6, "hashed_url_count": 6, "won": 6, "nor": 6, "longer": 6, "product_sku_hash_list": 6, "event_type_list": 6, "product_action_list": 6, "hashed_url_list": 6, "train_transform": [6, 35, 36], "coupl": 6, "578": 6, "972": 6, "420": 6, "328": 6, "126": 6, "233": 6, "925": 6, "410": 6, "766": 6, "955": 6, "745": 6, "210": 6, "940": 6, "688": 6, "986": 6, "524": 6, "425": 6, "298": 6, "304": [6, 37, 38, 39, 44, 45, 46], "393": 6, "697": 6, "706": 6, "313": 6, "834": 6, "83": [6, 27, 28, 37, 44], "502": 6, "221": 6, "915": 6, "658": 6, "802": 6, "180": [6, 21, 36], "580": 6, "702": 6, "339": 6, "645": 6, "436": 6, "358": 6, "84": [6, 36, 37, 38, 39, 44, 45, 46], "35": [6, 27, 28, 38, 45], "271": 6, "562": [6, 30], "498": [6, 27], "172": 6, "239": 6, "270": 6, "215": 6, "489": [6, 27], "278": 6, "153": 6, "717": 6, "540": [6, 30], "219": 6, "200": [6, 9, 12, 13, 19, 21, 22, 25, 27], "169": [6, 27], "419": 6, "875": 6, "725": 6, "926": 6, "770": 6, "160": [6, 9], "554": 6, "763": 6, "922": 6, "914": 6, "592": [6, 27], "842": 6, "916": 6, "137": 6, "928": 6, "615": 6, "318": 6, "506": 6, "281": 6, "191": [6, 19, 25], "480": 6, "965": 6, "399": 6, "761": 6, "preserv": 6, "grab": 6, "just": [6, 17, 20, 23, 29], "category_hash": 6, "price_bucket": 6, "07939800762120258": 6, "3465797761609977": 6, "186690": 6, "4275482879608162": 6, "30569476366666": 6, "1440": 6, "951997": 6, "31035419787213536": 6, "18070481533058008": 6, "973384": 6, "31319783485940356": 6, "11623980504981396": 6, "138": 6, "146260": 6, "25091279302969943": 6, "33473442518442525": 6, "119": 6, "808252": 6, "wf_sku": 6, "skus_d": 6, "836": 6, "979": 6, "469": 6, "118": 6, "emebed": 6, "to_npi": 6, "npy": 6, "know": [6, 25, 29, 33], "associ": [6, 10], "07939801": 6, "34657978": 6, "38269496": 6, "56307004": 6, "10142923": 6, "03702352": 6, "11606304": 6, "10070879": 6, "21879928": 6, "06107687": 6, "20743195": 6, "01330719": 6, "60182867": 6, "0920322": 6, "2648726": 6, "56061561": 6, "48643498": 6, "39045152": 6, "40012162": 6, "09153962": 6, "38351605": 6, "57134731": 6, "59986226": 6, "40321368": 6, "32984972": 6, "37559494": 6, "1554353": 6, "0413067": 6, "33814398": 6, "30678041": 6, "24001132": 6, "42737922": 6, "41554601": 6, "40451691": 6, "50428902": 6, "2004803": 6, "38297056": 6, "06580838": 6, "48285745": 6, "51406472": 6, "02268894": 6, "36343324": 6, "32497967": 6, "29736346": 6, "00538915": 6, "12329302": 6, "04998194": 6, "27843002": 6, "20212714": 6, "39019503": 6, "embeddingoper": 6, "lookup": 6, "hardwar": [6, 11, 29], "eas": 6, "npartit": 6, "part_mem_fract": [6, 28, 31], "broad": [6, 23], "lot": [6, 28], "permit": 6, "significantli": [6, 40], "increas": [6, 23], "embedding_oper": 6, "id_lookup_t": 6, "lookup_kei": 6, "embedding_nam": 6, "product_embed": 6, "835162": 6, "836068": 6, "836268": 6, "836425": 6, "836673": 6, "836849": 6, "837009": 6, "837114": 6, "837130": 6, "insruct": 6, "sample_batch": 6, "include_target": 6, "prepare_featur": 6, "dict_kei": 6, "input_block": 6, "pretrained_embed": 6, "pretrainedembed": 6, "normal": [6, 8, 11, 20, 27, 29], "norm": 6, "output_dim": 6, "amen": 6, "subsequ": [6, 11, 19], "everyth": [6, 11], "input_batch": 6, "remain": [6, 26, 28, 29, 35, 36], "itself": 6, "And": [6, 10, 23], "950175": 6, "59": 6, "285571": 6, "sequential_block_7": 6, "9800": 6, "0106": 6, "0033": 6, "0050": 6, "0011": 6, "9689": 6, "34": [6, 27], "9591": 6, "0031": 6, "0048": 6, "9363": 6, "9471": 6, "0107": 6, "0028": 6, "0046": 6, "9206": 6, "9398": 6, "0103": 6, "0030": 6, "0047": 6, "0010": 6, "9015": 6, "9375": 6, "0104": 6, "9095": 6, "0x7f55081d17c0": 6, "inference_oper": 6, "prepare_list_features_1_layer_call_fn": 6, "110": [6, 27], "tmpi3g8g7q7": 6, "five": 6, "2332087": 6, "1218574": 6, "390479": 6, "7735352": 6, "1954267": 6, "34523243": 6, "arriv": 6, "hash": [6, 20, 23], "predicted_hashed_url_id": 6, "argmax": 6, "academ": [7, 10, 11, 14, 15], "public": [7, 10, 11, 14, 15, 16, 28, 31, 40], "movi": [7, 8, 10, 11, 13, 35, 36], "rate": [7, 8, 9, 10, 11, 13, 17, 20, 21, 23, 35, 36, 40], "62": [7, 8, 10, 11, 37, 38, 39, 44, 45, 46], "000": [7, 10, 11], "162": [7, 10, 11, 25], "project": [7, 10, 11, 22, 23], "genr": [7, 8, 9, 10, 11, 12, 36], "improv": [7, 10, 11, 17, 22, 23, 40], "state": [7, 10, 19, 20, 25, 35, 36], "art": [7, 10], "hot": [7, 8, 10, 14, 15, 20, 21, 23], "download_fil": [7, 28], "df_lib": [7, 8, 13, 25, 27, 32, 33], "expandus": [7, 8, 9, 10, 11, 12, 13], "unzip": [7, 36], "grouplen": 7, "ml": [7, 18, 22, 24, 25, 27, 35, 36], "zip": [7, 17, 20, 27, 36], "262mb": 7, "18mb": 7, "00file": 7, "movieid": [7, 8, 9, 10, 11, 12, 13, 35, 36], "titl": [7, 36], "toi": [7, 8, 36], "stori": [7, 8, 36], "adventur": [7, 8], "anim": [7, 8], "children": [7, 8], "comedi": [7, 8], "fantasi": [7, 8], "jumanji": [7, 8], "grumpier": 7, "men": 7, "romanc": [7, 8], "exhal": 7, "drama": [7, 8], "father": [7, 8], "bride": [7, 8], "ii": [7, 8], "str": [7, 12, 13, 19, 25, 27, 28, 29, 32], "axi": [7, 21], "movies_convert": [7, 8], "userid": [7, 8, 9, 10, 11, 12, 13, 35, 36], "296": 7, "1147880044": 7, "306": 7, "1147868817": 7, "307": 7, "1147868828": 7, "665": 7, "1147878820": 7, "899": 7, "1147868510": 7, "train_df": 7, "num_valid": 7, "c": [8, 9, 22, 27, 28, 29, 30, 32], "One": [8, 10, 19, 20, 23, 29, 40], "vari": [8, 21, 22, 25], "often": [8, 15, 23, 40], "frequent": [8, 20, 21], "ones": [8, 20, 21, 40], "howev": [8, 33, 35, 36, 37, 38, 39, 40, 44, 45, 46], "better": [8, 17, 20, 21, 22, 23], "datapoint": [8, 10], "ad": [8, 20, 23, 24, 26, 27, 32, 41], "v0": [8, 43], "familiar": [8, 14, 15, 26, 27, 29, 30, 31], "joinextern": 8, "wise": 8, "shutil": [8, 29], "columngroup": 8, "overload": 8, "turn": [8, 27], "op1": 8, "op2": 8, "sound": 8, "complic": 8, "left": [8, 40], "road": 8, "categorical_column": [8, 12, 29], "label_column": [8, 12, 25, 27, 29, 32, 33], "acycl": 8, "visual": [8, 22, 40], "graphviz": 8, "contigu": [8, 20, 23, 29], "increment": 8, "fulfil": 8, "cat_featur": [8, 29], "int8": [8, 9, 11, 19], "mean": [8, 17, 20, 23, 25, 28, 32, 40], "deviat": 8, "account": [8, 21, 22, 23, 27], "manifest": 8, "matter": 8, "solv": [8, 40], "break": 8, "chunk": 8, "main": [8, 17, 30], "purpos": [8, 10, 18, 19, 35, 36], "reliabl": 8, "dask_cudf": [8, 17], "hood": 8, "decomposit": 8, "lazili": 8, "evalu": [8, 16, 17, 20, 22, 24], "handl": [8, 27], "global": [8, 20, 28, 30], "partit": [8, 17, 20], "instanti": [8, 29], "couldn": [8, 27, 29], "label": [8, 9, 12, 25, 27, 28, 29, 30, 32], "enforc": [8, 29], "dictionari": [8, 29], "give": [8, 10, 11, 17, 25, 27, 29, 40], "dict_dtyp": [8, 29], "train_dataset": [8, 10, 25, 27, 29], "valid_dataset": [8, 25, 27, 29], "ll": [8, 25, 29, 31], "scikit": 8, "148": [8, 27, 30], "692": 8, "m": [8, 10, 12, 13, 17, 20, 22, 23, 27, 29], "sy": [8, 10, 12, 13, 25, 27, 28, 29, 31], "354": 8, "wall": [8, 10, 12, 13, 25, 28, 29, 31], "0x7fb2d7fc42b0": 8, "clear": [8, 9, 32], "clean": [8, 29, 30], "rmtree": [8, 29], "_metadata": [8, 25], "per_partit": [8, 29], "cat": [8, 29, 32], "write_hugectr_keyset": 8, "speed": [8, 9, 10, 29, 30], "1e": [8, 10, 17, 22, 23, 27], "03": [8, 11, 22, 25, 27, 28, 30, 32, 36], "86": [8, 13, 35], "\u00b5": 8, "restor": [8, 29], "162541": 8, "162542": [8, 9, 10, 12], "56658": 8, "56659": [8, 10], "glob": [8, 10, 11, 28, 31, 32, 33], "train_path": [8, 10, 25, 27, 29], "valid_path": [8, 10, 25, 27, 29], "part_0": [8, 12, 25, 31], "1691": [8, 10], "332": [8, 10], "1001": [8, 10], "154": [8, 10], "967": [8, 10], "245": [8, 10], "150851": 8, "622": 8, "39553": 8, "1146": 8, "benefit": [9, 17, 21], "basic": [9, 14, 15, 20, 21, 22, 26, 27], "object": [9, 20, 25], "adopt": 9, "estim": [9, 20, 27, 30], "ctr": [9, 20, 30], "experienc": [9, 30], "dgx": [9, 30, 37, 38, 39, 44, 45, 46], "a100": [9, 29, 30, 37, 38, 39, 44, 45, 46], "prove": [9, 30], "fastest": [9, 30], "commerci": [9, 30], "parallel": [9, 11, 21, 22, 30], "dcn": [9, 11, 17, 21, 30, 40], "deepfm": [9, 17, 21, 30], "profici": 9, "oversubscrib": 9, "don": [9, 17, 20, 21, 27], "prefetch": 9, "asynchron": 9, "multithread": 9, "via": [9, 10, 22, 30, 33], "output_dtyp": 9, "numer": [9, 20, 27, 28], "previou": [9, 10, 11, 21, 22, 23, 27, 29, 32, 33], "model_dir": [9, 11, 12, 27], "movielens_hugectr": [9, 12], "mkdir": [9, 29, 30], "dedic": 9, "commandlin": [9, 30], "train_hugectr": 9, "afterward": [9, 29, 32, 33], "writefil": [9, 27], "get_embedding_s": 9, "mpi4pi": [9, 30], "mpi": [9, 30], "solver": [9, 30], "createsolv": [9, 30], "vvgpu": [9, 30], "batchsiz": [9, 30], "2048": 9, "batchsize_ev": [9, 30], "max_eval_batch": [9, 30], "i64_input_kei": [9, 30], "use_mixed_precis": [9, 30], "repeat_dataset": [9, 30], "createoptim": [9, 30], "optimizer_typ": [9, 30], "optimizer_t": [9, 30], "reader": [9, 26, 27, 29, 30], "datareaderparam": [9, 30], "data_reader_typ": [9, 30], "datareadertype_t": [9, 30], "_file_list": [9, 25, 30], "txt": [9, 17, 22, 25, 30], "eval_sourc": [9, 30], "check_typ": [9, 30], "check_t": [9, 30], "non": [9, 13, 30], "slot_size_arrai": [9, 30], "56586": 9, "label_dim": [9, 12, 30], "label_nam": [9, 30], "dense_dim": [9, 30], "dense_nam": [9, 30], "data_reader_sparse_param_arrai": [9, 30], "datareadersparseparam": [9, 30], "data1": [9, 30], "nnz_per_slot": 9, "is_fixed_length": 9, "slot_num": 9, "sparseembed": [9, 30], "embedding_typ": [9, 30], "embedding_t": [9, 30], "localizedslotsparseembeddinghash": [9, 30], "workspace_size_per_gpu_in_mb": [9, 30], "embedding_vec_s": [9, 30], "sum": [9, 10, 30], "sparse_embedding_nam": [9, 30], "sparse_embedding1": [9, 30], "bottom_nam": [9, 30], "denselay": [9, 30], "layer_typ": [9, 30], "layer_t": [9, 30], "top_nam": [9, 30], "reshape1": 9, "leading_dim": 9, "innerproduct": [9, 30], "fc1": [9, 30], "num_output": [9, 30], "relu1": [9, 30], "fc2": [9, 30], "relu2": [9, 30], "fc3": [9, 30], "binarycrossentropyloss": [9, 30], "summari": [9, 30, 41], "max_it": [9, 30], "2000": 9, "eval_interv": [9, 30], "snapshot": [9, 30, 32], "1900": 9, "graph_to_json": [9, 30], "graph_config_fil": [9, 30], "overwrit": [9, 27], "l": [9, 29, 30, 31], "0_opt_sparse_1900": 9, "_dense_1900": [9, 12], "_opt_dense_1900": 9, "0_sparse_1900": [9, 12], "emb_vector": [9, 32], "slot_id": [9, 32], "ask": 9, "mv": [9, 32], "observ": [10, 21, 23], "slow": 10, "bottleneck": 10, "nativ": 10, "veri": [10, 17, 20, 23, 25], "experi": [10, 21, 23, 25, 35, 36, 40], "tutori": [10, 17, 21, 22, 35, 36], "explain": [10, 11, 18, 20, 21, 22, 23, 31, 40], "gc": 10, "framework_util": 10, "process_epoch": 10, "tensor": [10, 25, 30], "recogn": 10, "accordingli": [10, 20, 23], "validation_dataset": 10, "train_load": 10, "valid_load": 10, "32768": [10, 22], "usual": [10, 40], "tupl": 10, "actual": [10, 18, 23], "reason": [10, 17, 32], "second": [10, 27], "3rd": 10, "4th": 10, "5th": 10, "third": 10, "6th": 10, "88555": 10, "88556": 10, "88557": 10, "848": 10, "1847": 10, "5456": 10, "3095": 10, "1062": 10, "3705": 10, "flat": 10, "del": 10, "concaten": [10, 11, 17, 23, 40], "forward": [10, 27], "batchnorm": 10, "detail": [10, 15, 17, 21, 23, 26, 27, 31, 32, 33, 40], "captur": [10, 23], "def": [10, 27], "extract_info": 10, "col_nam": 10, "single_hot_embedding_tables_shap": 10, "mutli_hot_embedding_tables_shap": 10, "embedding_table_shap": 10, "num_continu": 10, "emb_dropout": 10, "layer_hidden_dim": 10, "layer_dropout_r": 10, "initial_cat_lay": 10, "concatenatedembed": 10, "embedding_lay": 10, "modulelist": 10, "inplac": 10, "mh_cat_lay": 10, "multihotembed": 10, "embeddingbag": 10, "mode": [10, 12, 13, 17, 27], "initial_cont_lay": 10, "batchnorm1d": 10, "ep": 10, "momentum": 10, "affin": 10, "track_running_stat": 10, "linear": [10, 20, 23], "in_featur": 10, "1040": 10, "out_featur": 10, "bia": 10, "output_lay": 10, "lr": [10, 11, 17, 21, 22, 23, 31], "train_loss": 10, "y_pred": 10, "y": [10, 21, 22, 27], "loss_func": 10, "nn": 10, "bcewithlogitsloss": 10, "valid_loss": 10, "02d": 10, "4f": 10, "610": 10, "152": 10, "5204": 10, "2798": 10, "17": [10, 27, 31, 36], "323": 10, "cross": [11, 23], "streamlin": 11, "thu": [11, 20], "despit": 11, "fairli": 11, "elabor": 11, "great": 11, "fullest": 11, "datalod": 11, "optmiz": 11, "maximum": [11, 23, 28], "example_util": [11, 25], "workflow_fit_transform": [11, 25], "save_result": 11, "373230": 11, "40": [11, 12, 25, 27, 28, 32, 36, 37, 38, 39, 44, 45, 46], "514976": 11, "996": 11, "515413": 11, "515595": 11, "727125": 11, "728141": 11, "728351": 11, "728508": 11, "41": [11, 25, 27, 36], "475009": 11, "475231": 11, "475394": 11, "475508": 11, "475570": 11, "1637": 11, "train_d": 11, "valid_d": 11, "am": 11, "modifi": [11, 35, 36], "remove_col": [11, 32], "extrem": 11, "facilit": 11, "carri": 11, "dcnmodel": [11, 23], "depth": 11, "deep_block": 11, "binaryoutput": [11, 25], "intern": 11, "v2": [11, 12, 13, 17, 19, 21, 25], "upon": 11, "explicit": [11, 12, 13, 17, 20, 28], "complementari": 11, "implicit": [11, 20, 24, 40], "overal": 11, "depict": 11, "stack": [11, 23], "embbed": 11, "emb": 11, "\ud835\udc5b": 11, "\ud835\udc65": 11, "imag": [11, 17, 20, 22, 23, 43], "paper": [11, 23, 40], "opt": [11, 27], "adagrad": [11, 23], "learning_r": [11, 31], "1221": 11, "6m": 11, "6609": 11, "5281": 11, "6588": 11, "5626": 11, "6537": 11, "0x7f7b5959ed30": 11, "serving_oper": [11, 25, 27, 33], "remove_by_tag": 11, "ens_conf": [11, 25, 33], "node_conf": [11, 25, 33], "unsupport": [11, 31], "signatur": [11, 31], "binary_output": [11, 13, 19, 25], "tensorspec": [11, 31], "sample_weight": [11, 31], "negative_candidate_id": 11, "func_graph": [11, 31], "unknownargu": [11, 31], "0x7f7b58732c40": 11, "_wrapped_model": [11, 31], "charact": [11, 20, 31], "train_compute_metr": [11, 27, 31], "dense_6_layer_call_fn": 11, "dense_6_layer_call_and_return_conditional_loss": 11, "47": [11, 25, 27, 37, 44], "tmp5eyrbewk": 11, "0_predicttensorflowtriton": [11, 13], "hugecr": 12, "movielen": [12, 13, 35, 36, 40], "file_to_writ": [12, 30], "max_batch_s": [12, 32], "de": 12, "data_typ": 12, "type_fp32": 12, "dim": [12, 21, 23], "catcolumn": 12, "type_int64": 12, "rowindex": 12, "type_int32": 12, "output0": [12, 32], "instance_group": [12, 27], "kind_gpu": 12, "string_valu": 12, "gpucach": [12, 32], "hit_rate_threshold": [12, 32], "gpucachep": [12, 32], "slot": [12, 32], "cat_feature_num": 12, "des_feature_num": 12, "max_nnz": [12, 32], "embedding_vector_s": [12, 32], "embeddingkey_long_typ": 12, "utf": [12, 30, 32], "dump": [12, 32], "supportlonglong": [12, 32], "sparse_fil": [12, 32], "dense_fil": [12, 32], "network_fil": [12, 32], "num_of_worker_buffer_in_pool": [12, 32], "num_of_refresher_buffer_in_pool": [12, 32], "cache_refresh_percentage_per_iter": [12, 32], "deployed_device_list": [12, 32], "default_value_for_each_t": [12, 32], "maxnum_catfeature_query_per_table_per_sampl": [12, 32], "56632": 12, "embedding_vecsize_per_t": [12, 32], "httpclient": [12, 27], "path_to_model": [12, 13], "sinc": [12, 13, 18, 19, 27], "wont": 12, "haven": 12, "yet": [12, 13], "filterwarn": [12, 13], "ignor": [12, 13, 20, 23, 25, 35, 36], "tritonhttpcli": [12, 13], "try": [12, 13, 15, 19, 24, 25, 33, 35, 40], "triton_cli": [12, 13, 19, 25], "verbos": [12, 13, 19, 25, 27], "channel": [12, 13, 19, 25], "fail": [12, 13, 19, 25], "is_server_l": [12, 13, 19, 25], "health": [12, 13, 19, 25], "live": [12, 13, 19, 25], "header": [12, 13, 17, 19, 25, 27], "httpsocketpoolrespons": [12, 13, 19, 25], "plain": [12, 13, 19, 25], "get_model_repository_index": [12, 13, 19, 25], "bytearrai": [12, 13, 19, 25], "b": [12, 13, 19, 21, 22, 25, 27], "load_model": [12, 13, 25, 33], "model_nam": [12, 13, 27], "99": [12, 17, 22, 23, 39, 46], "continuous_column": [12, 29], "emb_size_arrai": 12, "29434": 12, "shift": 12, "insert": 12, "cumsum": 12, "test_df": 12, "rp_lst": 12, "cur": 12, "append": 12, "hstack": 12, "dense_featur": 12, "row_ptr": 12, "inferinput": [12, 13, 25, 27, 32, 33, 35, 36], "np_to_triton_dtyp": 12, "set_data_from_numpi": 12, "inferrequestedoutput": [12, 13, 25, 32, 33, 35, 36], "request_id": [12, 13, 25, 32, 33], "get_respons": 12, "wdl2predict": 12, "1851": 12, "settingwithcopywarn": 12, "slice": 12, "loc": 12, "row_index": 12, "col_index": 12, "caveat": 12, "pydata": 12, "doc": [12, 17, 22, 27], "user_guid": 12, "versu": 12, "self": 12, "_setitem_single_column": 12, "val": 12, "pi": 12, "1773": 12, "model_vers": 12, "numsampl": 12, "deviceid": 12, "datatyp": [12, 27, 29], "fp32": 12, "binary_data_s": [12, 27], "5346206": 12, "49736455": 12, "2987379": 12, "6282493": 12, "7548654": 12, "59079504": 12, "55132014": 12, "90419775": 12, "47409508": 12, "5124942": 12, "2020": 13, "hasn": 13, "merlin_triton": 13, "deactiv": 13, "endpoint": [13, 33], "verifi": [13, 25, 35, 36], "aliv": 13, "ensemble_model": 13, "623": 13, "craft": 13, "9098613": 13, "59266": 13, "553": 13, "2556225": 13, "17037": 13, "608": 13, "20514728": 13, "133412": 13, "40583": 13, "decod": 13, "60931313": 13, "6242866": 13, "6068166": 13, "reflect": 13, "movielens25m": [14, 15], "teach": [14, 17], "becom": 14, "convert": [14, 20, 25, 27, 29, 30, 32, 33, 34], "adapt": 15, "primarili": 15, "largest": [15, 27, 28, 31, 34], "publicli": [15, 34], "1tb": [15, 29, 31, 34], "uncompress": [15, 17, 28, 29, 31, 34], "billion": [15, 28, 29, 31, 34, 40], "amazon": [15, 27], "sdk": 15, "ecr": [15, 27], "boto3": [15, 27], "8797": [15, 24, 26], "8787": [15, 24, 26, 29, 36], "8796": [15, 24, 26], "8786": [15, 24, 26], "guid": [16, 17, 19, 30, 40], "power": [16, 17, 20, 23], "best": [16, 17], "practic": [16, 17], "typic": [16, 17, 20, 23, 40], "scienc": [16, 17], "involv": [16, 17, 18], "tune": 16, "meant": [16, 25], "effort": [16, 18], "possibl": [17, 20, 21, 23], "accuraci": [17, 22, 23], "problem": [17, 23, 40, 41], "hypertun": [17, 21], "feedback": [17, 20], "expos": [17, 23], "accord": [17, 20, 23], "switch": 17, "english": 17, "prefer": [17, 29], "term": [17, 27], "mail": [17, 40], "gb": [17, 29, 37, 38, 39, 44, 45, 46], "qk": [17, 20], "video": [17, 20], "contextu": [17, 20], "g": [17, 20, 22, 23, 24, 27, 40], "watch": 17, "493": [17, 30], "feel": 17, "free": [17, 22, 23, 27, 29], "reduc": [17, 21, 23, 30, 40], "v100": [17, 29, 30, 31], "truncat": 17, "10000001": 17, "10m": 17, "input_data_path": [17, 19], "insid": [17, 22, 25, 27, 32], "quick_start": [17, 18, 20, 22, 23], "r": [17, 19, 22, 27], "prototyp": 17, "dev": [17, 27, 29], "option": [17, 18, 23, 28, 29, 40, 41], "explan": 17, "discret": [17, 20, 29], "space": [17, 20, 22, 28, 29, 36], "binary_classif_target": [17, 20], "regression_target": [17, 20], "user_id_featur": [17, 20], "item_id_featur": [17, 20], "to_int32": [17, 20], "to_int16": [17, 20], "to_int8": [17, 20], "cast": 17, "lower": [17, 20, 21], "precis": [17, 31], "dataset_split_strategi": [17, 20], "random_by_us": [17, 20], "chosen": [17, 20], "random_split_eval_perc": [17, 20], "percentag": [17, 20], "filter_queri": [17, 20], "equal": [17, 21], "min_item_freq": [17, 20], "min_user_freq": [17, 20], "max_user_freq": [17, 20], "frequenc": [17, 19, 23], "threshold": [17, 20], "larger": [17, 20, 21, 23, 29, 30], "enable_dask_cuda_clust": [17, 20], "localcudaclust": [17, 20, 28, 29], "persist_intermediate_fil": [17, 20], "persist": [17, 20, 23], "cach": [17, 20, 27], "intermedi": [17, 20, 27], "paricular": 17, "data_path": [17, 20], "out_dataset_path": [17, 20, 23], "preproc": [17, 19, 20], "input_data_format": [17, 20], "csv_na_valu": [17, 20], "150": [17, 20], "num_max_rounds_filt": [17, 20], "video_categori": [17, 19, 20], "gender": [17, 19, 20, 36], "ag": [17, 19, 20, 36], "watching_tim": [17, 20], "2nd": [17, 23], "embeddings_dim": [17, 22, 23], "mlp_layer": [17, 22, 23], "decai": [17, 23], "lr_decay_r": [17, 22, 23], "lr_decay_step": [17, 22, 23], "regular": [17, 21], "l2_reg": [17, 22, 23], "embeddings_l2_reg": [17, 22, 23], "among": [17, 20, 22, 23], "train_batch_s": [17, 22, 23], "eval_batch_s": [17, 22, 23], "65536": [17, 22, 23], "faster": [17, 20, 22, 40], "train_steps_per_epoch": [17, 22, 23], "There": [17, 20, 23, 32, 40], "lead": [17, 21, 23], "unbal": 17, "stl_positive_class_weight": [17, 22, 23], "rarer": [17, 23], "cuda_visible_devic": [17, 22, 23, 28, 29], "train_data_path": [17, 22, 23], "eval_data_path": [17, 20, 22, 23], "joint": 17, "expert": [17, 21], "independ": [17, 23], "sub": [17, 19, 23, 35, 36], "mmoe_num_mlp_expert": [17, 23], "expert_mlp_lay": [17, 23], "gate": [17, 21, 23], "gate_dim": [17, 23], "averag": [17, 20, 21, 23], "harvest": 17, "tower_lay": [17, 23], "balanc": [17, 21, 23], "mtl_loss_weight_": [17, 21, 23], "mtl_pos_class_weight_": [17, 21, 23], "use_task_tow": [17, 23], "embedding_sizes_multipli": [17, 22, 23], "mtl_pos_class_weight_click": 17, "mtl_pos_class_weight_lik": 17, "mtl_pos_class_weight_shar": 17, "mtl_pos_class_weight_follow": 17, "mtl_loss_weight_click": 17, "mtl_loss_weight_lik": 17, "mtl_loss_weight_follow": 17, "mtl_loss_weight_shar": 17, "bias": [17, 21, 23], "sweep": [17, 21], "benchmark": [17, 29], "compar": [17, 21, 23, 35, 36], "empir": [17, 23], "curat": [17, 25, 27], "moreov": 17, "quick": [18, 19, 21, 23], "individu": [18, 21], "phase": 18, "skew": 18, "context": [18, 23, 28, 40], "encapsul": 18, "procedur": 18, "postprocess": 18, "section": [18, 20, 21, 23, 27], "br": [18, 19], "nvt_workflow_path": 18, "tf_saved_model_path": 18, "output_ensemble_path": 18, "load_model_path": [18, 23], "exact": 18, "save_model_path": [18, 23], "broken": [19, 35, 36], "preprocesss": 19, "locat": [19, 20, 27], "23": [19, 24, 27, 28, 43], "57": [19, 27, 37, 44], "737454": 19, "input_fold": [19, 32, 33], "workflow_stored_path": 19, "output_targets_path": 19, "outfil": 19, "whose": [19, 21, 23], "thing": [19, 29], "_cach": 19, "16794": 19, "221049": 19, "23542": 19, "61962": 19, "85886": 19, "281786": 19, "6016": 19, "26929": 19, "66043": 19, "30710": 19, "39752": 19, "222908": 19, "8365": 19, "273888": 19, "73739": 19, "280425": 19, "27552": 19, "28110": 19, "17866": 19, "69910": 19, "good": [19, 20, 21, 23, 25], "50231797": 19, "50405663": 19, "50262684": 19, "5003805": 19, "50613105": 19, "4995402": 19, "5027875": 19, "5036676": 19, "4998571": 19, "5052081": 19, "49693626": 19, "49303743": 19, "49347958": 19, "49609515": 19, "4981295": 19, "49890146": 19, "49202597": 19, "49149314": 19, "5004128": 19, "49684843": 19, "congratul": 19, "templat": [20, 23], "portal": 20, "social": [20, 23], "stream": [20, 23], "media": [20, 23], "dislik": 20, "cart": [20, 23], "purchas": [20, 23], "shown": 20, "desir": [20, 25, 28], "tenrec": [20, 21, 22, 23], "static": 20, "dynam": [20, 23, 29], "price": 20, "distinct": 20, "continuous_featur": 20, "natur": [20, 40], "approach": [20, 23, 40], "z": 20, "imput": 20, "constant": [20, 22], "float": [20, 23, 27], "median": [20, 21], "accept": [20, 23, 27], "nomin": 20, "meaning": 20, "represent": [20, 23, 25, 29, 40], "higher": [20, 21, 23], "infrequ": [20, 23, 29], "choic": [20, 21, 29, 40], "shard": [20, 23], "low": [20, 23, 27], "categ_min_freq_cap": 20, "cold": [20, 40], "fresh": 20, "major": 20, "long": [20, 29], "bucket": [20, 27], "collis": 20, "trade": 20, "off": [20, 27], "side": [20, 40], "respect": [20, 22], "udf": 20, "tempor": 20, "geograph": 20, "tip": 20, "inherit": 20, "preprocessingrunn": 20, "generate_nvt_featur": 20, "smooth": 20, "compress": [20, 27, 28, 37, 38, 39, 44, 45, 46], "small": [20, 21, 23, 40], "expand": [20, 21], "binar": 20, "ordin": 20, "period": 20, "dai": [20, 28, 29, 31], "besid": [20, 21], "whole": [20, 23], "segment": 20, "likelihood": [20, 23, 25, 40], "win": 20, "competit": [20, 28], "target_encoding_featur": 20, "target_encoding_target": 20, "week": 20, "month": 20, "year": 20, "quarter": 20, "hour": 20, "subtract": 20, "seen": [20, 21], "trend": 20, "accumul": 20, "engag": 20, "semest": 20, "treat": 20, "countri": 20, "latitud": 20, "longitud": 20, "airport": 20, "tourist": 20, "landmark": 20, "enrich": 20, "geoloc": 20, "censu": 20, "stratifi": 20, "dataset_split_temporal_timestamp": 20, "tsv": [20, 23, 29], "latter": 20, "columnar": 20, "setup": [20, 23, 27], "splitmust": 20, "predict_data_path": [20, 23], "csv_sep": 20, "tab": 20, "saveddefault": 20, "output_num_partit": 20, "filesdefault": 20, "whether": 20, "control_featur": 20, "comma": [20, 23], "kept": [20, 23], "sep": [20, 23, 28], "properli": [20, 22], "continuous_features_fillna": 20, "fill": [20, 29], "max_item_freq": 20, "timestamp_featur": 20, "session_id_featur": 20, "target_encoding_kfold": 20, "fold": 20, "consid": 20, "caus": 20, "overfit": [20, 23, 29], "target_encoding_smooth": 20, "factor": [20, 23, 40], "noisi": 20, "formula": 20, "sum_target_per_categ_valu": 20, "global_target_avg": 20, "categ_value_count": 20, "int16": 20, "to_float32": 20, "greater": [20, 23], "round": [20, 27], "interleav": 20, "proport": 20, "dask_cuda_visible_gpu_devic": 20, "visible_gpu_devic": 20, "dask_cuda_gpu_device_spill_frac": 20, "spill": [20, 29], "rais": [20, 28], "decid": [21, 23], "divid": 21, "mmoe": 21, "ple": 21, "bottom": [21, 23], "special": [21, 40], "fair": 21, "comparison": [21, 29], "bayesian": [21, 22], "bay": [21, 22], "reus": 21, "trial": [21, 22], "mayb": 21, "todo": 21, "maxim": [21, 22], "baselin": 21, "avg": 21, "highlight": 21, "bold": 21, "blue": [21, 40], "highest": 21, "sparser": [21, 23], "densiti": 21, "denser": 21, "necessarili": 21, "mantain": 21, "simpler": 21, "cost": [21, 27, 40], "evolv": [21, 22], "sensit": 21, "7023": 21, "7828": 21, "105": 21, "7826": 21, "chanc": 21, "accur": 21, "diminish": 21, "budget": 21, "bar": [21, 22], "chat": 21, "seach": 21, "spot": 21, "hotter": 21, "color": [21, 22], "smaller": [21, 28], "16k": 21, "reg": [21, 23], "analyz": [21, 25, 35, 36], "plot": [21, 22], "boxplot": 21, "hpo": 21, "curv": 21, "dot": [21, 40], "achiv": 21, "adjust": [22, 27], "interest": [22, 29, 31, 33], "cli": [22, 26, 27], "googl": [22, 27, 40], "vertex": 22, "azur": 22, "program": [22, 33], "toward": 22, "int_uniform": 22, "uniform": 22, "log_uniform": 22, "256": [22, 29, 30], "16384": 22, "preproc_data": 22, "easiest": 22, "preproc_input_data_path": 22, "reinstal": 22, "pypi": 22, "nee": 22, "author": [22, 23], "me": 22, "profil": 22, "hit": 22, "enter": 22, "press": [22, 29], "ctrl": 22, "quit": 22, "preproc_data_path": 22, "log_to_wandb": [22, 23], "wandb_ent": [22, 23], "wandb_project": [22, 23], "sucessfulli": 22, "websit": [22, 42], "rememb": 22, "sweep_nam": 22, "wandb_sweep": 22, "stl_mlp": 22, "stl_click_mlp": 22, "agent": 22, "sweep_id": 22, "team_nam": 22, "project_nam": 22, "identif": 22, "env": [22, 27], "background": 22, "nohup": 22, "nohup0": 22, "nice": [22, 23], "chart": 22, "hparam": 22, "inter": 23, "compani": 23, "heurist": 23, "matrix": [23, 40], "youtubednn": 23, "somewhat": 23, "correl": 23, "spend": 23, "resourc": [23, 41], "abil": 23, "capac": 23, "memor": 23, "wideanddeepmodel": 23, "deepfmmodel": 23, "drlm": 23, "bound": 23, "mixtur": 23, "mmoeblock": 23, "cgcblock": 23, "progress": [23, 27], "Their": 23, "pleblock": 23, "1d": 23, "cap": 23, "fraction": [23, 27], "unbalanc": 23, "alwai": 23, "in_batch_negatives_train": 23, "in_batch_negatives_ev": 23, "domin": 23, "tasks_sample_spac": 23, "match": [23, 27, 29, 31, 32], "empti": 23, "On": 23, "hand": 23, "prior": 23, "predict_output_path": 23, "keep_column": 23, "ignore_column": 23, "wide_n_deep": 23, "three": [23, 27], "selu": 23, "elu": 23, "tanh": 23, "sigmoid": [23, 27], "mlp_init": 23, "glorot_uniform": 23, "decreas": 23, "dcn_interacted_layer_num": 23, "wnd_hashed_cross_num_bin": 23, "10000": [23, 30], "wnd_wide_l2_reg": 23, "wnd_ignore_combin": 23, "_categori": 23, "multihot_max_seq_length": 23, "cgc_num_task_expert": 23, "cgc_num_shared_expert": 23, "ple_num_lay": 23, "mtl_gates_softmax_temperatur": 23, "train_metrics_step": 23, "bit": 23, "throughput": [23, 40], "validation_step": 23, "random_se": 23, "reproduc": 23, "shuffled_train": 23, "metrics_log_frequ": 23, "tensorboard": 23, "log_to_tensorboard": 23, "sign": 23, "wandb": 23, "consol": 23, "wandb_exp_group": 23, "Not": [23, 37, 38, 39, 44, 45, 46], "home": [23, 26, 27], "predict_output_format": 23, "predict_output_keep_col": 23, "tradit": 24, "xgboost": [24, 40], "soon": [24, 41], "xx": [24, 43], "600621": 25, "183": [25, 35], "070258": 25, "070303": 25, "226": [25, 27], "070448": 25, "1638": 25, "16249": 25, "gv100": 25, "tianchi": [25, 27], "aliyun": [25, 27], "1000000": 25, "synthetic_data": [25, 27], "category_temp_directori": 25, "out_path": [25, 29], "61": [25, 28], "68": [25, 35, 38, 45], "meta": 25, "transit": 25, "828993": 25, "executor": 25, "1209": 25, "abort": 25, "indic": [25, 28, 29], "invalid_argu": 25, "placehold": 25, "1563": [25, 31], "eta": [25, 27], "4998": [25, 27], "190967": 25, "69": 25, "38m": 25, "5000": [25, 27], "1min": 25, "2min": 25, "0x7f74b2a4b1c0": 25, "prepare_list_features_layer_call_fn": 25, "prepare_list_features_layer_call_and_return_conditional_loss": 25, "dense_9_layer_call_fn": 25, "96": [25, 35, 36, 37, 38, 39, 44, 45, 46], "wokflow": 25, "dl": 25, "std": [25, 32], "guarante": [25, 32], "edg": 25, "inferenc": 25, "remove_input": [25, 27, 33, 35, 36], "0x7f74b290f550": 25, "subclass": 25, "tf_model_path": 25, "illustr": 25, "meet": 25, "hard": 25, "someth": 25, "model_context_2_layer_call_fn": 25, "model_context_2_layer_call_and_return_conditional_loss": 25, "prepare_list_features_2_layer_call_fn": 25, "prepare_list_features_2_layer_call_and_return_conditional_loss": 25, "tmpomjyo5xq": 25, "772": 25, "773": 25, "789": 25, "790": 25, "67": [25, 27, 38, 45], "73": 25, "74": 25, "0_transformworkflow": [25, 27], "1_predicttensorflow": [25, 27], "compos": [25, 35, 36], "__null_dask_index__": [25, 27], "800000": 25, "85": [25, 31, 35, 36], "5936": 25, "2045": 25, "1670": 25, "484": 25, "830": 25, "88": 25, "800001": 25, "2850": 25, "982": 25, "1879": 25, "544": [25, 32], "934": 25, "800002": 25, "238": 25, "82": 25, "557": 25, "277": 25, "isol": 25, "declar": 25, "output_col": [25, 32, 33], "earlier": 25, "5002032": 25, "5001995": 25, "put": 25, "credenti": [26, 27], "mount": [26, 27], "restart": 27, "wtth": 27, "116": 27, "tar": 27, "gz": [27, 28], "kb": 27, "25hrequir": 27, "attr": 27, "pasta": 27, "importlib": 27, "importlib_metadata": 27, "whl": 27, "patho": 27, "25hcollect": 27, "protobuf3": 27, "py2": 27, "smdebug_rulesconfig": 27, "s3transfer": 27, "botocor": 27, "jmespath": 27, "six": 27, "zipp": 27, "pypars": 27, "pytz": 27, "2017": 27, "dateutil": 27, "pox": 27, "dill": 27, "multiprocess": 27, "py38": 27, "ppft": 27, "contextlib2": 27, "urllib3": [27, 31, 32], "wheel": 27, "25ldone": 27, "25h": 27, "filenam": 27, "809052": 27, "sha256": 27, "f446dd6eed6d268b7f3f2709f8f11c1ba153e382fbea9b2caedd517c1fb71215": 27, "3e": 27, "cb": 27, "b1": 27, "5b13ff7b150aa151e4a11030a6c41b1e457c31a52ea1ef11b0": 27, "protobuf3_to_dict": 27, "4029": 27, "8f99baaa875ba544d54f624f95dfbf4fd52ca96d52ce8af6d05c1ff2bb8435b2": 27, "fc": 27, "2d1e23d8b9a9013a83fbb418a0b17b1e6f81c8db8f53b53934": 27, "smdebug": 27, "rulesconfig": 27, "gather": 27, "traffic": 27, "taobao": 27, "retail": 27, "1_000_000": 27, "archiv": 27, "made": 27, "s3": 27, "upload": 27, "argpars": 27, "tempfil": 27, "runtim": 27, "tf_memory_alloc": 27, "logger": 27, "getlogg": 27, "__name__": 27, "setlevel": 27, "addhandl": 27, "streamhandl": 27, "stdout": 27, "parse_arg": 27, "parser": 27, "argumentpars": 27, "add_argu": 27, "train_dir": 27, "sm_channel_train": 27, "valid_dir": 27, "sm_channel_valid": 27, "sm_model_dir": 27, "parse_known_arg": 27, "create_nvtabular_workflow": 27, "create_ensembl": 27, "arg": [27, 29], "mkdtemp": 27, "workflow_path": 27, "train_data": 27, "valid_data": 27, "model_path": 27, "ensemble_path": 27, "__main__": 27, "_": [27, 35, 36], "pip3": 27, "push": 27, "build_and_push_imag": 27, "sh": 27, "outsid": 27, "region": 27, "euo": 27, "pipefail": 27, "algorithm": [27, 30], "algorithm_nam": 27, "east": 27, "st": 27, "caller": 27, "west": 27, "dkr": 27, "amazonaw": 27, "image_uri": 27, "login": 27, "usernam": 27, "stdin": 27, "ne": 27, "fi": [27, 30], "role": 27, "sess": 27, "prefix": 27, "get_execution_rol": 27, "get_rol": 27, "arn": 27, "awso": 27, "iam": 27, "843263297212": 27, "data_loc": 27, "upload_data": 27, "key_prefix": 27, "sts_client": 27, "get_caller_ident": 27, "my_sess": 27, "region_nam": 27, "ecr_imag": 27, "g4dn": 27, "equip": [27, 35, 36], "t4": [27, 37, 38, 39, 44, 45, 46], "entry_point": 27, "behind": 27, "scene": 27, "sagemaker_program": 27, "training_instance_typ": 27, "xlarg": 27, "instance_count": 27, "instance_typ": 27, "1_024": 27, "trainingprofilerreport": 27, "1667989110": 27, "inprogress": 27, "34m": [27, 32], "34mnvidia": 27, "42766143": 27, "34mcopyright": 27, "2018": 27, "affili": 27, "34mvariou": 27, "modif": 27, "34mthi": 27, "34mby": 27, "34mhttp": 27, "34mnote": 27, "driver": [27, 37, 38, 39, 44, 45, 46], "515": 27, "kernel": 27, "510": 27, "34m2022": 27, "405": 27, "neuron": 27, "detect": [27, 28], "438": 27, "473": 27, "485": 27, "invok": 27, "34mtrain": 27, "additional_framework_paramet": 27, "channel_input_dir": 27, "current_host": 27, "algo": 27, "current_instance_group": 27, "homogeneousclust": 27, "current_instance_group_host": 27, "current_instance_typ": 27, "distribution_host": 27, "distribution_instance_group": 27, "framework_modul": 27, "input_config_dir": 27, "input_data_config": 27, "traininginputmod": 27, "s3distributiontyp": 27, "fullyrepl": 27, "recordwrappertyp": 27, "input_dir": 27, "instance_groups_dict": 27, "instance_group_nam": 27, "is_hetero": 27, "is_mast": 27, "is_modelparallel_en": 27, "is_smddpmprun_instal": 27, "job_nam": 27, "376": [27, 30], "log_level": 27, "master_hostnam": 27, "module_dir": 27, "sourcedir": 27, "module_nam": 27, "network_interface_nam": 27, "eth0": 27, "num_cpu": 27, "num_gpu": [27, 29], "num_neuron": 27, "output_dir": 27, "output_intermediate_dir": 27, "resource_config": 27, "current_group_nam": 27, "user_entry_point": 27, "34menviron": 27, "34msm_host": 27, "34msm_network_interface_nam": 27, "34msm_hp": 27, "34msm_user_entry_point": 27, "34msm_framework_param": 27, "34msm_resource_config": 27, "34msm_input_data_config": 27, "34msm_output_data_dir": 27, "34msm_channel": 27, "34msm_current_host": 27, "34msm_current_instance_typ": 27, "34msm_current_instance_group": 27, "34msm_current_instance_group_host": 27, "34msm_instance_group": 27, "34msm_instance_groups_dict": 27, "34msm_distribution_instance_group": 27, "34msm_is_hetero": 27, "34msm_module_nam": 27, "34msm_log_level": 27, "34msm_framework_modul": 27, "34msm_input_dir": 27, "34msm_input_config_dir": 27, "34msm_output_dir": 27, "34msm_num_cpu": 27, "34msm_num_gpu": 27, "34msm_num_neuron": 27, "34msm_model_dir": 27, "34msm_module_dir": 27, "34msm_training_env": 27, "34msm_user_arg": 27, "34msm_output_intermediate_dir": 27, "34msm_channel_train": 27, "34msm_channel_valid": 27, "34msm_hp_batch_siz": 27, "34msm_hp_epoch": 27, "34mpythonpath": 27, "python38": 27, "dynload": 27, "merlin_hp": 27, "34minvok": 27, "486": 27, "debugg": 27, "761711": 27, "slightli": 27, "tf_enable_onednn_opt": 27, "818302": 27, "991": [27, 35, 36], "819693": 27, "819906": 27, "894084": 27, "34mto": 27, "895367": 27, "895631": 27, "895807": 27, "651703": 27, "651981": 27, "652183": 27, "653025": 27, "1532": [27, 31], "10752": 27, "tesla": [27, 29, 30, 31, 37, 38, 39, 44, 45, 46], "34mworkflow": [27, 32], "tmp5fpdavsc": 27, "34mbatch_siz": 27, "34mepoch": 27, "34m684": 27, "684": 27, "5026": 27, "4990": 27, "11m": 27, "6922": 27, "5222": 27, "6941": 27, "4989": 27, "6858": 27, "5509": 27, "6991": 27, "4994": 27, "6790": 27, "5660": 27, "7052": 27, "4993": 27, "6751": 27, "5722": 27, "7096": 27, "6722": 27, "5755": 27, "7184": 27, "4991": 27, "6700": 27, "5777": 27, "7289": 27, "6687": 27, "5792": 27, "7404": 27, "6678": 27, "5801": 27, "7393": 27, "4988": 27, "89": [27, 31, 32, 36], "requestsdependencywarn": [27, 31, 32], "chardet": [27, 31, 32], "34mwarn": 27, "output_layer_layer_call_fn": [27, 31], "output_layer_layer_call_and_return_conditional_loss": [27, 31], "97": [27, 39, 46], "34minfo": 27, "34mmodel": 27, "34mensembl": 27, "589": 27, "billabl": 27, "model_data": 27, "s3download": 27, "s3down": 27, "xvzf": 27, "artificat": 27, "sagemaker_triton_default_model_nam": 27, "sm_client": 27, "service_nam": 27, "modeldataurl": 27, "sagemaker_triton_tensorflow_vers": 27, "strftime": 27, "d": 27, "h": 27, "gmtime": 27, "create_model_respons": 27, "create_model": 27, "modelnam": 27, "executionrolearn": 27, "primarycontain": 27, "model_arn": 27, "modelarn": 27, "endpoint_instance_typ": 27, "endpoint_config_nam": 27, "create_endpoint_config_respons": 27, "create_endpoint_config": 27, "endpointconfignam": 27, "productionvari": 27, "instancetyp": 27, "initialvariantweight": 27, "initialinstancecount": 27, "variantnam": 27, "alltraff": 27, "endpoint_config_arn": 27, "endpointconfigarn": 27, "endpoint_nam": 27, "create_endpoint_respons": 27, "create_endpoint": 27, "endpointnam": 27, "endpoint_arn": 27, "endpointarn": 27, "describe_endpoint": 27, "endpointstatu": 27, "sleep": 27, "60": [27, 36], "rv": 27, "inservic": 27, "700000": 27, "700001": 27, "80": [27, 36, 37, 38, 39, 44, 45, 46], "5621": 27, "1936": 27, "700002": 27, "776": 27, "700003": 27, "970": 27, "334": 27, "700004": 27, "1939": 27, "668": 27, "700005": 27, "15893": 27, "5474": 27, "700006": 27, "102": 27, "7172": 27, "2470": 27, "700007": 27, "1163": 27, "401": 27, "700008": 27, "582": 27, "201": 27, "700009": 27, "4458": 27, "1536": 27, "636": 27, "983": 27, "1965": 27, "2890": 27, "1214": 27, "694": 27, "521": 27, "174": 27, "184": 27, "316": 27, "285": 27, "569": 27, "977": [27, 30], "837": 27, "1436": 27, "352": 27, "604": 27, "345": 27, "151": 27, "259": 27, "103": 27, "payload": 27, "request_bodi": 27, "header_length": 27, "generate_request_bodi": 27, "binary_data_output": 27, "x0c": 27, "x00": 27, "x12": 27, "x003": 27, "x16": 27, "x04": 27, "x02": 27, "x1e": 27, "x05": 27, "x06": 27, "x0b": 27, "x07": 27, "x18": 27, "x03": 27, "x00p": 27, "x0e": 27, "x1c": 27, "xe2": 27, "x00f": 27, "x11": 27, "xc2": 27, "xf5": 27, "x15": 27, "x08": 27, "xca": 27, "x93": 27, "x8b": 27, "x00j": 27, "x00c": 27, "x90": 27, "x01": 27, "x00n": 27, "x9c": 27, "x00b": 27, "xa6": 27, "x91": 27, "xc9": 27, "xd7": 27, "xad": 27, "xbe": 27, "xb6": 27, "xae": 27, "xb8": 27, "x1d": 27, "x009": 27, "x00e": 27, "x97": 27, "xe9": 27, "xd1": 27, "x00y": 27, "x00w": 27, "x004": 27, "x00g": 27, "kserv": 27, "protocol": [27, 29], "runtime_sm_cli": 27, "invoke_endpoint": 27, "contenttyp": 27, "vnd": 27, "bodi": 27, "header_length_prefix": 27, "header_length_str": 27, "parse_response_bodi": 27, "output_data": 27, "48595208": 27, "4647554": 27, "50048226": 27, "53553176": 27, "5209902": 27, "54944164": 27, "5032344": 27, "475241": 27, "5077254": 27, "5009623": 27, "incur": 27, "delete_model": 27, "delete_endpoint_config": 27, "delete_endpoint": 27, "responsemetadata": 27, "requestid": 27, "6ad24616": 27, "5c7c": 27, "4525": 27, "a63c": 27, "62d1b06ee8ad": 27, "httpstatuscod": 27, "httpheader": 27, "amzn": 27, "amz": 27, "wed": 27, "nov": 27, "gmt": 27, "retryattempt": 27, "tb": [28, 29, 31], "span": [28, 29, 31], "record": 28, "criteolab": 28, "kaggl": 28, "15gb": 28, "47gb": 28, "30min": 28, "download_criteo": 28, "input_path": [28, 31, 33], "crit_orig": 28, "number_dai": 28, "valueerror": 28, "minimum": 28, "day_": [28, 29], "storag": 28, "googleapi": 28, "cail": 28, "day_0": [28, 29], "3gb": 28, "0mb": 28, "day_1": [28, 29], "7gb": 28, "3mb": 28, "15min": 28, "3min": [28, 31], "19min": 28, "23min": 28, "orient": 28, "dask_cuda": [28, 29, 36], "device_mem_s": [28, 29], "get_rmm_siz": 28, "frac_siz": 28, "rmm_pool_siz": [28, 29], "local_directori": [28, 29], "564": 28, "preload": [28, 29, 36], "cont_nam": 28, "cat_nam": 28, "hex": 28, "hexadecim": 28, "file_list": 28, "384": [28, 35, 36], "futurewarn": 28, "set_dask_cli": 28, "preserve_fil": 28, "prerequisit": 29, "8min": [29, 31], "1x": 29, "9min": 29, "8x": 29, "100x": 29, "10000x": 29, "clip": 29, "publish": 29, "strong": 29, "altern": 29, "numba": 29, "fillmiss": 29, "pynvml_mem_s": 29, "test_dask": [29, 30, 31, 32, 33], "use_hugectr": 29, "bool": 29, "stats_path": 29, "stat": 29, "dask_workdir": 29, "workdir": 29, "worker": [29, 30, 36], "fname": 29, "num_dai": 29, "listdir": 29, "dashboard": 29, "dashboard_port": 29, "tcp": [29, 36], "ucx": 29, "is_avail": 29, "visible_devic": 29, "device_limit_frac": 29, "device_pool_frac": 29, "part_mem_frac": 29, "device_s": 29, "device_limit": 29, "device_pool_s": 29, "occupi": 29, "fmem": 29, "1e9": 29, "bewar": 29, "schedul": 29, "port": [29, 36], "n_worker": 29, "device_memory_limit": 29, "dashboard_address": 29, "242": 29, "16739757": 29, "7554": 29, "11ed": 29, "88cc": 29, "2a33bb9638f6": 29, "c4f394c8": 29, "thread": 29, "gib": 29, "a605479f": 29, "02de": 29, "4ca6": 29, "8661": 29, "fd02b84655ac": 29, "comm": 29, "44059": 29, "45833": 29, "38541": 29, "nanni": 29, "43849": 29, "9hxldmq1": 29, "sxm2": [29, 30, 31], "32gb": [29, 30, 31], "75": [29, 38, 45], "35261": 29, "46685": 29, "37015": 29, "nsltf61z": 29, "still": 29, "isn": 29, "ideal": 29, "consumpt": 29, "fact": 29, "magnitud": 29, "uncent": 29, "concis": 29, "worth": 29, "handi": 29, "ve": 29, "occurr": 29, "signal": 29, "10000000": [29, 30], "categorify_op": 29, "cont_featur": 29, "min_valu": 29, "out_dtyp": 29, "label_featur": 29, "output_train_dir": 29, "output_valid_dir": 29, "260": 29, "0x7fdacec4fdc0": 29, "output_fil": 29, "cont": [29, 32], "980": 29, "202": 29, "801": 29, "203": 29, "exploit": 30, "cubla": 30, "nccl": 30, "criteo_hugectr": [30, 32], "noqa": [30, 33], "sgd": [30, 31], "3014529": 30, "400781": 30, "2209": 30, "11869": 30, "38713": 30, "584616": 30, "12883": 30, "109": [30, 35], "17177": 30, "7425": 30, "20266": 30, "7085": 30, "1535": 30, "6000": 30, "relu3": 30, "interaction1": 30, "fc4": 30, "relu4": 30, "fc5": 30, "relu5": 30, "fc6": 30, "relu6": 30, "fc7": 30, "relu7": 30, "fc8": 30, "3200": 30, "1000": 30, "run_tim": 30, "hctr": 30, "374": 30, "rk0": 30, "2831956451": 30, "490": 30, "peer": 30, "fulli": [30, 40], "all2al": 30, "warmup": 30, "491": 30, "num": 30, "dataread": 30, "vocabulari": 30, "54120457": 30, "541": 30, "max_vocabulary_size_per_gpu_": 30, "12288000": 30, "analysi": 30, "resolv": 30, "055": 30, "gpu0": 30, "107": 30, "111": 30, "warm": 30, "113": 30, "125": 30, "9122965335846": 30, "942873": 31, "579306": 31, "579506": 31, "16255": 31, "582211": 31, "582328": 31, "30655": 31, "8a": 31, "path_to_train_data": 31, "path_to_valid_data": 31, "part_1": 31, "2867": 31, "224": 31, "72m": 31, "0378": 31, "recal": 31, "0745e": 31, "binary_accuraci": 31, "9677": 31, "5673": 31, "1385": 31, "val_precis": 31, "val_recal": 31, "val_binary_accuraci": 31, "9680": 31, "6279": 31, "1405": 31, "13min": 31, "21min": 31, "45": 31, "0x7f58457b1c40": 31, "eval_metr": 31, "2787": 31, "31m": 31, "13853003084659576": 31, "9679937958717346": 31, "6279259324073792": 31, "14048266410827637": 31, "predictionoutput": 31, "positive_item_id": 31, "label_relevant_count": 31, "valid_negatives_mask": 31, "negative_item_id": 31, "0x7f57c5f8ecd0": 31, "c1": [31, 32], "c10": [31, 32], "c11": [31, 32], "c12": [31, 32], "c13": [31, 32], "c14": [31, 32], "c15": [31, 32], "c16": [31, 32], "c17": [31, 32], "c18": [31, 32], "c19": [31, 32], "c2": [31, 32], "c20": [31, 32], "c21": [31, 32], "c22": [31, 32], "c23": [31, 32], "c24": [31, 32], "c25": [31, 32], "c26": [31, 32], "c3": [31, 32], "c4": [31, 32], "c5": [31, 32], "c6": [31, 32], "c7": [31, 32], "c8": [31, 32], "c9": [31, 32], "i1": 31, "i10": [31, 32], "i11": [31, 32], "i12": [31, 32], "i13": [31, 32], "i2": 31, "i3": 31, "i4": [31, 32], "i5": [31, 32], "i6": [31, 32], "i7": [31, 32], "i8": [31, 32], "i9": [31, 32], "161": 31, "temporari": 32, "export_hugectr_ensembl": 32, "original_data_path": [32, 33], "9600": 32, "model_infer": 32, "hugectr_model_path": 32, "hugectr_param": 32, "n_output": 32, "tree": 32, "test_merlin_criteo_hugectr": 32, "00m": 32, "34mcriteo": 32, "34m1": 32, "34m0_sparse_9600": 32, "_dense_9600": 32, "_opt_dense_9600": 32, "34mcriteo_en": 32, "34mcriteo_nvt": 32, "34m__pycache__": 32, "34mcategori": 32, "0_sparse_9600": 32, "input_col": [32, 33], "70000": 32, "2714039": 32, "29401": 32, "11464": 32, "1122": 32, "9355": 32, "6370": 32, "1010": 32, "1865651": 32, "208215": 32, "952671": 32, "955872": 32, "944922": 32, "139380": 32, "994092": 32, "056103": 32, "547473": 32, "709442": 32, "930728": 32, "70001": 32, "3514299": 32, "27259": 32, "8072": 32, "395": 32, "9361": 32, "862": 32, "3292987": 32, "171709": 32, "759526": 32, "795019": 32, "716366": 32, "134964": 32, "516737": 32, "065577": 32, "129782": 32, "471361": 32, "386101": 32, "70002": 32, "1304577": 32, "5287": 32, "7367": 32, "2033": 32, "2899": 32, "712": 32, "640": 32, "6415968": 32, "880028": 32, "347701": 32, "207892": 32, "753950": 32, "371013": 32, "759502": 32, "201477": 32, "192447": 32, "085893": 32, "957961": 32, "na": [32, 33], "workaround": [32, 33], "fillna": [32, 33], "criteo_en": 32, "52164096": 32, "50390565": 32, "4957397": 32, "though": 33, "affect": 33, "f401": 33, "extend": [35, 36], "friendli": [35, 36], "accustom": [35, 36], "accommod": [35, 36], "100k": [35, 36], "proce": [35, 36], "breez": [35, 36], "bayesianpersonalizedrank": 35, "entertain": [35, 36], "get_movielen": [35, 36], "variant": [35, 36], "henc": 35, "train_workflow": [35, 36], "681291": 35, "681740": 35, "681877": 35, "frame": [35, 36], "777": 35, "52it": 35, "train_auc": 35, "incom": [35, 36], "lifecycl": [35, 36], "redefin": [35, 36], "subtl": [35, 36], "inf_workflow": [35, 36], "wrap": [35, 36], "predictimplicit": 35, "implicit_model": 35, "ten_exampl": [35, 36], "to_fram": 35, "143": 35, "609": 35, "858": 35, "predictions_from_triton": [35, 36], "local_predict": [35, 36], "assert_allclos": 35, "xgb": 36, "preprocess_categori": 36, "preprocess_r": 36, "rating_binari": 36, "num_boost_round": 36, "verbose_ev": 36, "446602": 36, "447091": 36, "447227": 36, "94mb": 36, "45mb": 36, "32file": 36, "947": 36, "diskutil": 36, "stale": 36, "lock": 36, "oqemvhkv": 36, "purg": 36, "41809": 36, "got": 36, "rmse": 36, "36952": 36, "95316": 36, "92447": 36, "90741": 36, "89437": 36, "89138": 36, "fil": 36, "predictforest": 36, "te_movieid_r": 36, "userid_count": 36, "zip_cod": 36, "77": 36, "779876": 36, "572154": 36, "231": 36, "896619": 36, "goldeney": 36, "366": 36, "954632": 36, "room": 36, "093809": 36, "shorti": 36, "383": 36, "539376": 36, "copycat": 36, "perhap": 36, "35647": 36, "allclos": 36, "station": [37, 38, 39, 44, 45, 46], "ubuntu": [37, 38, 39, 44, 45, 46], "certifi": [37, 38, 39, 44, 45, 46], "465": [37, 38, 39, 44, 45, 46], "center": [37, 38, 39, 44, 45, 46], "formerli": [37, 38, 39, 44, 45, 46], "418": [37, 38, 39, 44, 45, 46], "r418": [37, 38, 39, 44, 45, 46], "440": [37, 38, 39, 44, 45, 46], "r440": [37, 38, 39, 44, 45, 46], "450": [37, 38, 39, 44, 45, 46], "r450": [37, 38, 39, 44, 45, 46], "r460": [37, 38, 39, 44, 45, 46], "amper": [37, 38, 39, 44, 45, 46], "ture": [37, 38, 39, 44, 45, 46], "volta": [37, 38, 39, 44, 45, 46], "pascal": [37, 38, 39, 44, 45, 46], "023": [37, 38, 39, 44, 45, 46], "011": [37, 38, 39, 44, 45, 46], "065": [37, 38, 39, 44, 45, 46], "rmm": [37, 38, 39, 44, 45, 46], "00a": [37, 38, 39, 44, 45, 46], "gf6bf047": [37, 38, 39, 44, 45, 46], "dirti": [37, 38, 39, 44, 45, 46], "g6ca81bbc78": [37, 38, 39, 44, 45, 46], "131": [37, 38, 39, 44, 45, 46], "cuda11": [37, 38, 39, 44, 45, 46], "95": [37, 44], "017": [37, 38, 39, 44, 45, 46], "014": [37, 38, 39, 44, 45, 46], "022": [37, 38, 39, 44, 45, 46], "76": [37, 38, 39, 44, 45, 46], "g185c18e6": [37, 38, 39, 44, 45, 46], "319": [37, 38, 39, 44, 45, 46], "g97422602b8": [37, 38, 39, 44, 45, 46], "g1354dcf": [37, 38, 39, 44, 45, 46], "g042ee198": [37, 39, 44, 46], "g886cf6de": [37, 38, 39, 44, 45, 46], "gf1e2bf5": [37, 39, 44, 46], "g712b04d": [37, 38, 39, 44, 45, 46], "g32ed5fa12": [37, 38, 39, 44, 45, 46], "g3367d725": [37, 39, 44, 46], "g046077b8": [38, 45], "g07bf4ab": [38, 45], "1a": 40, "1b": 40, "ncf": 40, "criteo": 40, "guidanc": 40, "2b": 40, "yellow": 40, "cosin": 40, "latenc": 40, "2a": 40, "mf": 40, "delai": 40, "retrain": 40, "hi": 40, "her": 40, "intent": 40, "viewed_item_id": 40, "4a": 40, "nlp": 40, "rnn": 40, "gru": 40, "lstm": 40, "4b": 40, "othersequenti": 40, "casual": 40, "clm": 40, "mlm": 40, "gru4rec": 40, "lstm4rec": 40, "gpt": 40, "bert": 40, "xl": 40, "electra": 40, "introduct": [40, 42], "bi": 40, "flexibl": 40, "divers": 41, "bring": 42, "older": 43}, "objects": {}, "objtypes": {}, "objnames": {}, "titleterms": {"nvidia": [0, 15], "merlin": [0, 1, 2, 3, 10, 14, 15, 16, 17, 19, 24, 25, 26, 27, 31, 33, 35, 36, 42, 43, 44, 45, 46], "benefit": 0, "compon": 0, "instal": 0, "exampl": [0, 4, 15, 24, 26], "notebook": [0, 4, 15, 24, 26], "tutori": 0, "i": 0, "built": 0, "On": 0, "feedback": 0, "support": [0, 23, 43, 44, 45, 46], "contain": [1, 17, 27, 37, 38, 39, 44, 45, 46], "structur": 1, "chang": 1, "begin": 1, "22": [1, 37, 38, 39, 44, 45, 46], "06": 1, "releas": 1, "build": [2, 15, 25, 27], "intellig": 2, "recommend": [2, 3, 4, 12, 13, 25, 27, 35, 36, 40, 41], "system": [2, 3, 4, 19, 25, 35, 36, 41], "overview": [2, 5, 6, 8, 9, 10, 11, 12, 29, 30, 31, 32, 33, 35, 36], "learn": [2, 5, 6, 8, 10, 11, 17, 19, 21, 23, 29, 30, 31, 32, 33, 35, 36], "object": [2, 3, 5, 6, 8, 10, 11, 19, 29, 30, 31, 32, 33, 35, 36], "import": [2, 3, 21, 25], "requir": [2, 3], "librari": [2, 3, 25], "function": [2, 3], "set": [2, 3, 20, 23], "up": [2, 3, 27], "featur": [2, 3, 9, 20, 23, 25], "store": [2, 3], "feast": 2, "export": [2, 3, 25, 32], "user": 2, "item": [2, 5, 6], "engin": [2, 20, 25], "nvtabular": [2, 5, 8, 25, 29, 33], "train": [2, 5, 6, 7, 9, 10, 11, 15, 17, 20, 23, 24, 25, 26, 27, 30, 31, 35, 36], "retriev": [2, 3, 25, 27, 35, 36, 40], "model": [2, 5, 6, 9, 11, 12, 13, 17, 18, 19, 21, 23, 24, 25, 27, 31, 33, 35, 36, 40], "two": 2, "tower": 2, "queri": 2, "rank": [2, 17, 18, 19, 21, 23, 24, 25, 40], "dlrm": [2, 23, 25, 31], "extract": 2, "save": [2, 11, 25, 31, 33], "embed": [2, 6, 23], "creat": [2, 3, 6, 18, 22, 25, 27, 35, 36], "definit": [2, 11, 20], "next": [2, 5, 6, 8, 31, 32, 33], "step": [2, 8, 31, 32, 33], "deploi": [3, 4, 15, 18, 19, 24, 25, 29, 32, 33], "multi": [3, 4, 15, 17, 21, 23], "stage": [3, 4, 15, 23], "recsi": [3, 15, 23], "product": 3, "triton": [3, 5, 12, 13, 17, 18, 19, 25, 27, 32, 33, 35, 36], "infer": [3, 5, 11, 13, 17, 18, 19, 25, 27, 32, 33, 35, 36], "server": [3, 5, 12, 13, 17, 18, 19, 25, 27, 32, 33, 35, 36], "regist": [3, 27], "our": [3, 8, 9, 11, 29], "load": [3, 12, 13], "from": [3, 12, 13, 25, 27, 32, 33, 35, 36], "offlin": 3, "an": [3, 17, 35, 36], "onlin": 3, "faiss": 3, "index": 3, "client": [3, 22], "ensembl": [3, 18, 25, 32, 33, 35, 36], "graph": [3, 18, 25, 35, 36], "start": [3, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 19, 22, 25, 32, 35, 36], "run": [4, 8, 15, 21, 24, 26], "transform": [5, 6, 27], "base": [5, 6, 23, 40], "architectur": [5, 6, 10, 23], "predict": [5, 6, 13, 32, 33], "task": [5, 6, 17, 20, 21, 23], "download": [5, 6, 7, 17, 28], "prepar": [5, 6, 11, 17, 35, 36], "dataset": [5, 6, 7, 14, 15, 17, 20, 27, 28, 34, 35, 36], "preprocess": [5, 8, 17, 20, 29], "construct": [5, 6], "evalu": [5, 23, 31], "serv": [5, 6, 12, 13, 15, 26, 27, 35, 36], "us": [5, 9, 27], "summari": [5, 6, 19, 25, 31, 32, 33], "pretrain": 6, "clean": [6, 27], "data": [6, 11, 17, 20, 29, 35, 36], "gener": [6, 27], "synthet": 6, "workflow": [6, 33], "get": [7, 8, 9, 10, 11, 12, 14, 15, 32, 33], "movielen": [7, 8, 9, 10, 11, 14, 15], "convert": [7, 28], "movielens25m": [7, 10, 11], "split": [7, 20], "valid": 7, "etl": [8, 29], "defin": [8, 9, 10, 20, 23, 29], "pipelin": [8, 29], "check": 8, "pre": 8, "process": [8, 21, 22], "output": [8, 20, 23], "hugectr": [9, 12, 30, 32, 44], "why": 9, "other": 9, "scale": [9, 15, 28, 29, 30, 31, 32, 33, 34], "acceler": 9, "let": 9, "": 9, "pytorch": [10, 45], "initi": 10, "dataload": 10, "neural": [10, 21], "network": 10, "tensorflow": [11, 13, 31, 33, 46], "specifi": 11, "hyperparamet": [11, 17, 21, 22], "launch": [13, 18], "inventori": 15, "larg": [15, 34], "criteo": [15, 28, 29, 30, 31, 32, 33, 34], "aw": [15, 26, 27], "sagemak": [15, 26, 27], "quick": [16, 17, 22], "setup": [17, 21, 22], "tenrec": 17, "docker": [17, 27], "mmoe": [17, 23], "tune": [17, 21, 22], "deploy": 17, "send": [18, 27], "request": [18, 27], "command": [18, 20, 23], "line": [18, 20, 23], "argument": [18, 20, 23], "input": [18, 20, 23], "script": [20, 23, 27, 28], "best": [20, 21, 23], "practic": [20, 23], "deal": [20, 23], "high": [20, 23], "cardin": [20, 23], "target": 20, "encod": 20, "cast": 20, "filter": 20, "frequenc": [20, 29], "cap": 20, "eval": 20, "cuda": 20, "cluster": [20, 29], "option": 20, "benchmark": 21, "singl": 21, "stl": 21, "result": [21, 22], "most": 21, "mtl": [21, 23], "analysi": 21, "varianc": 21, "accuraci": 21, "its": 21, "improv": 21, "over": 21, "refin": 21, "search": 21, "space": [21, 23], "weight": [22, 23], "bias": 22, "sweep": 22, "1": [22, 27], "config": 22, "file": 22, "2": [22, 27], "environ": 22, "wandb": 22, "3": [22, 27], "test": 22, "4": 22, "5": 22, "hypertun": 22, "monitor": 22, "analys": 22, "baselin": 23, "categor": 23, "size": 23, "regular": 23, "class": 23, "neg": 23, "sampl": 23, "loss": 23, "dcn": 23, "v2": 23, "deepfm": 23, "wide": 23, "deep": 23, "cgc": 23, "ple": 23, "expert": 23, "log": [23, 28], "verif": 25, "artifact": [25, 27, 32], "part": 27, "imag": [27, 37, 38, 39, 44, 45, 46], "dockerfil": 27, "your": 27, "python": 27, "sdk": 27, "raw": 27, "termin": 27, "endpoint": 27, "1tb": 28, "click": 28, "convers": 28, "csv": 28, "parquet": 28, "prep": 29, "distribut": 29, "dask": 29, "threshold": 29, "implicit": 35, "tool": [35, 36], "prerequisit": [35, 36], "xgboost": 36, "23": [37, 38, 39, 44, 45, 46], "xx": [37, 38, 39, 44, 45, 46], "resourc": [40, 42], "sequenti": 40, "session": 40, "guid": 41, "relat": 42, "matrix": [43, 44, 45, 46]}, "envversion": {"sphinx.domains.c": 2, "sphinx.domains.changeset": 1, "sphinx.domains.citation": 1, "sphinx.domains.cpp": 8, "sphinx.domains.index": 1, "sphinx.domains.javascript": 2, "sphinx.domains.math": 2, "sphinx.domains.python": 3, "sphinx.domains.rst": 2, "sphinx.domains.std": 2, "sphinx.ext.viewcode": 1, "sphinx.ext.intersphinx": 1, "sphinx": 57}, "alltitles": {"NVIDIA Merlin": [[0, "nvidia-merlin"]], "Benefits": [[0, "benefits"]], "Components of NVIDIA Merlin": [[0, "components-of-nvidia-merlin"]], "Installation": [[0, "installation"]], "Example Notebooks and Tutorials": [[0, "example-notebooks-and-tutorials"]], "Merlin Is Built On": [[0, "merlin-is-built-on"]], "Feedback and Support": [[0, "feedback-and-support"]], "Merlin Containers": [[1, "merlin-containers"]], "Structural Changes Beginning with the 22.06 Releases": [[1, "structural-changes-beginning-with-the-22-06-releases"]], "Building Intelligent Recommender Systems with Merlin": [[2, "building-intelligent-recommender-systems-with-merlin"]], "Overview": [[2, "overview"], [5, "overview"], [6, "overview"], [8, "overview"], [9, "overview"], [10, "overview"], [11, "overview"], [12, "overview"], [29, "overview"], [30, "overview"], [31, "overview"], [32, "overview"], [33, "overview"], [35, "overview"], [36, "overview"]], "Learning objectives": [[2, "learning-objectives"], [5, "learning-objectives"], [6, "learning-objectives"], [8, "learning-objectives"], [10, "learning-objectives"], [11, "learning-objectives"], [29, "learning-objectives"], [30, "learning-objectives"], [31, "learning-objectives"], [32, "learning-objectives"], [33, "learning-objectives"], [35, "learning-objectives"], [36, "learning-objectives"]], "Import required libraries and functions": [[2, "import-required-libraries-and-functions"], [3, "import-required-libraries-and-functions"]], "Set up a feature store with Feast": [[2, "set-up-a-feature-store-with-feast"]], "Exporting user and item features": [[2, "exporting-user-and-item-features"]], "Feature Engineering with NVTabular": [[2, "feature-engineering-with-nvtabular"], [2, "id1"], [25, "feature-engineering-with-nvtabular"]], "Training a Retrieval Model with Two-Tower Model": [[2, "training-a-retrieval-model-with-two-tower-model"]], "Exporting query (user) model": [[2, "exporting-query-user-model"]], "Training a Ranking Model with DLRM": [[2, "training-a-ranking-model-with-dlrm"]], "Extract and save Item embeddings": [[2, "extract-and-save-item-embeddings"]], "Create feature definitions": [[2, "create-feature-definitions"]], "Next Steps": [[2, "next-steps"], [8, "next-steps"]], "Deploying a Multi-Stage RecSys into Production with Merlin Systems and Triton Inference Server": [[3, "deploying-a-multi-stage-recsys-into-production-with-merlin-systems-and-triton-inference-server"]], "Register our features on feature store": [[3, "register-our-features-on-feature-store"]], "Loading features from offline store into an online store": [[3, "loading-features-from-offline-store-into-an-online-store"]], "Set up Faiss index, create feature store client and objects for the Triton ensemble": [[3, "set-up-faiss-index-create-feature-store-client-and-objects-for-the-triton-ensemble"]], "Export Graph as Ensemble": [[3, "export-graph-as-ensemble"], [25, "export-graph-as-ensemble"]], "Starting Triton Server": [[3, "starting-triton-server"], [25, "starting-triton-server"]], "Retrieving Recommendations from Triton": [[3, "retrieving-recommendations-from-triton"]], "Deploying a Multi-Stage Recommender System": [[4, "deploying-a-multi-stage-recommender-system"]], "Running the Example Notebooks": [[4, "running-the-example-notebooks"], [15, "running-the-example-notebooks"], [24, "running-the-example-notebooks"]], "Transformer-based architecture for next-item prediction task": [[5, "transformer-based-architecture-for-next-item-prediction-task"]], "Downloading and preparing the dataset": [[5, "downloading-and-preparing-the-dataset"], [6, "downloading-and-preparing-the-dataset"]], "Preprocessing with NVTabular": [[5, "preprocessing-with-nvtabular"]], "Constructing the model": [[5, "constructing-the-model"]], "Model training": [[5, "model-training"]], "Model evaluation": [[5, "model-evaluation"]], "Serving predictions using the Triton Inference Server": [[5, "serving-predictions-using-the-triton-inference-server"]], "Summary": [[5, "summary"], [6, "summary"], [19, "summary"], [25, "summary"], [31, "summary"], [32, "summary"], [33, "summary"]], "Transformer-based architecture for next-item prediction task with pretrained embeddings": [[6, "transformer-based-architecture-for-next-item-prediction-task-with-pretrained-embeddings"]], "Clean downloaded data": [[6, "clean-downloaded-data"]], "Generate synthetic data": [[6, "generate-synthetic-data"]], "Constructing a workflow": [[6, "constructing-a-workflow"]], "Creating and training the model": [[6, "creating-and-training-the-model"]], "Serving predictions": [[6, "serving-predictions"]], "Getting Started MovieLens: Download and Convert": [[7, "getting-started-movielens-download-and-convert"]], "MovieLens25M": [[7, "movielens25m"], [10, "movielens25m"], [11, "movielens25m"]], "Download the dataset": [[7, "download-the-dataset"]], "Convert the dataset": [[7, "convert-the-dataset"]], "Splitting into train and validation dataset": [[7, "splitting-into-train-and-validation-dataset"]], "Getting Started MovieLens: ETL with NVTabular": [[8, "getting-started-movielens-etl-with-nvtabular"]], "ETL with NVTabular": [[8, "etl-with-nvtabular"], [29, "etl-with-nvtabular"]], "Defining our Preprocessing Pipeline": [[8, "defining-our-preprocessing-pipeline"], [29, "defining-our-preprocessing-pipeline"]], "Running the pipeline": [[8, "running-the-pipeline"]], "Checking the pre-processing outputs": [[8, "checking-the-pre-processing-outputs"]], "Getting Started MovieLens: Training with HugeCTR": [[9, "getting-started-movielens-training-with-hugectr"]], "Why use HugeCTR?": [[9, "why-use-hugectr"]], "Other Features of HugeCTR": [[9, "other-features-of-hugectr"]], "Getting Started": [[9, "getting-started"], [12, "getting-started"]], "Scaling Accelerated training with HugeCTR": [[9, "scaling-accelerated-training-with-hugectr"]], "Let\u2019s define our model": [[9, "let-s-define-our-model"]], "Getting Started MovieLens: Training with PyTorch": [[10, "getting-started-movielens-training-with-pytorch"]], "Initializing the Merlin Dataloader for PyTorch": [[10, "initializing-the-merlin-dataloader-for-pytorch"]], "Defining Neural Network Architecture": [[10, "defining-neural-network-architecture"]], "Getting Started MovieLens: Training with TensorFlow": [[11, "getting-started-movielens-training-with-tensorflow"]], "Data Preparation": [[11, "data-preparation"]], "Training our model": [[11, "training-our-model"], [11, "id1"]], "Model definition": [[11, "model-definition"]], "Specifying Hyperparameters": [[11, "specifying-hyperparameters"]], "Saving the model for inference": [[11, "saving-the-model-for-inference"]], "Serve Recommendations from the HugeCTR Model": [[12, "serve-recommendations-from-the-hugectr-model"]], "Load Models on Triton Server": [[12, "load-models-on-triton-server"]], "Serve Recommendations from the TensorFlow Model": [[13, "serve-recommendations-from-the-tensorflow-model"]], "Launching and Starting the Triton Server": [[13, "launching-and-starting-the-triton-server"]], "Loading the Tensorflow Model with Triton Inference Server": [[13, "loading-the-tensorflow-model-with-triton-inference-server"]], "Predicting": [[13, "predicting"]], "Getting Started with Merlin and the MovieLens Dataset": [[14, "getting-started-with-merlin-and-the-movielens-dataset"]], "NVIDIA Merlin Example Notebooks": [[15, "nvidia-merlin-example-notebooks"]], "Inventory": [[15, "inventory"]], "Building and Deploying a multi-stage RecSys": [[15, "building-and-deploying-a-multi-stage-recsys"]], "Getting Started with MovieLens": [[15, "getting-started-with-movielens"]], "Scaling Large Datasets with Criteo": [[15, "scaling-large-datasets-with-criteo"], [34, "scaling-large-datasets-with-criteo"]], "Training and Serving with Merlin on AWS SageMaker": [[15, "training-and-serving-with-merlin-on-aws-sagemaker"]], "Quick-start for Merlin": [[16, "quick-start-for-merlin"]], "Quick-start for ranking models with Merlin": [[17, "quick-start-for-ranking-models-with-merlin"]], "Setup": [[17, "setup"]], "Download the TenRec dataset": [[17, "download-the-tenrec-dataset"]], "Preparing the data": [[17, "preparing-the-data"]], "Start Docker container": [[17, "start-docker-container"]], "Preprocessing": [[17, "preprocessing"]], "Training a ranking model": [[17, "training-a-ranking-model"]], "Training a ranking model with multi-task learning": [[17, "training-a-ranking-model-with-multi-task-learning"]], "Training an MMOE model": [[17, "training-an-mmoe-model"]], "Hyperparameter tuning": [[17, "hyperparameter-tuning"]], "Model Deployment on Triton Inference Server": [[17, "model-deployment-on-triton-inference-server"]], "Deploying a Ranking model on Triton Inference Server": [[18, "deploying-a-ranking-model-on-triton-inference-server"]], "Creating the Ensemble Graph": [[18, "creating-the-ensemble-graph"]], "Launching Triton Inference Server": [[18, "launching-triton-inference-server"]], "Sending request to Triton": [[18, "sending-request-to-triton"]], "Command line arguments": [[18, "command-line-arguments"], [20, "command-line-arguments"], [23, "command-line-arguments"]], "Inputs": [[18, "inputs"], [20, "inputs"], [23, "inputs"]], "Deploying Ranking Models with Merlin Systems": [[19, "deploying-ranking-models-with-merlin-systems"]], "Learning Objectives": [[19, "learning-objectives"]], "Starting Triton Inference Server": [[19, "starting-triton-inference-server"]], "Deploy models on Triton Inference Server": [[19, "deploy-models-on-triton-inference-server"]], "Preprocessing script": [[20, "preprocessing-script"]], "Best practices": [[20, "best-practices"], [23, "best-practices"]], "Dataset": [[20, "dataset"], [35, "dataset"], [36, "dataset"]], "Defining the task": [[20, "defining-the-task"]], "Preprocessing features": [[20, "preprocessing-features"]], "Dealing with high-cardinality data": [[20, "dealing-with-high-cardinality-data"]], "Feature Engineering": [[20, "feature-engineering"]], "Data set splitting": [[20, "data-set-splitting"]], "Outputs": [[20, "outputs"], [23, "outputs"]], "Features and targets definition": [[20, "features-and-targets-definition"]], "Target encoding features": [[20, "target-encoding-features"]], "Data casting and filtering": [[20, "data-casting-and-filtering"]], "Filtering and frequency capping": [[20, "filtering-and-frequency-capping"]], "Dataset splitting (train and eval sets)": [[20, "dataset-splitting-train-and-eval-sets"]], "CUDA cluster options": [[20, "cuda-cluster-options"]], "Benchmark of ranking models": [[21, "benchmark-of-ranking-models"]], "Neural ranking models.": [[21, "neural-ranking-models"]], "Hyperparameter tuning setup": [[21, "hyperparameter-tuning-setup"]], "Single-task learning": [[21, "single-task-learning"]], "STL Benchmark results": [[21, "stl-benchmark-results"]], "Most important hyperparameters": [[21, "most-important-hyperparameters"], [21, "id1"]], "Best runs hyperparameters": [[21, "best-runs-hyperparameters"], [21, "id2"]], "Multi-task learning": [[21, "multi-task-learning"], [23, "multi-task-learning"]], "MTL Benchmark results": [[21, "mtl-benchmark-results"]], "Analysis of the hyperparameter tuning process": [[21, "analysis-of-the-hyperparameter-tuning-process"]], "Variance of accuracy and its improvement over runs": [[21, "variance-of-accuracy-and-its-improvement-over-runs"]], "Refine the search space": [[21, "refine-the-search-space"]], "Hyperparameter tuning with Quick-start and Weights&Biases Sweeps": [[22, "hyperparameter-tuning-with-quick-start-and-weights-biases-sweeps"]], "1. Creating the Sweeps config file": [[22, "creating-the-sweeps-config-file"]], "2. Environment setup": [[22, "environment-setup"]], "2.1 Setup wandb client": [[22, "setup-wandb-client"]], "3. Testing the environment": [[22, "testing-the-environment"]], "4. Creating the Sweep at Weights&Biases": [[22, "creating-the-sweep-at-weights-biases"]], "5. Starting the hypertuning process": [[22, "starting-the-hypertuning-process"]], "5. Monitoring the hypertuning process & analysing results": [[22, "monitoring-the-hypertuning-process-analysing-results"]], "Ranking script": [[23, "ranking-script"]], "Ranking in multi-stage RecSys": [[23, "ranking-in-multi-stage-recsys"]], "Multi-task learning for ranking models": [[23, "multi-task-learning-for-ranking-models"]], "Supported models": [[23, "supported-models"]], "Baseline ranking architectures": [[23, "baseline-ranking-architectures"]], "Multi-task learning architectures": [[23, "multi-task-learning-architectures"]], "Modeling inputs features": [[23, "modeling-inputs-features"]], "Dealing with high-cardinality categorical features": [[23, "dealing-with-high-cardinality-categorical-features"]], "Defining the embedding size": [[23, "defining-the-embedding-size"]], "Regularization": [[23, "regularization"]], "Classes weights": [[23, "classes-weights"]], "Negative sampling": [[23, "negative-sampling"], [23, "id1"]], "Losses weights": [[23, "losses-weights"]], "Setting tasks sample space": [[23, "setting-tasks-sample-space"]], "Tasks": [[23, "tasks"]], "Model": [[23, "model"]], "DCN-v2": [[23, "dcn-v2"]], "DLRM and DeepFM": [[23, "dlrm-and-deepfm"]], "Wide&Deep": [[23, "wide-deep"]], "Wide&Deep and DeepFM": [[23, "wide-deep-and-deepfm"]], "MMOE": [[23, "mmoe"]], "CGC and PLE": [[23, "cgc-and-ple"]], "Expert-based MTL models": [[23, "expert-based-mtl-models"]], "Multi-task learning models": [[23, "multi-task-learning-models"]], "Training and evaluation": [[23, "training-and-evaluation"]], "Logging": [[23, "logging"]], "Training and Deploying Ranking models with Merlin": [[24, "training-and-deploying-ranking-models-with-merlin"]], "Exporting Ranking Models": [[25, "exporting-ranking-models"]], "Importing Libraries": [[25, "importing-libraries"]], "Build and Train a DLRM model": [[25, "build-and-train-a-dlrm-model"]], "Save model": [[25, "save-model"]], "Deploying the model with Merlin Systems": [[25, "deploying-the-model-with-merlin-systems"]], "Create the Ensemble Graph": [[25, "create-the-ensemble-graph"], [35, "create-the-ensemble-graph"], [36, "create-the-ensemble-graph"]], "Verification of Ensemble Artifacts": [[25, "verification-of-ensemble-artifacts"]], "Retrieving Recommendations from Triton Inference Server": [[25, "retrieving-recommendations-from-triton-inference-server"], [35, "retrieving-recommendations-from-triton-inference-server"], [36, "retrieving-recommendations-from-triton-inference-server"]], "Training and Serving Merlin on AWS SageMaker": [[26, "training-and-serving-merlin-on-aws-sagemaker"], [27, "training-and-serving-merlin-on-aws-sagemaker"]], "Running the Example Notebook": [[26, "running-the-example-notebook"]], "Part 1: Generating Dataset and Docker image": [[27, "part-1-generating-dataset-and-docker-image"]], "Generating Dataset": [[27, "generating-dataset"]], "Training Script": [[27, "training-script"]], "Create the Dockerfile": [[27, "create-the-dockerfile"]], "Building and registering the container": [[27, "building-and-registering-the-container"]], "Part 2: Training your Merlin model on Sagemaker": [[27, "part-2-training-your-merlin-model-on-sagemaker"]], "Training on Sagemaker using the Python SDK": [[27, "training-on-sagemaker-using-the-python-sdk"]], "Part 3: Retrieving Recommendations from Triton Inference Server": [[27, "part-3-retrieving-recommendations-from-triton-inference-server"]], "Send a Request to Triton Inference Server to Transform a Raw Dataset": [[27, "send-a-request-to-triton-inference-server-to-transform-a-raw-dataset"]], "Terminate endpoint and clean up artifacts": [[27, "terminate-endpoint-and-clean-up-artifacts"]], "Scaling Criteo: Download and Convert": [[28, "scaling-criteo-download-and-convert"]], "Criteo 1TB Click Logs dataset": [[28, "criteo-1tb-click-logs-dataset"]], "Conversion Script for Criteo Dataset (CSV-to-Parquet)": [[28, "conversion-script-for-criteo-dataset-csv-to-parquet"]], "Scaling Criteo: ETL with NVTabular": [[29, "scaling-criteo-etl-with-nvtabular"]], "Data Prep": [[29, "data-prep"]], "Deploy a Distributed-Dask Cluster": [[29, "deploy-a-distributed-dask-cluster"]], "Frequency Thresholding": [[29, "frequency-thresholding"]], "Scaling Criteo: Training with HugeCTR": [[30, "scaling-criteo-training-with-hugectr"]], "Training with HugeCTR": [[30, "training-with-hugectr"]], "Scaling Criteo: Training with Merlin Models TensorFlow": [[31, "scaling-criteo-training-with-merlin-models-tensorflow"]], "Training a DLRM model": [[31, "training-a-dlrm-model"]], "Evaluate the model": [[31, "evaluate-the-model"]], "Save the model": [[31, "save-the-model"]], "Next steps": [[31, "next-steps"], [32, "next-steps"], [33, "next-steps"]], "Scaling Criteo: Triton Inference with HugeCTR": [[32, "scaling-criteo-triton-inference-with-hugectr"]], "Deploying Ensemble to Triton Inference Server": [[32, "deploying-ensemble-to-triton-inference-server"], [33, "deploying-ensemble-to-triton-inference-server"]], "Export artifacts": [[32, "export-artifacts"]], "Start Triton Inference Server": [[32, "start-triton-inference-server"]], "Get prediction from Triton Inference Server": [[32, "get-prediction-from-triton-inference-server"], [33, "get-prediction-from-triton-inference-server"]], "Scaling Criteo: Triton Inference with Merlin Models TensorFlow": [[33, "scaling-criteo-triton-inference-with-merlin-models-tensorflow"]], "Saved NVTabular workflow and Merlin Models": [[33, "saved-nvtabular-workflow-and-merlin-models"]], "Serving an Implicit Model with Merlin Systems": [[35, "serving-an-implicit-model-with-merlin-systems"]], "Tools": [[35, "tools"], [36, "tools"]], "Prerequisite: Preparing the data and Training Implicit": [[35, "prerequisite-preparing-the-data-and-training-implicit"]], "Starting the Triton Inference Server": [[35, "starting-the-triton-inference-server"], [36, "starting-the-triton-inference-server"]], "Serving an XGBoost Model with Merlin Systems": [[36, "serving-an-xgboost-model-with-merlin-systems"]], "Prerequisite: Preparing the data and Training XGBoost": [[36, "prerequisite-preparing-the-data-and-training-xgboost"]], "23.xx Container Images": [[37, "xx-container-images"], [38, "xx-container-images"], [39, "xx-container-images"], [44, "xx-container-images"], [45, "xx-container-images"], [46, "xx-container-images"]], "22.xx Container Images": [[37, "id13"], [38, "id13"], [39, "id13"], [44, "id13"], [45, "id13"], [46, "id13"]], "Recommender Models": [[40, "recommender-models"]], "Ranking Models": [[40, "ranking-models"]], "Resources for Ranking Models:": [[40, "resources-for-ranking-models"]], "Retrieval Models": [[40, "retrieval-models"]], "Resources for Retrieval Models:": [[40, "resources-for-retrieval-models"]], "Sequential and Session-Based Models": [[40, "sequential-and-session-based-models"]], "Resources for Session-Based Models:": [[40, "resources-for-session-based-models"]], "Recommender Systems Guide": [[41, "recommender-systems-guide"]], "Merlin": [[42, "merlin"]], "Related Resources": [[42, "related-resources"]], "Merlin Support Matrix": [[43, "merlin-support-matrix"]], "Merlin HugeCTR Support Matrix": [[44, "merlin-hugectr-support-matrix"]], "Merlin PyTorch Support Matrix": [[45, "merlin-pytorch-support-matrix"]], "Merlin TensorFlow Support Matrix": [[46, "merlin-tensorflow-support-matrix"]]}, "indexentries": {}}) \ No newline at end of file +Search.setIndex({"docnames": ["README", "containers", "examples/Building-and-deploying-multi-stage-RecSys/01-Building-Recommender-Systems-with-Merlin", "examples/Building-and-deploying-multi-stage-RecSys/02-Deploying-multi-stage-RecSys-with-Merlin-Systems", "examples/Building-and-deploying-multi-stage-RecSys/index", "examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction", "examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction-with-pretrained-embeddings", "examples/getting-started-movielens/01-Download-Convert", "examples/getting-started-movielens/02-ETL-with-NVTabular", "examples/getting-started-movielens/03-Training-with-HugeCTR", "examples/getting-started-movielens/03-Training-with-PyTorch", "examples/getting-started-movielens/03-Training-with-TF", "examples/getting-started-movielens/04-Triton-Inference-with-TF", "examples/getting-started-movielens/index", "examples/index", "examples/quick_start/index", "examples/quick_start/ranking", "examples/quick_start/scripts/inference/index", "examples/quick_start/scripts/inference/inference", "examples/quick_start/scripts/preproc/index", "examples/quick_start/scripts/ranking/hypertuning/index", "examples/quick_start/scripts/ranking/hypertuning/tutorial_with_wb_sweeps", "examples/quick_start/scripts/ranking/index", "examples/ranking/index", "examples/ranking/tf/Training-and-Deploying-DLRM-model-with-Models-and-Systems", "examples/sagemaker-tensorflow/index", "examples/sagemaker-tensorflow/sagemaker-merlin-tensorflow", "examples/scaling-criteo/01-Download-Convert", "examples/scaling-criteo/02-ETL-with-NVTabular", "examples/scaling-criteo/03-Training-with-HugeCTR", "examples/scaling-criteo/03-Training-with-Merlin-Models-TensorFlow", "examples/scaling-criteo/04-Triton-Inference-with-Merlin-Models-TensorFlow", "examples/scaling-criteo/index", "examples/traditional-ml/Serving-An-Implicit-Model-With-Merlin-Systems", "examples/traditional-ml/Serving-An-XGboost-Model-With-Merlin-Systems", "generated/nvcr.io-nvidia-merlin-merlin-hugectr", "generated/nvcr.io-nvidia-merlin-merlin-pytorch", "generated/nvcr.io-nvidia-merlin-merlin-tensorflow", "guide/recommender_models", "guide/recommender_system_guide", "index", "support_matrix/index", "support_matrix/support_matrix_merlin_hugectr", "support_matrix/support_matrix_merlin_pytorch", "support_matrix/support_matrix_merlin_tensorflow"], "filenames": ["README.md", "containers.rst", "examples/Building-and-deploying-multi-stage-RecSys/01-Building-Recommender-Systems-with-Merlin.ipynb", "examples/Building-and-deploying-multi-stage-RecSys/02-Deploying-multi-stage-RecSys-with-Merlin-Systems.ipynb", "examples/Building-and-deploying-multi-stage-RecSys/index.md", "examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction.ipynb", "examples/Next-Item-Prediction-with-Transformers/tf/transformers-next-item-prediction-with-pretrained-embeddings.ipynb", "examples/getting-started-movielens/01-Download-Convert.ipynb", "examples/getting-started-movielens/02-ETL-with-NVTabular.ipynb", "examples/getting-started-movielens/03-Training-with-HugeCTR.ipynb", "examples/getting-started-movielens/03-Training-with-PyTorch.ipynb", "examples/getting-started-movielens/03-Training-with-TF.ipynb", "examples/getting-started-movielens/04-Triton-Inference-with-TF.ipynb", "examples/getting-started-movielens/index.md", "examples/index.md", "examples/quick_start/index.md", "examples/quick_start/ranking.md", "examples/quick_start/scripts/inference/index.md", "examples/quick_start/scripts/inference/inference.ipynb", "examples/quick_start/scripts/preproc/index.md", "examples/quick_start/scripts/ranking/hypertuning/index.md", "examples/quick_start/scripts/ranking/hypertuning/tutorial_with_wb_sweeps.md", "examples/quick_start/scripts/ranking/index.md", "examples/ranking/index.md", "examples/ranking/tf/Training-and-Deploying-DLRM-model-with-Models-and-Systems.ipynb", "examples/sagemaker-tensorflow/index.md", "examples/sagemaker-tensorflow/sagemaker-merlin-tensorflow.ipynb", "examples/scaling-criteo/01-Download-Convert.ipynb", "examples/scaling-criteo/02-ETL-with-NVTabular.ipynb", "examples/scaling-criteo/03-Training-with-HugeCTR.ipynb", "examples/scaling-criteo/03-Training-with-Merlin-Models-TensorFlow.ipynb", "examples/scaling-criteo/04-Triton-Inference-with-Merlin-Models-TensorFlow.ipynb", "examples/scaling-criteo/index.md", "examples/traditional-ml/Serving-An-Implicit-Model-With-Merlin-Systems.ipynb", "examples/traditional-ml/Serving-An-XGboost-Model-With-Merlin-Systems.ipynb", "generated/nvcr.io-nvidia-merlin-merlin-hugectr.rst", "generated/nvcr.io-nvidia-merlin-merlin-pytorch.rst", "generated/nvcr.io-nvidia-merlin-merlin-tensorflow.rst", "guide/recommender_models.rst", "guide/recommender_system_guide.rst", "index.rst", "support_matrix/index.rst", "support_matrix/support_matrix_merlin_hugectr.rst", "support_matrix/support_matrix_merlin_pytorch.rst", "support_matrix/support_matrix_merlin_tensorflow.rst"], "titles": ["NVIDIA Merlin", "Merlin Containers", "Building Intelligent Recommender Systems with Merlin", "Deploying a Multi-Stage RecSys into Production with Merlin Systems and Triton Inference Server", "Deploying a Multi-Stage Recommender System", "Transformer-based architecture for next-item prediction task", "Transformer-based architecture for next-item prediction task with pretrained embeddings", "Getting Started MovieLens: Download and Convert", "Getting Started MovieLens: ETL with NVTabular", "Getting Started MovieLens: Training with HugeCTR", "Getting Started MovieLens: Training with PyTorch", "Getting Started MovieLens: Training with TensorFlow", "Serve Recommendations from the TensorFlow Model", "Getting Started with Merlin and the MovieLens Dataset", "NVIDIA Merlin Example Notebooks", "Quick-start for Merlin", "Quick-start for ranking models with Merlin", "Deploying a Ranking model on Triton Inference Server", "Deploying Ranking Models with Merlin Systems", "Preprocessing script", "Benchmark of ranking models", "Hyperparameter tuning with Quick-start and Weights&Biases Sweeps", "Ranking script", "Training and Deploying Ranking models with Merlin", "Exporting Ranking Models", "Training and Serving Merlin on AWS SageMaker", "Training and Serving Merlin on AWS SageMaker", "Scaling Criteo: Download and Convert", "Scaling Criteo: ETL with NVTabular", "Scaling Criteo: Training with HugeCTR", "Scaling Criteo: Training with Merlin Models TensorFlow", "Scaling Criteo: Triton Inference with Merlin Models TensorFlow", "Scaling Large Datasets with Criteo", "Serving an Implicit Model with Merlin Systems", "Serving an XGBoost Model with Merlin Systems", "23.xx Container Images", "23.xx Container Images", "23.xx Container Images", "Recommender Models", "Recommender Systems Guide", "Merlin", "Merlin Support Matrix", "Merlin HugeCTR Support Matrix", "Merlin PyTorch Support Matrix", "Merlin TensorFlow Support Matrix"], "terms": {"an": [0, 2, 5, 6, 7, 8, 9, 10, 11, 12, 14, 17, 18, 19, 21, 22, 24, 26, 27, 28, 29, 30, 31, 38, 40], "open": [0, 2, 4, 5, 14, 17, 18, 23, 29, 33, 34, 40], "sourc": [0, 2, 3, 5, 9, 11, 17, 18, 26, 29, 33, 34, 40], "librari": [0, 1, 4, 5, 6, 7, 8, 11, 12, 14, 15, 16, 17, 18, 19, 20, 22, 26, 27, 28, 29, 30, 31, 33, 34, 38, 40], "acceler": [0, 8, 10, 11, 13, 14, 16, 18, 19, 25, 26, 28, 29, 33, 34, 40], "recommend": [0, 5, 7, 8, 9, 10, 11, 13, 14, 15, 16, 17, 18, 19, 22, 23, 25, 27, 28, 29, 30, 31, 32, 40], "system": [0, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 17, 19, 22, 23, 25, 26, 27, 28, 29, 30, 31, 32, 35, 36, 37, 38, 40, 42, 43, 44], "gpu": [0, 1, 2, 3, 4, 5, 6, 8, 9, 11, 13, 14, 16, 19, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 32, 35, 36, 37, 38, 40, 42, 43, 44], "The": [0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 11, 12, 13, 14, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44], "enabl": [0, 2, 3, 5, 6, 11, 18, 19, 21, 22, 24, 26, 29, 30, 42, 43, 44], "data": [0, 2, 3, 4, 5, 7, 8, 9, 10, 12, 13, 14, 15, 17, 18, 22, 24, 26, 27, 29, 30, 31, 35, 36, 37, 38, 42, 43, 44], "scientist": 0, "machin": [0, 2, 21, 22, 24, 28, 38], "learn": [0, 9, 13, 14, 19, 21, 23, 24, 25, 26, 38], "engin": [0, 4, 6, 8, 11, 13, 14, 15, 16, 25, 26, 27, 28, 31, 32, 38], "research": [0, 38], "build": [0, 3, 4, 16, 17, 22, 23, 38, 40], "high": [0, 8, 13, 14, 26, 28, 38, 40], "perform": [0, 1, 2, 3, 4, 5, 6, 11, 14, 16, 18, 19, 20, 22, 23, 24, 26, 30, 33, 34, 40], "scale": [0, 5, 6, 8, 18, 19, 22, 24, 33, 34, 38, 40], "includ": [0, 1, 5, 6, 14, 16, 17, 18, 19, 20, 21, 22, 26, 27, 33, 34], "tool": [0, 2, 5, 6, 20, 24], "address": [0, 12], "common": [0, 16, 19, 22, 23, 26, 28, 38], "featur": [0, 4, 5, 7, 8, 10, 11, 13, 14, 15, 16, 20, 25, 26, 27, 28, 30, 31, 32, 33, 34, 38], "train": [0, 1, 3, 4, 8, 12, 13, 15, 17, 18, 20, 21, 27, 28, 31, 32, 38, 41, 42, 43, 44], "infer": [0, 1, 2, 4, 6, 8, 9, 13, 14, 22, 23, 25, 27, 29, 32, 35, 36, 37, 41, 42, 43, 44], "challeng": [0, 2, 6, 8], "each": [0, 1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 17, 18, 19, 20, 21, 22, 24, 26, 27, 28, 29, 30, 31, 32, 33, 34, 38, 39, 41], "stage": [0, 2, 12, 15, 17, 20, 25, 26, 38], "pipelin": [0, 2, 3, 4, 5, 6, 9, 10, 14, 16, 17, 18, 20, 22, 24, 26, 29, 31, 33, 34], "optim": [0, 2, 3, 5, 6, 9, 10, 11, 18, 20, 21, 22, 24, 26, 29, 30], "hundr": [0, 19, 22, 38], "terabyt": [0, 8, 28], "which": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 14, 15, 16, 17, 18, 19, 20, 21, 22, 24, 26, 28, 29, 30, 31, 38], "all": [0, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 42, 43, 44], "access": [0, 1, 6, 14, 18, 19, 29, 38], "through": [0, 3, 5, 6, 8, 9, 10, 11, 15, 19, 22, 24, 26, 28, 31, 33, 34, 38], "easi": [0, 9, 16, 17, 18, 21, 22, 26, 29, 33, 34], "us": [0, 1, 2, 3, 4, 6, 7, 8, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 42, 43, 44], "api": [0, 5, 8, 9, 13, 14, 16, 17, 21, 22, 26, 28, 29, 30, 31, 33, 34], "For": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 16, 17, 18, 19, 20, 21, 22, 24, 25, 27, 28, 31, 33, 34, 38, 40], "more": [0, 1, 2, 3, 4, 5, 6, 8, 9, 10, 11, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 28, 29, 30, 31, 33, 34, 38, 39, 40], "inform": [0, 1, 2, 4, 5, 6, 7, 8, 9, 10, 11, 12, 16, 17, 18, 19, 21, 22, 23, 24, 28, 29, 31, 33, 34, 38, 40], "see": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 17, 18, 19, 20, 24, 25, 26, 27, 28, 29, 30, 31, 33, 34, 38, 40, 41], "develop": [0, 2, 5, 9, 18, 23, 26, 29, 33, 34, 39, 40], "web": 0, "site": 0, "scalabl": [0, 5, 17, 18, 22, 38], "solut": [0, 9, 19, 21, 22, 24, 29], "make": [0, 3, 5, 6, 8, 14, 16, 18, 19, 20, 21, 22, 24, 25, 26, 28, 33, 34], "from": [0, 1, 2, 4, 5, 6, 7, 8, 9, 10, 11, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 27, 28, 29, 30, 38, 39], "end": [0, 2, 5, 14, 18, 21, 22, 29, 31, 33, 34, 38], "With": [0, 1, 3, 5, 20, 25, 26, 31, 33, 34], "you": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 42, 43, 44], "can": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 14, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 38, 41, 42, 43, 44], "transform": [0, 2, 8, 12, 17, 18, 24, 28, 33, 38], "etl": [0, 4, 9, 10, 14, 24, 27, 34], "preprocess": [0, 2, 4, 6, 9, 12, 13, 14, 15, 17, 18, 21, 22, 24, 25, 26, 33, 34], "your": [0, 2, 3, 4, 5, 6, 11, 12, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 31, 33, 34], "exist": [0, 2, 5, 6, 8, 9, 10, 21, 22, 26, 27, 28], "tensorflow": [0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 13, 14, 16, 18, 21, 22, 23, 24, 25, 26, 27, 28, 32, 33, 34, 37, 38, 41], "pytorch": [0, 1, 2, 3, 5, 6, 7, 8, 13, 14, 18, 23, 24, 27, 28, 33, 34, 36, 38, 41], "fastai": 0, "leverag": [0, 6, 8, 10, 11, 14, 16, 22, 28, 29, 33, 34, 38], "custom": [0, 2, 5, 6, 8, 16, 19, 22, 24, 26, 33, 34, 38], "loader": [0, 3, 6, 9, 10], "larg": [0, 8, 9, 16, 19, 22, 29, 30, 38], "deep": [0, 2, 3, 5, 6, 8, 9, 11, 16, 18, 20, 24, 26, 28, 29, 30, 33, 34, 38], "model": [0, 1, 3, 4, 8, 10, 13, 14, 15, 19, 21, 25, 28, 29, 32, 35, 36, 37, 39, 41, 42, 43, 44], "distribut": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 19, 21, 22, 24, 26, 27, 29, 30, 31, 33, 34, 35, 36, 37, 42, 43, 44], "embed": [0, 1, 3, 4, 5, 8, 9, 10, 11, 16, 19, 20, 28, 29, 35, 36, 37, 38, 42, 43, 44], "tabl": [0, 1, 3, 9, 19, 20, 22, 29], "exce": 0, "avail": [0, 1, 4, 6, 9, 12, 14, 15, 16, 19, 20, 21, 22, 23, 24, 26, 27, 29, 30, 32, 33, 34, 38, 40, 41], "cpu": [0, 2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 16, 18, 19, 22, 24, 26, 27, 28, 30], "memori": [0, 2, 3, 5, 6, 8, 9, 11, 16, 19, 22, 24, 26, 28, 29, 30], "deploi": [0, 2, 12, 16, 19, 20, 21, 22, 25, 26, 30, 32, 33, 34, 38, 41, 42, 43, 44], "product": [0, 2, 4, 5, 6, 8, 9, 12, 17, 18, 19, 20, 24, 30, 31, 33, 34, 38, 42, 43, 44], "onli": [0, 2, 5, 7, 8, 9, 10, 11, 16, 17, 19, 20, 22, 24, 26, 30, 33, 34, 38], "few": [0, 11, 22], "line": [0, 2, 3, 4, 6, 11, 14, 16, 20, 21, 23, 24, 26], "code": [0, 2, 3, 5, 6, 8, 11, 17, 18, 19, 24, 26, 28, 29, 33, 34, 38], "consist": [0, 3, 5, 6, 8, 17, 18, 33, 34], "follow": [0, 1, 2, 3, 4, 5, 6, 9, 11, 12, 13, 14, 16, 17, 18, 19, 20, 22, 23, 24, 25, 26, 28, 30, 31, 32, 33, 34, 35, 36, 37, 41, 42, 43, 44], "nvtabular": [0, 3, 4, 6, 7, 9, 10, 11, 13, 14, 16, 17, 18, 19, 25, 26, 27, 30, 32, 33, 34, 35, 36, 37, 42, 43, 44], "tabular": [0, 5, 6, 8, 19, 22, 28], "quickli": [0, 8, 15, 16, 18, 28, 33, 34], "easili": [0, 2, 3, 5, 8, 16, 17, 22, 24, 28], "manipul": [0, 8, 28], "size": [0, 5, 6, 8, 10, 14, 16, 20, 26, 27, 28, 29, 32, 35, 36, 37, 42, 43, 44], "dataset": [0, 2, 3, 8, 9, 10, 11, 12, 15, 17, 18, 20, 21, 22, 23, 24, 28, 29, 30, 31, 38, 39], "ar": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 33, 34, 38, 39], "base": [0, 2, 3, 4, 7, 8, 9, 10, 11, 12, 14, 15, 16, 18, 19, 23, 24, 26, 27, 28, 31, 33, 34, 35, 36, 37, 39, 42, 43, 44], "offer": [0, 5, 9, 26, 29], "level": [0, 8, 13, 14, 16, 20, 22, 26, 28], "defin": [0, 2, 3, 5, 6, 7, 11, 12, 16, 17, 18, 21, 24, 26, 27, 29, 30, 31, 33, 34, 38], "complex": [0, 2, 8, 22, 39], "workflow": [0, 2, 3, 5, 8, 9, 11, 14, 16, 17, 18, 19, 24, 26, 28, 30, 33, 34, 42, 43, 44], "prepar": [0, 14, 18, 19, 21, 24, 26, 28], "experiment": 0, "so": [0, 2, 5, 6, 7, 8, 10, 11, 14, 16, 17, 18, 19, 20, 21, 22, 24, 26, 28, 33, 34], "process": [0, 2, 5, 6, 11, 14, 15, 16, 19, 22, 24, 28, 29, 30, 31, 33, 34, 38, 40], "without": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 15, 16, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34, 38], "have": [0, 2, 3, 5, 6, 8, 9, 10, 12, 14, 16, 17, 18, 19, 20, 21, 22, 24, 26, 28, 33, 34, 38], "worri": 0, "about": [0, 1, 2, 5, 6, 8, 10, 11, 14, 16, 17, 21, 23, 24, 25, 26, 28, 29, 33, 34, 38, 40, 41], "focu": [0, 14, 18, 33, 34, 38], "what": [0, 2, 6, 9, 11, 12, 16, 28, 33, 34], "do": [0, 3, 4, 5, 6, 9, 12, 16, 17, 18, 19, 24, 26, 31, 33, 34], "how": [0, 1, 2, 4, 5, 6, 7, 8, 10, 11, 14, 16, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 31, 32, 33, 34, 38], "abstract": [0, 8, 24, 28], "oper": [0, 2, 3, 4, 5, 6, 8, 11, 18, 19, 22, 24, 26, 30, 33, 34, 35, 36, 37, 42, 43, 44], "hugectr": [0, 1, 7, 8, 13, 14, 27, 28, 32, 35, 36, 37, 38, 41, 43, 44], "framework": [0, 3, 9, 11, 14, 18, 23, 24, 28, 29, 30, 33, 34, 38], "across": [0, 8, 9, 19, 21, 28], "multipl": [0, 2, 3, 5, 6, 8, 9, 10, 14, 16, 17, 19, 20, 21, 22, 28, 29, 32, 38], "node": [0, 3, 5, 6, 9, 11, 14, 24, 26, 28, 29, 32, 33, 34], "contain": [0, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 15, 18, 19, 20, 21, 22, 23, 24, 25, 27, 28, 29, 30, 31, 32, 33, 34, 38, 41], "provid": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 15, 16, 17, 19, 20, 21, 22, 23, 24, 26, 27, 28, 30, 31, 32, 38, 39, 41], "strategi": [0, 8, 16, 19, 38], "beyond": [0, 2], "over": [0, 2, 8, 9, 10, 15, 16, 21, 27, 28, 29, 30], "load": [0, 2, 5, 6, 7, 8, 9, 10, 11, 17, 18, 22, 24, 26, 28, 30, 31, 33, 34], "subset": [0, 27], "coars": 0, "grain": 0, "demand": [0, 8], "manner": [0, 28], "dure": [0, 2, 5, 8, 11, 16, 17, 18, 20, 21, 22, 24, 31, 33, 34], "standard": [0, 5, 6, 8, 17, 18, 19, 26, 28], "aim": [0, 22], "qualiti": [0, 38], "implement": [0, 5, 8, 10, 11, 19, 22, 26, 38], "rang": [0, 10, 19, 20, 21, 22, 27, 28], "classic": [0, 38], "highli": [0, 9, 10, 11, 12, 29], "advanc": [0, 5, 6, 19, 20, 22], "rank": [0, 3, 4, 15, 19, 21, 25, 26, 31, 34, 39], "up": [0, 4, 6, 8, 9, 10, 16, 18, 21, 22, 25, 28, 29, 33, 34], "10x": [0, 9, 29], "iter": [0, 8, 10, 15, 16, 27, 28], "rapidli": 0, "explor": [0, 2, 13, 16, 20, 21, 24, 32, 38], "map": [0, 2, 3, 5, 6, 8, 10, 18, 19, 21, 22, 24, 28, 29], "creat": [0, 4, 5, 7, 8, 9, 10, 11, 12, 14, 16, 18, 19, 22, 27, 28, 29, 30, 31], "input": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 16, 18, 21, 24, 26, 27, 28, 29, 30, 31, 33, 34, 38], "layer": [0, 5, 6, 8, 10, 11, 16, 20, 22, 28, 38], "automat": [0, 5, 10, 17, 19, 22, 27], "chang": [0, 3, 5, 6, 8, 19], "either": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 19, 20, 21, 22, 24, 26, 27, 28, 29, 30, 31, 33, 34], "impact": 0, "other": [0, 2, 3, 5, 6, 8, 11, 16, 18, 19, 20, 21, 22, 24, 26, 30], "assembl": 0, "connect": [0, 12, 17, 18, 24, 27, 28, 38], "block": [0, 5, 16, 22, 24, 38], "recsi": [0, 2, 4, 15, 16, 19], "architectur": [0, 2, 7, 8, 9, 11, 16, 20, 24, 28, 29, 30, 31, 35, 36, 37, 38, 39, 42, 43, 44], "new": [0, 2, 3, 8, 19, 22, 26, 28, 34, 38], "transformers4rec": [0, 35, 36, 37, 38, 42, 43, 44], "sequenti": [0, 3, 10, 17, 21, 39], "session": [0, 5, 6, 15, 26, 39], "modular": 0, "compat": [0, 2, 16, 19, 26, 28, 33, 34, 35, 36, 37, 42, 43, 44], "modul": [0, 2, 3, 5, 6, 8, 10, 11, 18, 22, 24, 27, 28, 34], "design": [0, 5, 8, 9, 11, 15, 16, 17, 19, 20, 23, 28], "tower": [0, 16, 20, 22, 38], "head": [0, 2, 5, 6, 7, 8, 16, 19, 20, 22, 24, 34], "task": [0, 2, 3, 11, 18, 21, 24, 26, 30, 34, 38], "loss": [0, 2, 5, 6, 9, 10, 11, 16, 20, 24, 26, 29, 30], "ani": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 14, 18, 19, 22, 24, 26, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 42, 43, 44], "take": [0, 2, 3, 5, 6, 7, 8, 10, 11, 18, 20, 22, 24, 27, 28, 34], "advantag": [0, 9, 29, 38], "integr": [0, 2, 9, 10, 24], "seamless": [0, 24], "next": [0, 3, 7, 10, 15, 16, 17, 20, 24, 28, 29, 38], "item": [0, 3, 4, 7, 8, 10, 11, 16, 18, 19, 22, 24, 26, 33, 34, 38], "predict": [0, 2, 3, 8, 11, 14, 16, 18, 19, 20, 22, 24, 26, 30, 33, 34, 38], "well": [0, 1, 5, 7, 8, 10, 11, 14, 16, 27, 28, 32, 38], "binari": [0, 2, 3, 5, 6, 8, 9, 11, 14, 16, 18, 19, 20, 22, 24, 26, 30], "classif": [0, 14, 16, 19, 22], "regress": [0, 16, 19, 22, 34], "combin": [0, 3, 5, 6, 9, 10, 11, 19, 22, 29], "element": 0, "like": [0, 2, 4, 5, 6, 8, 11, 14, 16, 17, 18, 19, 20, 21, 22, 23, 24, 26, 27, 28, 33, 34], "store": [0, 4, 6, 10, 14, 16, 19, 22, 26, 28, 31, 38], "nearest": [0, 2, 3, 31, 38], "neighbor": [0, 3, 31, 38], "search": [0, 2, 3, 4, 14, 16, 21], "serv": [0, 2, 3, 4, 11, 13, 17, 18, 20, 24, 31, 38], "triton": [0, 2, 4, 6, 9, 11, 13, 14, 15, 23, 25, 32, 35, 36, 37, 38, 42, 43, 44], "server": [0, 2, 4, 6, 9, 11, 13, 14, 21, 23, 25, 29, 32, 35, 36, 37, 42, 43, 44], "start": [0, 2, 4, 5, 6, 17, 19, 20, 22, 23, 26, 27, 28, 29, 30, 31, 38], "platform": [0, 2, 3, 5, 6, 11, 18, 19, 24, 26, 30], "graph": [0, 5, 8, 26, 29, 31], "gener": [0, 2, 3, 8, 9, 10, 16, 17, 18, 19, 20, 22, 24, 31, 33, 34, 38], "popular": [0, 2, 7, 9, 10, 11, 13, 14, 16, 19, 22, 29, 30, 38], "commonli": 0, "found": [0, 2, 3, 5, 6, 8, 9, 10, 11, 20, 24, 26, 28, 30, 33, 34], "core": [0, 2, 3, 5, 6, 7, 8, 11, 12, 14, 16, 18, 22, 24, 26, 27, 28, 30, 31, 33, 34, 35, 36, 37, 42, 43, 44], "function": [0, 5, 6, 8, 10, 11, 17, 19, 22, 24, 26, 28, 30, 38], "throughout": [0, 28, 33, 34], "ecosystem": 0, "schema": [0, 2, 3, 5, 6, 8, 10, 11, 12, 16, 18, 19, 22, 24, 26, 28, 30, 31, 33, 34], "identifi": [0, 1, 4, 5, 14, 23, 24, 28], "kei": [0, 1, 5, 6, 9, 19, 21, 22], "autom": [0, 16], "routin": 0, "simplifi": [0, 8, 17, 24, 28, 40], "share": [0, 16, 19, 20, 21, 22, 38], "construct": [0, 8, 33, 34], "simplest": 0, "wai": [0, 5, 6, 8, 11, 17, 18, 19, 21, 33, 34, 38], "run": [0, 1, 2, 3, 5, 6, 9, 16, 17, 18, 19, 21, 22, 24, 26, 28, 31, 33, 34, 35, 36, 37, 42, 43, 44], "docker": [0, 1, 4, 12, 14, 19, 21, 22, 23, 25], "cloud": [0, 1, 14, 21, 23, 24], "ngc": [0, 1, 2, 4, 14, 26], "depend": [0, 6, 7, 8, 9, 10, 11, 12, 14, 16, 22, 24, 28, 29], "receiv": [0, 10, 12, 16, 17, 24, 31, 33, 34, 38], "unit": 0, "test": [0, 2, 5, 6, 7, 16, 17, 24, 26, 27, 33], "page": [0, 2, 4, 16, 23, 41], "To": [0, 1, 2, 3, 4, 5, 6, 10, 11, 14, 16, 18, 23, 24, 25, 26, 30, 41], "contribut": 0, "review": [0, 38], "document": [0, 5, 6, 8, 15, 16, 18, 19, 20, 22, 23, 24, 25, 28, 31, 33, 34, 41], "environ": [0, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 18, 24, 25, 26, 27, 28, 29, 30, 31, 33, 34], "set": [0, 4, 5, 6, 7, 8, 11, 16, 17, 18, 20, 21, 24, 25, 26, 27, 28, 33, 34, 38, 39], "conda": 0, "pip": [0, 2, 3, 16, 21, 24, 26], "step": [0, 3, 5, 6, 10, 11, 12, 14, 16, 17, 18, 22, 23, 24, 25, 26, 33, 34, 38], "http": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 18, 21, 22, 23, 24, 26, 27, 28, 29, 30, 31, 33, 34, 40], "github": [0, 3, 5, 8, 29, 40], "io": [0, 2, 4, 5, 8, 9, 10, 11, 16, 21, 23, 24, 26, 28, 30, 33], "master": 0, "hugectr_contributor_guid": 0, "html": [0, 2, 5, 8, 9, 10, 11, 24, 33], "A": [0, 5, 8, 9, 16, 19, 22, 27, 38], "collect": [0, 8, 10, 14, 17, 26, 28], "form": [0, 5, 19, 22, 28], "jupyt": [0, 2, 4, 5, 8, 9, 10, 11, 14, 18, 23, 24, 33, 34], "demonstr": [0, 3, 4, 14, 16, 17, 18, 21, 22, 23, 24, 32, 33, 34, 38], "download": [0, 13, 24, 26, 28, 32, 33, 34], "These": [0, 3, 5, 6, 11, 14, 17, 19, 24, 26, 30], "differ": [0, 2, 4, 7, 10, 14, 16, 19, 20, 22, 24, 26, 28, 30, 38, 39], "wide": [0, 9, 16, 20, 29, 38], "real": [0, 2, 24, 26], "world": [0, 2, 26], "case": [0, 3, 5, 6, 15, 16, 19, 20, 21, 22, 23, 26, 38, 39], "rapid": [0, 8, 28], "cudf": [0, 2, 3, 5, 6, 7, 8, 12, 16, 18, 19, 24, 28, 33, 34, 35, 36, 37, 42, 43, 44], "reli": 0, "datafram": [0, 2, 5, 7, 8, 12, 16, 19, 24, 26], "dask": [0, 8, 16, 19, 27, 34], "within": [0, 5, 9, 26], "dataload": [0, 3, 6, 11, 13, 14, 16, 22, 30, 35, 36, 37, 42, 43, 44], "report": [0, 20, 26], "bug": [0, 33, 34], "get": [0, 2, 3, 5, 6, 12, 16, 18, 21, 22, 24, 26, 27, 28, 29, 30, 33, 34, 38], "help": [0, 3, 5, 6, 8, 14, 22, 24], "pleas": [0, 2, 3, 5, 6, 8, 9, 10, 11, 12, 17, 18, 21, 23, 24, 26, 28, 33, 34], "issu": [0, 2, 16, 19, 28], "compon": [1, 6, 14, 18, 22, 24, 33, 34, 40], "nvidia": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 16, 17, 18, 21, 23, 24, 25, 26, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 38, 39, 40, 42, 43, 44], "ncg": 1, "catalog": [1, 2, 3, 4, 14, 22, 38], "com": [1, 3, 5, 14, 24, 26, 27, 40], "name": [1, 2, 3, 5, 6, 8, 10, 11, 12, 16, 18, 20, 21, 22, 24, 26, 27, 29, 30, 33, 34], "url": [1, 4, 6, 12, 14, 18, 21, 23, 24, 26], "org": [1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 22, 24, 26, 27, 28, 29, 30, 31, 33, 34], "team": [1, 5, 21, 22], "plugin": 1, "must": [1, 5, 6, 11, 17, 22, 24, 26, 33, 34], "instal": [1, 2, 3, 5, 6, 16, 18, 21, 24, 26, 31, 33, 34], "toolkit": [1, 26], "support": [1, 3, 8, 9, 10, 16, 19, 23, 24, 26, 27, 29, 30, 31, 35, 36, 37], "link": [1, 5, 6, 16, 21, 39], "referenc": [1, 6, 12, 31], "preced": [1, 2], "launch": [1, 3, 4, 21, 24, 26], "befor": [1, 2, 3, 6, 12, 19, 22, 24, 26, 38, 41], "one": [1, 4, 5, 6, 8, 10, 11, 14, 16, 17, 18, 19, 21, 22, 23, 24, 26, 27, 30, 31, 33, 34, 38], "separ": [1, 8, 19, 20, 21, 22, 38, 41], "copyright": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "2023": [2, 3, 5, 6, 11, 18, 24], "corpor": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "right": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "reserv": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "licens": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "under": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 23, 24, 26, 27, 28, 29, 30, 31, 33, 34], "apach": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "version": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 23, 24, 26, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 41, 42, 43, 44], "2": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 19, 20, 22, 24, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 42, 43, 44], "0": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 17, 18, 19, 20, 21, 22, 23, 24, 26, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 38, 42, 43, 44], "mai": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "thi": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 33, 34, 38, 39, 42, 43, 44], "file": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 17, 18, 19, 22, 24, 26, 27, 28, 29, 30, 31, 33, 34], "except": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 20, 24, 26, 27, 28, 29, 30, 31, 33, 34], "complianc": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "obtain": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 21, 24, 26, 27, 28, 29, 30, 31, 33, 34], "copi": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 18, 19, 23, 24, 26, 27, 28, 29, 30, 31, 33, 34], "www": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "unless": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "applic": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 42, 43, 44], "law": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "agre": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "write": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "softwar": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 17, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34, 41], "i": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 16, 17, 18, 19, 20, 21, 22, 23, 24, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 42, 43, 44], "AS": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "basi": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 21, 24, 26, 27, 28, 29, 30, 31, 33, 34], "warranti": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "OR": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "condit": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 19, 24, 26, 27, 28, 29, 30, 31, 33, 34], "OF": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "kind": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "express": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 19, 24, 26, 27, 28, 29, 30, 31, 33, 34], "impli": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "specif": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 19, 22, 24, 26, 27, 28, 29, 30, 31, 33, 34, 38], "languag": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34, 38], "govern": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 19, 20, 24, 26, 27, 28, 29, 30, 31, 33, 34], "permiss": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "limit": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "respons": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 17, 18, 22, 24, 26, 27, 28, 29, 30, 31, 33, 34], "check": [2, 3, 5, 6, 7, 9, 10, 11, 12, 16, 18, 19, 20, 21, 22, 24, 26, 27, 28, 29, 30, 31, 33, 34], "content": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34, 38, 39], "determin": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "suitabl": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 19, 24, 26, 27, 28, 29, 30, 31, 33, 34], "intend": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 17, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "notebook": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 16, 17, 18, 24, 26, 27, 28, 29, 30, 31, 32, 33, 34], "latest": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 21, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 38], "stabl": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34], "modern": 2, "internet": 2, "catalyst": 2, "human": 2, "decis": 2, "becaus": [2, 5, 6, 8, 16, 20, 22, 24, 26], "offlin": 2, "filter": [2, 5, 6, 14, 16, 22], "order": [2, 3, 5, 6, 8, 10, 14, 16, 17, 19, 22, 24, 26, 28, 33, 34, 38], "etc": [2, 19, 22, 38], "work": [2, 5, 6, 8, 15, 21, 24, 26], "togeth": [2, 19, 22, 40], "seamlessli": 2, "effici": [2, 3, 9, 10, 29], "biggest": 2, "practition": 2, "lack": 2, "understand": [2, 20], "around": 2, "look": [2, 3, 4, 5, 6, 7, 8, 10, 11, 14, 16, 23, 24, 26, 28, 38], "gap": 2, "between": [2, 3, 6, 8, 19, 20, 22, 24, 38], "exampl": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 15, 16, 17, 18, 19, 20, 21, 22, 24, 26, 27, 28, 29, 30, 31, 32, 33, 34, 38, 39], "simpl": [2, 7, 10, 22, 24, 38], "readi": [2, 3, 5, 6, 11, 12, 16, 18, 21, 24, 28, 31, 33, 34], "figur": [2, 3, 11, 20, 27, 38], "below": [2, 3, 5, 6, 9, 11, 12, 17, 18, 19, 20, 21, 24, 26, 33, 34], "repres": [2, 3, 5, 8, 10, 17, 19, 22, 24, 28], "four": [2, 5, 20, 27, 28, 30, 34], "than": [2, 5, 6, 10, 16, 19, 20, 22, 26, 28, 29, 38], "singl": [2, 5, 6, 8, 9, 10, 13, 14, 16, 17, 18, 21, 22, 28, 38], "much": [2, 15, 19, 20, 26], "realist": [2, 19], "closer": 2, "": [2, 3, 5, 6, 7, 8, 10, 11, 12, 14, 16, 17, 18, 24, 26, 27, 28, 29, 30, 33, 34, 38], "happen": [2, 22], "In": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 16, 17, 18, 19, 20, 21, 22, 24, 26, 28, 29, 31, 33, 34, 38, 39], "seri": [2, 3, 18], "we": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 26, 27, 28, 29, 30, 31, 32, 33, 34, 38, 39], "go": [2, 3, 8, 15, 16, 19, 21, 31, 33, 34], "showcas": [2, 20], "let": [2, 3, 5, 6, 8, 10, 11, 12, 21, 24, 28, 30, 33, 34], "concept": [2, 11, 13, 14, 25, 26], "briefli": 2, "narrow": 2, "down": [2, 8], "million": [2, 16, 19, 22], "thousand": [2, 22], "candid": [2, 3, 22, 31, 38], "relev": [2, 3, 16, 19, 22, 38], "top": [2, 3, 16, 19, 20, 22], "k": [2, 3, 19, 26], "exclud": 2, "alreadi": [2, 3, 6, 8, 16, 17, 19, 21, 22, 24, 26, 27, 28, 34], "interact": [2, 5, 11, 14, 15, 16, 19, 22, 29, 38], "undesir": 2, "appli": [2, 3, 6, 8, 24, 28, 33, 34, 38], "busi": [2, 24], "logic": [2, 24], "rule": [2, 38], "although": [2, 7, 10, 26], "skip": [2, 5, 6, 27, 33], "score": [2, 3, 12, 14, 18, 19, 22, 31, 38], "also": [2, 5, 8, 16, 18, 19, 20, 21, 22, 24, 26], "known": [2, 24], "here": [2, 3, 5, 6, 8, 10, 11, 15, 16, 19, 20, 21, 22, 23, 26, 28], "being": [2, 5, 6, 19, 23], "abl": [2, 3, 5, 6, 9, 10, 16, 17, 22, 26, 28, 29, 33, 34], "our": [2, 5, 6, 7, 10, 12, 16, 17, 18, 19, 20, 21, 22, 24, 26, 30, 31, 32, 33, 34, 38, 40], "At": [2, 3, 12, 17, 28, 38], "final": [2, 3, 5, 18, 19, 21, 24, 26, 30, 31, 38], "want": [2, 6, 7, 8, 9, 14, 16, 19, 20, 21, 22, 24, 26, 28, 31, 33, 34], "re": [2, 5, 6, 26, 28, 35, 36, 37, 42, 43, 44], "align": 2, "output": [2, 3, 4, 5, 6, 9, 10, 11, 12, 16, 17, 18, 21, 24, 26, 27, 28, 29, 30, 31, 33, 34, 38], "need": [2, 3, 5, 6, 8, 9, 10, 11, 12, 16, 18, 19, 21, 22, 24, 25, 26, 28, 29, 31, 33], "constraint": 2, "criteria": 2, "listen": 2, "even": [2, 5, 6, 31], "oldridg": 2, "move": [2, 3, 5, 6, 9, 11, 24], "talk": [2, 38], "kdd": 2, "21": [2, 5, 8, 9, 10, 24, 26, 35, 36, 37, 42, 43, 44], "read": [2, 3, 5, 6, 8, 11, 18, 24, 26, 28, 29, 31, 33, 34], "blog": [2, 28, 38], "post": [2, 5, 12, 18, 22, 24, 38], "approxim": [2, 3, 31, 38], "neighbour": 2, "ann": [2, 3, 14, 38], "addit": [2, 7, 8, 14, 19, 22, 31, 38], "client": [2, 5, 6, 12, 18, 24, 26, 27, 28, 31, 33, 34], "extern": [2, 7, 8, 9, 10, 11, 12, 19, 28], "faiss": [2, 4, 14, 26], "similar": [2, 4, 8, 9, 14, 19, 20, 23, 24, 28, 29, 38], "cluster": [2, 3, 27, 34], "dens": [2, 9, 10, 11, 22, 29], "vector": [2, 3, 6, 11, 22, 38], "find": [2, 3, 4, 16, 20, 21, 22, 23, 29], "tag": [2, 4, 5, 6, 8, 10, 11, 16, 18, 19, 22, 23, 24, 26, 28, 30, 31, 33, 34], "most": [2, 6, 8, 13, 16, 19, 21, 22, 23, 28, 38], "recent": [2, 22, 38], "releas": [2, 4, 8, 15, 23, 26, 35, 36, 37, 41, 42, 43, 44], "refer": [2, 3, 8, 19, 20, 21, 25, 28], "31": [2, 3, 5, 12, 26, 28, 35, 36, 37, 42, 43, 44], "uncom": [2, 26], "uninstal": [2, 21, 26], "o": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31, 33, 34], "nvt": [2, 6, 7, 8, 9, 10, 11, 12, 24, 26, 27, 28, 33, 34], "op": [2, 3, 5, 6, 8, 9, 11, 19, 24, 26, 28, 31, 33, 34], "renam": [2, 5, 11, 30, 33], "dropna": 2, "lambdaop": [2, 5, 8], "categorifi": [2, 5, 6, 8, 19, 24, 26, 28, 33, 34], "tagasuserfeatur": [2, 24, 26], "tagasuserid": [2, 8, 24, 26, 33], "tagasitemfeatur": [2, 24, 26], "tagasitemid": [2, 6, 8, 24, 26, 33], "addmetadata": [2, 24, 26, 28], "dag": [2, 3, 5, 6, 8, 11, 24, 26, 31, 33, 34], "subgraph": 2, "tf": [2, 3, 5, 6, 8, 10, 11, 24, 26, 30, 31], "mm": [2, 5, 6, 11, 24, 26, 30, 31], "ecommerc": [2, 5], "transform_aliccp": 2, "comment": [2, 19, 22], "out": [2, 3, 5, 6, 8, 10, 11, 14, 16, 18, 19, 21, 22, 23, 24, 28, 30, 31], "tf_gpu_alloc": [2, 5, 11, 16, 17, 18, 21, 22, 24, 30, 31], "cuda_malloc_async": [2, 5, 11, 16, 17, 18, 21, 22, 24, 30, 31], "06": [2, 3, 5, 6, 8, 24, 27, 28, 29, 30, 35, 36, 37, 41, 42, 43, 44], "29": [2, 3, 5, 6, 17, 18, 24, 26, 30, 33, 34, 35, 36, 37, 42, 43, 44], "19": [2, 3, 5, 26, 28, 35, 36, 37, 42, 43, 44], "49": [2, 35, 37, 42, 44], "32": [2, 3, 10, 11, 16, 21, 22, 24, 26, 27, 28, 30, 33], "836544": 2, "cpu_feature_guard": [2, 3, 5, 6, 11, 18, 24, 26, 30], "cc": [2, 3, 5, 6, 11, 17, 18, 24, 26, 30, 33, 34], "194": [2, 3, 5, 6, 11, 18, 26, 30], "oneapi": [2, 3, 5, 6, 11, 18, 26, 30], "neural": [2, 3, 5, 6, 7, 8, 9, 11, 18, 19, 22, 26, 28, 30, 38], "network": [2, 3, 5, 6, 7, 8, 9, 11, 16, 18, 19, 22, 26, 28, 30, 38], "onednn": [2, 3, 5, 6, 11, 18, 26, 30], "instruct": [2, 3, 5, 6, 11, 12, 17, 18, 24, 26, 30, 33, 34], "critic": [2, 3, 5, 6, 11, 18, 24, 26, 30], "sse3": [2, 3, 5, 6, 11, 18, 24, 26, 30], "sse4": [2, 3, 5, 6, 11, 18, 24, 26, 30], "1": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 18, 19, 20, 22, 23, 24, 25, 27, 28, 29, 30, 31, 33, 34, 35, 36, 37, 38, 41, 42, 43, 44], "avx": [2, 3, 5, 6, 11, 18, 24, 26, 30], "them": [2, 3, 5, 6, 8, 10, 11, 16, 17, 18, 19, 22, 24, 26, 28, 30, 31, 38, 39], "rebuild": [2, 3, 5, 6, 11, 18, 24, 26, 30], "appropri": [2, 3, 5, 6, 11, 18, 24, 26, 30], "compil": [2, 3, 5, 6, 9, 11, 18, 24, 26, 29, 30], "flag": [2, 3, 4, 5, 6, 11, 12, 18, 24, 26, 30], "usr": [2, 3, 5, 6, 8, 9, 10, 11, 12, 18, 24, 26, 27, 30, 33, 34], "local": [2, 3, 5, 6, 8, 9, 10, 11, 12, 14, 18, 21, 24, 26, 27, 28, 33, 34], "lib": [2, 3, 5, 6, 8, 9, 10, 11, 12, 18, 24, 26, 27, 30, 33, 34], "python3": [2, 3, 5, 6, 8, 9, 10, 11, 12, 18, 24, 26, 27, 30, 33, 34], "8": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 21, 24, 26, 27, 28, 29, 30, 33, 34, 35, 36, 37, 42, 43, 44], "dist": [2, 3, 5, 6, 8, 9, 10, 11, 12, 18, 24, 26, 27, 30, 33, 34], "packag": [2, 3, 5, 6, 8, 9, 10, 11, 12, 18, 21, 24, 25, 26, 27, 30, 33, 34, 35, 36, 37, 42, 43, 44], "dtype": [2, 3, 5, 6, 8, 9, 10, 11, 12, 18, 24, 27, 28, 30, 34], "torch": [2, 3, 5, 6, 10, 18, 24], "py": [2, 3, 5, 6, 8, 9, 10, 11, 12, 16, 17, 18, 19, 21, 22, 24, 26, 27, 29, 30, 33, 34], "43": [2, 3, 5, 6, 18, 24, 34, 37, 44], "userwarn": [2, 3, 5, 6, 8, 10, 18, 24, 26, 33, 34], "did": [2, 3, 5, 6, 8, 10, 18, 24, 33, 34], "successfulli": [2, 3, 5, 6, 8, 10, 18, 24, 26], "due": [2, 3, 5, 6, 8, 10, 18, 20, 24, 26], "error": [2, 3, 5, 6, 8, 10, 16, 18, 19, 24, 26], "No": [2, 3, 5, 6, 8, 10, 11, 18, 24, 26], "warn": [2, 3, 5, 6, 8, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 33, 34], "f": [2, 3, 5, 6, 8, 10, 11, 17, 18, 24, 26, 28, 29], "exc": [2, 3, 5, 6, 8, 10, 18, 24], "msg": [2, 3, 5, 6, 8, 10, 18, 24], "fix": [2, 5, 6, 11, 22, 24, 28], "python": [2, 3, 5, 6, 9, 11, 14, 16, 17, 19, 21, 22, 23, 24, 27, 29, 30], "track": [2, 5, 6, 11, 21, 24], "data_structur": [2, 5, 6, 11, 24], "ha": [2, 3, 5, 6, 7, 8, 10, 11, 16, 19, 20, 22, 24, 27, 33, 34, 38], "been": [2, 5, 6, 8, 11, 12, 19, 22, 24, 26], "trackabl": [2, 5, 6, 11, 24], "old": [2, 5, 6, 7, 11, 24], "delet": [2, 5, 6, 11, 24, 27], "11": [2, 5, 6, 11, 24, 26, 29, 30, 35, 36, 37, 42, 43, 44], "info": [2, 3, 5, 6, 11, 22, 24, 26, 27, 28, 29, 30, 34], "sparse_operation_kit": [2, 5, 6, 24], "sok": [2, 5, 6, 24], "merlin_sok": [2, 5, 6, 24, 26], "4": [2, 5, 6, 7, 8, 10, 14, 16, 18, 22, 24, 26, 28, 29, 30, 32, 33, 34, 35, 36, 37, 38, 42, 43, 44], "py3": [2, 5, 6, 24, 26], "linux": [2, 5, 6, 24, 26], "x86_64": [2, 5, 6, 24, 26], "egg": [2, 5, 6, 24, 26], "libsok_experi": [2, 5, 6, 24], "37": [2, 3, 26, 29], "094972": 2, "38": [2, 6, 26, 28, 35, 42], "134481": 2, "w": [2, 5, 6, 20, 21, 24, 29], "common_runtim": [2, 3, 5, 6, 11, 24, 26, 30], "gpu_bfc_alloc": [2, 5, 6, 24], "42": [2, 5, 6, 22, 33, 35, 37, 42, 44], "overrid": [2, 5, 6, 19, 24], "orig_valu": [2, 5, 6, 24], "tf_force_gpu_allow_growth": [2, 5, 6, 24], "variabl": [2, 3, 5, 6, 16, 21, 24, 26, 27, 28], "origin": [2, 5, 6, 7, 10, 11, 19, 22, 24, 27], "config": [2, 3, 5, 6, 11, 12, 17, 24, 26, 31], "valu": [2, 3, 5, 6, 10, 11, 18, 19, 21, 22, 24, 26, 27, 28, 31, 33, 34], "wa": [2, 5, 6, 8, 11, 16, 19, 20, 24, 26, 27, 31, 33, 34, 38], "134526": 2, "gpu_devic": [2, 3, 5, 6, 11, 24, 26, 30], "1621": [2, 3, 5, 6], "devic": [2, 3, 5, 6, 8, 10, 11, 19, 22, 24, 26, 27, 28, 29, 30], "job": [2, 3, 5, 6, 11, 20, 21, 24, 26, 30], "localhost": [2, 3, 5, 6, 11, 12, 18, 24, 26, 30, 31, 33, 34], "replica": [2, 3, 5, 6, 11, 24, 26, 30], "24576": [2, 3, 5, 6, 11], "mb": [2, 3, 5, 6, 11, 24, 26, 30], "quadro": [2, 3, 5, 6, 11, 24], "rtx": [2, 3, 5, 6, 11], "8000": [2, 3, 4, 5, 6, 11, 12, 17, 18, 24, 33, 34], "pci": [2, 3, 5, 6, 11, 24, 26, 30], "bu": [2, 3, 5, 6, 11, 24, 26, 30], "id": [2, 3, 5, 6, 8, 10, 11, 12, 16, 19, 21, 22, 24, 26, 27, 30, 38], "0000": [2, 3, 5, 6, 11, 24, 26, 30], "15": [2, 3, 8, 16, 24, 28, 29, 33, 35, 36, 37, 42, 43, 44], "00": [2, 3, 5, 6, 7, 10, 11, 24, 26, 28, 30, 33, 34, 35, 36, 37, 42, 43, 44], "comput": [2, 3, 5, 6, 8, 11, 19, 20, 21, 22, 24, 26, 28, 30, 33, 34, 38], "capabl": [2, 3, 5, 6, 11, 14, 24, 26, 30], "7": [2, 3, 5, 6, 8, 10, 11, 16, 18, 19, 24, 26, 27, 28, 30, 33, 34, 35, 36, 37, 41, 42, 43, 44], "5": [2, 3, 5, 6, 7, 8, 11, 16, 18, 19, 20, 22, 24, 26, 28, 30, 33, 34, 35, 36, 37, 42, 43, 44], "135533": 2, "135562": 2, "2d": [2, 3, 24], "tqdm": [2, 5, 8, 9, 10, 11, 24, 33], "auto": [2, 5, 8, 9, 10, 11, 24, 33], "tqdmwarn": [2, 5, 8, 9, 10, 11, 24, 33], "iprogress": [2, 5, 8, 9, 10, 11, 24, 33], "updat": [2, 3, 5, 6, 8, 9, 10, 11, 24, 26, 31, 33, 38], "ipywidget": [2, 5, 8, 9, 10, 11, 24, 33], "readthedoc": [2, 5, 8, 9, 10, 11, 24, 33], "en": [2, 5, 8, 9, 10, 11, 24, 33], "user_instal": [2, 5, 8, 9, 10, 11, 24, 33], "autonotebook": [2, 5, 8, 9, 10, 11, 24, 33], "notebook_tqdm": [2, 5, 8, 9, 10, 11, 24, 33], "initi": [2, 3, 5, 6, 8, 16, 19, 22, 24, 26, 27, 28, 30, 34], "finish": [2, 3, 5, 6, 17, 21, 24], "commun": [2, 3, 5, 6, 24, 26, 30, 38], "horovod": [2, 5, 6, 24], "disabl": 2, "debug": [2, 22, 24, 26], "log": [2, 4, 14, 16, 19, 21, 23, 26, 28, 30, 32], "everywher": 2, "synthet": [2, 22, 24, 26, 38], "mimick": [2, 24, 26], "ali": [2, 24, 26], "ccp": [2, 24, 26], "alibaba": [2, 24, 26], "click": [2, 3, 9, 14, 16, 18, 19, 20, 21, 22, 24, 26, 28, 30, 32, 38], "convers": [2, 14, 24, 26], "first": [2, 3, 5, 6, 7, 8, 10, 12, 16, 18, 19, 20, 21, 22, 24, 27, 28, 29, 31], "path": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 16, 17, 18, 19, 21, 22, 24, 26, 27, 28, 29, 30, 31], "repo": [2, 3], "data_fold": [2, 3, 24, 26], "workspac": [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 24, 26, 34], "dir": [2, 3, 27], "base_dir": [2, 3, 27, 28, 29, 31], "multi": [2, 7, 8, 10, 13, 15, 18, 19, 25, 26, 27, 28, 38], "Then": [2, 21, 22], "generate_data": [2, 6, 24, 26], "util": [2, 3, 5, 6, 7, 10, 11, 18, 24, 26, 27, 34], "num_row": [2, 24, 26], "100_000": [2, 6], "train_raw": 2, "valid_raw": 2, "aliccp": [2, 24, 26], "raw": [2, 3, 5, 8, 16, 17, 18, 19, 24, 31], "int": [2, 3, 5, 6, 7, 10, 24, 26, 28, 30], "set_siz": [2, 24, 26], "3": [2, 5, 6, 7, 8, 9, 10, 12, 16, 18, 22, 24, 27, 28, 30, 31, 33, 34, 35, 36, 37, 38, 42, 43, 44], "If": [2, 5, 6, 16, 18, 19, 21, 22, 24, 26, 27, 29, 30, 31, 33, 34, 38], "would": [2, 5, 6, 19, 22, 24, 26, 33, 34, 38], "get_aliccp": [2, 24, 26], "instead": [2, 4, 12, 20, 22, 24, 26, 34], "csv": [2, 5, 6, 7, 8, 16, 18, 19, 22, 24, 26], "parquet": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 19, 22, 24, 26, 28, 29, 30, 31], "directli": [2, 3, 5, 6, 11, 24, 26, 30], "fed": [2, 10, 19, 22, 38], "abov": [2, 3, 5, 6, 9, 12, 21, 24, 38], "onto": [2, 11, 26], "repositori": [2, 3, 5, 6, 8, 9, 10, 11, 12, 17, 18, 24, 25, 26, 29, 31, 33, 34, 40], "customiz": 2, "infrastructur": [2, 38], "manag": [2, 16, 20, 21, 33, 34], "time": [2, 3, 5, 6, 8, 10, 12, 16, 19, 20, 21, 22, 24, 26, 27, 28, 29, 30], "current": [2, 7, 8, 15, 19, 22, 23, 24, 26, 31, 38], "directori": [2, 4, 6, 7, 8, 9, 10, 11, 12, 14, 23, 24, 26, 27, 28, 29, 30, 31, 34], "u": [2, 5, 6, 11, 24, 26, 28, 33, 34], "rm": [2, 9, 14, 16, 21, 23, 25, 29], "rf": [2, 9, 29], "feast_repo": [2, 3], "cd": [2, 3, 16, 17, 19, 21, 22, 26], "init": [2, 6, 21, 29], "1m": [2, 3], "32m": 2, "raid": [2, 3, 27, 28, 29, 30, 31], "workshar": [2, 3], "0m": [2, 3, 26], "should": [2, 3, 4, 6, 8, 9, 12, 14, 19, 20, 21, 22, 23, 27, 38], "messag": [2, 3, 4, 5, 6, 14, 17, 18, 21, 23, 24, 33, 34], "print": [2, 3, 5, 10, 12, 18, 21, 24, 26, 28, 29, 31], "now": [2, 3, 5, 6, 8, 9, 11, 12, 17, 18, 21, 24, 28, 33, 34], "navig": [2, 4, 14, 23], "feature_repo": [2, 3], "folder": [2, 3, 5, 9, 15, 16, 17, 18, 21, 22, 23, 24, 27], "remov": [2, 3, 4, 5, 8, 12, 16, 18, 19, 24, 26, 27, 31], "demo": [2, 26], "default": [2, 5, 6, 19, 21, 22, 26, 27], "feature_repo_path": 2, "join": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 18, 24, 26, 27, 28, 29, 30, 31], "example_repo": 2, "driver_stat": 2, "unique_rows_by_featur": 2, "user_featur": [2, 3, 19, 24, 26], "user_id": [2, 3, 5, 8, 16, 18, 19, 22, 24, 26, 33, 38], "reset_index": [2, 18, 33], "drop": [2, 6, 7, 18, 33, 34], "true": [2, 3, 5, 6, 8, 9, 10, 12, 16, 18, 22, 24, 26, 27, 28, 29, 30, 33], "artifici": 2, "add": [2, 3, 5, 8, 9, 12, 15, 19, 21, 22, 24, 29], "datetim": 2, "timestamp": [2, 7, 19, 38], "column": [2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 19, 22, 24, 26, 27, 28, 31, 33, 34, 38], "creation": [2, 12, 18, 24, 26], "when": [2, 3, 4, 5, 6, 8, 9, 14, 16, 17, 18, 19, 20, 21, 22, 23, 24, 26, 28, 29, 33, 34, 38], "astyp": [2, 3, 5, 6, 8], "datetime64": 2, "n": [2, 8, 12, 16, 19, 22, 26, 28], "user_shop": [2, 3, 24, 26], "user_profil": [2, 3, 24, 26], "user_group": [2, 3, 24, 26], "user_gend": [2, 3, 22, 24, 26], "user_ag": [2, 3, 24, 26, 38], "user_consumption_1": 2, "user_consumption_2": [2, 3, 24, 26], "user_is_occupi": [2, 3, 24, 26], "user_geographi": [2, 3, 24, 26], "user_intent": [2, 3, 24, 26], "user_brand": [2, 3, 24, 26], "user_categori": [2, 3, 24, 26], "6": [2, 5, 6, 8, 10, 16, 18, 21, 22, 24, 26, 28, 33, 35, 36, 37, 42, 43, 44], "590": 2, "171": 2, "293": 2, "50": [2, 3, 5, 6, 17, 18, 22, 24, 28, 33, 34, 35, 36, 37, 42, 43, 44], "300270": 2, "303330": 2, "to_parquet": [2, 5, 6, 7, 8, 24, 26, 27, 28], "item_featur": [2, 3, 19, 22, 24, 26], "item_id": [2, 3, 5, 8, 16, 18, 19, 22, 24, 26], "item_categori": [2, 3, 24, 26], "item_shop": [2, 3, 24, 26], "item_brand": [2, 3, 24, 26], "item_intent": 2, "410715": 2, "412307": 2, "412": 2, "142": 2, "66": [2, 24], "12": [2, 5, 24, 26, 27, 28, 30, 35, 36, 37, 42, 43, 44], "824": 2, "284": 2, "132": [2, 26], "18": [2, 3, 5, 6, 17, 18, 24, 26, 28, 33, 34, 36, 37, 43, 44], "1236": 2, "426": 2, "197": [2, 34], "24": [2, 5, 26, 27, 28, 29, 30, 34], "1648": 2, "568": 2, "263": 2, "disk": [2, 5, 6, 7, 8, 11, 16, 27, 28, 30, 31, 33, 34], "output_path": [2, 8, 16, 17, 19, 21, 22, 24, 26, 27, 28], "processed_nvt": [2, 3], "notic": [2, 16, 20, 21], "row": [2, 5, 6, 8, 12, 16, 18, 20, 22, 31, 38], "miss": [2, 6, 19, 28], "after": [2, 3, 4, 5, 6, 9, 10, 11, 14, 16, 17, 18, 20, 21, 22, 23, 24, 25, 26, 30, 31, 33, 34], "doe": [2, 5, 18, 20, 24, 35, 36, 37, 42, 43, 44], "null": [2, 19, 26], "entri": 2, "might": [2, 6, 14, 16, 17, 19, 20, 21, 22, 26], "own": [2, 15, 16, 20, 21, 25, 26], "therefor": [2, 24], "while": [2, 3, 5, 6, 8, 11, 20, 21, 22, 24, 26, 30], "regist": [2, 6, 16], "user_id_raw": [2, 3], "item_id_raw": 2, "avoid": [2, 5, 6, 16, 17, 19, 22], "potenti": 2, "occur": [2, 22, 28], "postfix": 2, "_raw": 2, "lambda": [2, 5, 6, 8, 19], "col": [2, 5, 8, 12, 24, 27, 28, 31, 33, 34], "int32": [2, 3, 10, 18, 19, 24, 26, 27, 28, 31, 34], "item_cat": 2, "subgraph_item": 2, "items_cat": 2, "subgraph_us": 2, "target": [2, 3, 5, 6, 8, 11, 16, 18, 20, 22, 24, 26, 28, 30, 31, 34, 38], "binary_classif": [2, 24, 26, 28], "nvt_wkflow": 2, "call": [2, 3, 5, 6, 8, 17, 20, 26, 38], "fit": [2, 5, 6, 8, 9, 11, 22, 24, 26, 28, 29, 30, 33, 34], "nvt_workflow": [2, 3], "workflow_nam": 2, "visit": [2, 5, 6, 23, 24], "05": [2, 5, 8, 10, 16, 18, 21, 22, 33, 34, 35, 36, 37, 42, 43, 44], "ipynb": [2, 3, 17, 24], "categor": [2, 5, 6, 7, 8, 10, 13, 14, 16, 19, 21, 24, 27, 28], "encod": [2, 7, 10, 22, 29], "output_path2": 2, "train_tt": 2, "valid_tt": 2, "valid": [2, 5, 8, 9, 10, 11, 12, 18, 22, 24, 26, 27, 28, 29, 30], "select": [2, 5, 16, 22, 28], "posit": [2, 5, 16, 19, 20, 22, 38], "where": [2, 3, 5, 16, 19, 20, 21, 22, 26, 28, 38], "column_nam": [2, 3, 5, 6, 8, 11, 18, 24, 26, 30, 31, 33, 34], "df": [2, 5, 6, 8], "pbtxt": [2, 3, 16, 24, 26], "protobuf": [2, 26], "text": [2, 8, 12, 18, 24, 26, 27], "02": [2, 9, 18, 24, 35, 36, 37, 42, 43, 44], "part_siz": [2, 6, 8, 28], "500mb": [2, 6], "select_by_tag": [2, 5, 6, 11, 24, 26, 30, 31], "model_tt": 2, "twotowermodel": 2, "query_tow": [2, 3], "mlpblock": [2, 5, 6, 11, 22, 24, 26, 30], "128": [2, 5, 6, 9, 10, 16, 21, 22, 24, 26, 29, 30], "64": [2, 5, 6, 11, 12, 16, 21, 22, 24, 26, 28, 29, 30], "no_activation_last_lay": [2, 5, 6], "sampler": 2, "inbatchsampl": 2, "embedding_opt": 2, "embeddingopt": 2, "infer_embedding_s": 2, "adam": [2, 5, 6, 9, 10, 22, 24, 26], "run_eagerli": [2, 5, 6, 11, 24, 26, 30], "fals": [2, 5, 6, 7, 8, 9, 10, 11, 18, 19, 24, 26, 29, 30], "categorical_crossentropi": [2, 5, 6], "metric": [2, 5, 6, 11, 17, 18, 20, 21, 22, 24, 26, 33, 34], "recallat": 2, "10": [2, 3, 9, 16, 18, 19, 21, 22, 24, 26, 27, 28, 33, 34, 35, 36, 37, 42, 43, 44], "ndcgat": 2, "validation_data": [2, 5, 11, 24, 26, 30], "batch_siz": [2, 5, 6, 10, 11, 24, 26, 30], "1024": [2, 10, 11, 22, 26, 29, 30], "epoch": [2, 5, 6, 10, 16, 21, 22, 26, 30], "kera": [2, 5, 6, 9, 11, 14, 22, 24, 26, 29, 30, 31], "initializers_v2": [2, 5, 6], "120": [2, 5, 6], "truncatednorm": [2, 5, 6], "unseed": [2, 5, 6], "return": [2, 3, 5, 6, 8, 10, 11, 12, 19, 24, 26, 33, 34], "ident": [2, 5, 6, 26], "seed": [2, 5, 6, 22, 29], "same": [2, 3, 5, 6, 8, 16, 17, 18, 19, 21, 22, 24, 27, 28, 31, 33, 34], "instanc": [2, 5, 6, 21, 24, 26], "onc": [2, 3, 5, 6, 8, 17, 26, 28, 33, 34], "9": [2, 6, 11, 18, 24, 26, 28, 33, 35, 36, 37, 42, 43, 44], "275m": 2, "9538": 2, "recall_at_10": [2, 5, 6], "0101": 2, "ndcg_at_10": [2, 5, 6], "0067": 2, "regularization_loss": [2, 5, 6, 11, 24, 26, 30], "0000e": [2, 5, 6, 11, 24, 26, 30], "loss_batch": [2, 5, 6, 11, 24, 30], "8711": 2, "val_loss": [2, 11, 24, 26, 30], "9179": 2, "val_recall_at_10": 2, "0212": 2, "val_ndcg_at_10": 2, "0155": 2, "val_regularization_loss": [2, 11, 24, 26, 30], "val_loss_batch": [2, 11, 24, 30], "5806": [2, 5], "callback": [2, 5, 6, 11, 24, 30], "histori": [2, 5, 6, 11, 24, 30, 38], "0x7fd4b04139d0": 2, "later": [2, 5, 6, 8, 19, 22, 26, 35, 36, 37, 42, 43, 44], "deploy": [2, 5, 9, 15, 17, 18, 20, 24, 32, 38], "retrieval_block": 2, "query_block": 2, "target_column": [2, 11, 24, 26], "propos": [2, 11, 22], "facebook": [2, 9, 26, 29, 30, 38], "2019": 2, "introduc": [2, 3, 19, 22], "person": [2, 22, 38], "spars": [2, 22, 28], "multilay": 2, "perceptron": [2, 22, 38], "mlp": [2, 20, 21, 22, 38], "explicitli": [2, 22], "statist": [2, 8, 16, 19, 28], "techniqu": [2, 5, 6, 19, 22, 38], "architetctur": 2, "gh": 2, "dlrmmodel": [2, 14, 22, 24, 26, 30], "embedding_dim": [2, 22, 24, 26, 30], "bottom_block": [2, 24, 26, 30], "top_block": [2, 24, 26, 30], "prediction_task": [2, 11, 24, 26, 30], "binaryclassificationtask": [2, 26, 30], "auc": [2, 11, 20, 21, 24, 26, 29, 30], "16": [2, 5, 6, 8, 9, 10, 11, 18, 24, 26, 27, 35, 36, 37, 42, 43, 44], "305m": 2, "6932": [2, 24, 26], "5005": 2, "6931": [2, 24, 26], "val_auc": [2, 11, 24, 26, 30], "5029": 2, "0x7fd449398a30": 2, "back": [2, 22, 24, 38], "cell": [2, 3, 5, 17, 24, 26], "predicttensorflow": [2, 3, 5, 6, 11, 24, 26, 31], "transformworkflow": [2, 3, 5, 6, 11, 24, 26, 31, 33, 34], "get_subworkflow": [2, 3], "item_block": 2, "item_embed": [2, 3], "fit_transform": [2, 5, 6, 33, 34], "to_ddf": [2, 24, 26], "tail": [2, 19], "output_1": [2, 3], "453": 2, "945": 2, "012117806822061539": 2, "02241620607674122": 2, "454": 2, "948": 2, "455": 2, "956": 2, "456": [2, 6], "1437": 2, "457": 2, "1469": 2, "timedelta": 2, "entiti": [2, 3], "field": [2, 26, 33, 34], "featureview": [2, 3], "valuetyp": 2, "type": [2, 8, 12, 16, 18, 22, 24, 26, 28, 38], "infra": 2, "offline_stor": 2, "file_sourc": 2, "filesourc": 2, "timestamp_field": 2, "created_timestamp_column": 2, "value_typ": 2, "join_kei": 2, "user_features_view": 2, "ttl": 2, "onlin": [2, 6, 22, 26, 38], "dict": [2, 5, 6, 10, 26], "format": [2, 7, 8, 9, 19, 22, 26, 27, 28], "close": [2, 20], "item_features_view": 2, "checkout": [2, 5], "structur": [2, 3, 4, 24, 27], "seedir": [2, 3, 24], "satisfi": [2, 24, 26], "natsort": [2, 24], "sd": [2, 3, 24], "style": [2, 3, 24], "itemlimit": [2, 3, 24], "depthlimit": [2, 3, 24], "exclude_fold": [2, 3, 24], "ipynb_checkpoint": [2, 3, 24], "sort": [2, 3, 5, 8, 10, 21, 24, 30, 31, 38], "readm": [2, 3, 18], "md": [2, 3, 12], "__init__": [2, 3, 12, 26, 30], "__pycache__": [2, 3], "cpython": 2, "pyc": 2, "test_workflow": [2, 3], "feature_stor": [2, 3], "yaml": [2, 3, 21], "ti": [2, 3, 5, 6, 11, 24, 31, 33, 34], "ensembl": [2, 4, 5, 6, 11, 12, 14, 18, 26], "result": [2, 7, 8, 10, 16, 18, 19, 26, 33, 34, 38], "given": [2, 3, 5, 6, 7, 10, 11, 12, 18, 19, 20, 22, 33, 34, 38], "request": [2, 3, 12, 14, 16, 18, 24, 30, 31, 33, 34, 38], "user": [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 18, 19, 22, 24, 26, 27, 28, 29, 30, 31, 33, 34, 38], "point": [3, 9, 10, 26, 28], "reach": 3, "expect": [3, 8, 19, 22, 24, 33, 34], "execut": [3, 4, 5, 9, 14, 16, 17, 18, 21, 23, 24, 26, 28, 29, 31, 33, 34], "01": [3, 7, 10, 11, 21, 26, 28, 30, 35, 36, 37, 42, 43, 44], "queri": [3, 16, 18, 19, 26, 33, 34, 38], "describ": [3, 17, 19, 20, 22, 26, 28], "built": [3, 5, 6, 16, 20, 23, 26], "usag": [3, 15], "assum": [3, 21, 22, 25, 26, 30], "feast": [3, 4, 14], "script": [3, 9, 14, 15, 16, 17, 18, 20, 21], "numpi": [3, 5, 6, 8, 11, 18, 24, 26, 27, 28, 33, 34], "np": [3, 5, 6, 8, 11, 18, 24, 26, 27, 28, 33, 34], "panda": [3, 6, 7, 8, 12, 18, 24, 26, 33, 34], "pd": [3, 6, 18], "columnschema": [3, 6, 12, 24], "softmax_sampl": 3, "softmaxsampl": 3, "unroll_featur": 3, "unrollfeatur": 3, "send_triton_request": [3, 18], "metadata": [3, 7, 8, 10, 11, 16, 24, 26, 33, 34], "orc_column_statistics_pb2": 3, "deprecationwarn": [3, 12], "deprec": [3, 5, 8, 12, 26, 27], "filedescriptor": 3, "note": [3, 5, 9, 10, 12, 16, 17, 18, 26, 27, 33, 34, 35, 36, 37, 42, 43, 44], "unlink": 3, "descriptor": 3, "awai": [3, 8, 24], "descriptor_pool": 3, "_descriptor": 3, "fielddescriptor": 3, "30": [3, 5, 16, 19, 24, 26], "_integerstatist": 3, "56": [3, 29], "885234": 3, "tensor_shape_pb2": 3, "36": [3, 6, 24], "_tensorshapeproto_dim": 3, "types_pb2": 3, "33": [3, 35, 36, 37, 42, 43, 44], "enumvaluedescriptor": 3, "27": [3, 26, 27, 28, 34, 35, 36, 37, 42, 43, 44], "enumdescriptor": 3, "_datatyp": 3, "287": 3, "280": 3, "_serializeddtyp": 3, "resource_handle_pb2": 3, "20": [3, 6, 8, 11, 26, 27, 30, 34, 35, 36, 37, 42, 43, 44], "39": [3, 6, 11, 34, 36, 43], "futur": [3, 5, 8, 12, 19, 26, 27, 31], "merlinload": 3, "registri": 3, "central": 3, "definit": [3, 24], "relat": [3, 20], "organ": [3, 5, 6, 22], "those": [3, 15, 16, 19, 22, 24, 33, 34], "both": [3, 6, 8, 19, 24, 38], "tell": 3, "command": [3, 4, 5, 6, 9, 12, 14, 16, 18, 21, 23, 24, 25, 26, 30, 31, 33, 34], "direct": [3, 8], "view": [3, 4, 14, 21, 23, 24, 38, 41], "sqlite": 3, "online_stor": 3, "db": 3, "feast_repo_path": 3, "32muser_id": 3, "32mitem_id": 3, "32muser_featur": 3, "32mitem_featur": 3, "32mfeast_repo_item_featur": 3, "32mfeast_repo_user_featur": 3, "materi": [3, 8], "done": [3, 18, 24, 29], "keep": [3, 16, 19, 20, 22, 28], "date": [3, 19, 26], "As": [3, 8, 9, 10, 16, 20, 21, 29, 31, 38], "continu": [3, 5, 8, 10, 19, 22, 24, 26, 28], "1995": [3, 7, 34], "2025": 3, "some": [3, 6, 16, 18, 19, 20, 21, 22, 24, 25, 26, 28, 30, 33, 34, 38], "01t01": 3, "32m2": 3, "32m1995": 3, "32m2025": 3, "32msqlite": 3, "100": [3, 5, 7, 9, 16, 21, 22, 28, 29, 33, 34], "460": [3, 35, 36, 37, 42, 43, 44], "2521": 3, "27it": 3, "458": 3, "3335": 3, "12it": 3, "again": [3, 26], "ran": [3, 18, 20], "isdir": [3, 28], "faiss_index": 3, "makedir": [3, 11, 27, 28], "faiss_index_path": 3, "retrieval_model_path": 3, "ranking_model_path": 3, "dlrm": [3, 9, 16, 20, 21, 23, 26, 29, 31, 38], "queryfaiss": 3, "interfac": [3, 21], "nearbi": 3, "setup_faiss": 3, "l2": [3, 6, 16, 22], "distanc": [3, 19], "read_parquet": [3, 8, 12, 18, 24, 26, 31], "embedding_column": 3, "centroid": 3, "least": [3, 5, 6, 11, 19, 26, 33, 34], "1248": 3, "featurestor": 3, "fetch": 3, "queryfeast": 3, "ensur": [3, 5, 6, 8, 11, 18, 19, 21, 24, 31, 33, 34], "correctli": [3, 21, 24, 26], "tritonserv": [3, 5, 6, 12, 17, 18, 24, 26, 31, 33, 34], "user_attribut": 3, "from_feature_view": 3, "include_id": 3, "32m1": 3, "32m2023": 3, "51": [3, 5, 24, 26, 35, 36, 37, 42, 43, 44], "0it": 3, "user_subgraph": 3, "backend": [3, 12, 31, 38], "prevent": 3, "claim": [3, 26], "tf_util": [3, 5, 6], "configure_tensorflow": 3, "dlpack": 3, "from_dlpack": 3, "dlcapsul": 3, "topk_retriev": 3, "topk": [3, 38], "07": [3, 24, 33, 35, 36, 37, 42, 43, 44], "269579": 3, "430459": 3, "431356": 3, "absl": [3, 5, 6, 11, 24, 26, 30], "untrac": [3, 5, 6, 11, 24, 26, 30], "restored_function_bodi": 3, "save": [3, 4, 5, 6, 7, 8, 9, 12, 16, 17, 18, 19, 20, 21, 22, 26, 28], "show": [3, 4, 5, 6, 7, 11, 14, 17, 18, 21, 22, 23, 24, 25, 26, 28, 30, 33, 34], "52": [3, 5, 8, 10, 26, 27, 28, 35, 37, 42, 44], "callabl": [3, 5, 6, 11, 24, 26, 30], "asset": [3, 5, 6, 11, 22, 24, 26, 30], "written": [3, 5, 6, 9, 11, 24, 29, 30, 33, 34], "tmp": [3, 5, 6, 11, 21, 24, 26], "tmpdalflmaz": 3, "item_attribut": 3, "candidate_id": 3, "output_prefix": 3, "14": [3, 5, 24, 26, 27, 28, 35, 36, 37, 42, 43, 44], "item_subgraph": 3, "merg": [3, 5, 6], "were": [3, 5, 16, 18, 19, 20, 22, 41], "unrol": 3, "broadcast": 3, "user_features_to_unrol": 3, "combined_featur": 3, "feed": [3, 6, 8, 10, 11, 24], "98": [3, 21, 24], "tmpqdd_jn5e": 3, "descend": 3, "random": [3, 5, 7, 16, 19, 22, 28], "sampl": [3, 4, 5, 7, 10, 14, 19, 24, 27, 28, 30, 33, 38], "softmax": [3, 16, 22], "top_k": 3, "relevance_col": 3, "binary_classification_task": [3, 26], "temperatur": [3, 5, 6, 22], "00000001": 3, "last": [3, 5, 12, 17, 19, 24, 27, 28, 31, 33, 34, 38], "artifact": [3, 16, 17, 31], "consum": [3, 6, 24], "class": [3, 8, 16, 17, 19, 20, 24], "entir": [3, 22, 24], "It": [3, 5, 6, 8, 9, 10, 11, 15, 16, 19, 20, 21, 22, 24, 26, 27, 28, 29, 30, 31, 33, 34, 38], "interpret": [3, 17, 24], "correct": [3, 6, 12, 17, 24], "suppli": [3, 24], "list": [3, 5, 6, 7, 8, 9, 18, 19, 21, 22, 24, 28, 38], "poc_ensembl": 3, "send": [3, 5, 6, 12, 18, 24, 31, 33, 34], "request_schema": [3, 12], "export_path": [3, 5, 6, 11, 24, 31, 33, 34], "ens_config": 3, "node_config": 3, "output_schema": [3, 5, 6, 8, 24, 26, 31, 33, 34], "ordered_id": 3, "ordered_scor": 3, "0_transformworkflowtriton": [3, 18, 24], "categori": [3, 5, 8, 19, 24, 26, 28], "uniqu": [3, 5, 8, 24, 26, 33], "json": [3, 8, 9, 12, 18, 24, 26, 29], "pkl": [3, 24, 26], "1_predicttensorflowtriton": [3, 5, 6, 18, 24], "savedmodel": [3, 5, 6, 11, 24, 26, 30], "fingerprint": [3, 24], "pb": [3, 24, 26], "keras_metadata": [3, 24, 26], "saved_model": [3, 5, 6, 22, 24, 26], "00000": [3, 24, 26], "00001": [3, 24, 26], "2_transformworkflowtriton": 3, "3_predicttensorflowtriton": 3, "input_schema": [3, 5, 6, 18, 24, 26, 31, 33, 34], "executor_model": [3, 5, 6, 18, 24, 26, 31, 33, 34], "termin": [3, 4, 5, 6, 9, 14, 17, 18, 23, 24, 33, 34], "ensemble_export_path": [3, 17, 18, 33, 34], "argument": [3, 5, 6, 8, 16, 18, 24, 26, 27, 28, 31, 33, 34], "specifi": [3, 4, 5, 6, 8, 17, 18, 19, 21, 24, 26, 27, 28, 29, 31, 33, 34], "previous": [3, 5, 6, 17, 18, 24, 31, 33, 34], "method": [3, 5, 6, 19, 20, 21, 22, 24, 26, 28, 31, 33, 34], "statu": [3, 12, 18, 24, 26, 28], "dispatch": [3, 5, 7, 8, 12, 24, 26, 31], "make_df": 3, "sent": [3, 16, 24, 26], "per": [3, 7, 8, 9, 20, 22, 24, 38], "arrai": [3, 5, 6, 18], "168": 3, "324": 3, "79": [3, 6, 26], "361": 3, "294": 3, "267": [3, 26], "289": 3, "397": 3, "189": [3, 6], "5016385": 3, "50176895": 3, "5017176": 3, "5024097": 3, "5018236": 3, "5018286": 3, "50162375": 3, "5015677": 3, "50175667": 3, "5014358": 3, "float32": [3, 5, 6, 8, 11, 18, 19, 28, 30], "That": [3, 16, 19, 21, 22, 28], "two": [4, 5, 8, 10, 11, 20, 22, 24, 38], "merlin": [4, 5, 6, 7, 8, 9, 11, 12, 17, 19, 20, 21, 22, 27, 28, 29, 32, 35, 36, 37, 38, 39], "retriev": [4, 14, 15, 19, 22, 31, 39], "export": [4, 5, 6, 11, 16, 17, 18, 26, 31, 33, 34], "index": [4, 10, 12, 14, 18, 24, 26, 28, 38], "pull": [4, 14, 16, 21, 23, 26], "nvcr": [4, 16, 21, 23, 26], "nightli": 4, "p": [4, 9, 10, 14, 16, 23, 25, 28, 29], "8001": [4, 5, 6, 12, 17, 18, 24, 31, 33, 34], "8002": [4, 5, 6, 17, 18, 24, 33, 34], "8888": [4, 14, 23, 25], "v": [4, 16, 21, 25, 26], "ipc": [4, 14, 16, 21, 23, 25], "host": [4, 12, 14, 16, 21, 23, 25, 26, 28, 34], "bin": [4, 14, 16, 19, 21, 22, 23, 25, 26], "bash": [4, 14, 16, 21, 23, 25, 26], "shell": [4, 14, 23, 26], "complet": [4, 14, 17, 18, 23, 24, 26], "prompt": [4, 14, 21, 23], "root": [4, 14, 23, 25, 26], "2efa5b50b909": [4, 14, 23], "jupyterlab": [4, 14, 23], "lab": [4, 14, 23], "allow": [4, 8, 11, 14, 15, 16, 19, 20, 21, 22, 23], "ip": [4, 12, 14, 23], "notebookapp": 4, "token": [4, 14, 23], "password": [4, 26], "Or": [4, 14, 23], "past": [4, 14, 19, 21, 23], "9b537d1fda9e4e9cadc673ba2a472e247deee69a6229ff8d": [4, 14, 23], "127": [4, 14, 23, 28, 34], "browser": [4, 14, 21, 23], "2022": [5, 6, 18, 26, 27, 28, 30, 31, 33, 34], "requir": [5, 6, 7, 8, 9, 10, 11, 12, 16, 17, 18, 19, 21, 22, 24, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 42, 43, 44], "anda": [5, 6], "book": 5, "166": 5, "835": 5, "anonym": [5, 38], "hotel": [5, 19], "378": [5, 6], "667": 5, "part": [5, 7, 8, 17, 18, 22, 24], "trip": 5, "utrip_id": 5, "consecut": 5, "reshap": [5, 6, 9], "full": [5, 6, 16, 21, 24, 26, 28, 33, 34], "itinerari": 5, "chronolog": [5, 6], "goal": [5, 6, 21, 24, 38], "city_id": 5, "import": [5, 6, 7, 8, 9, 10, 11, 12, 16, 18, 19, 21, 26, 27, 28, 29, 30, 31, 33, 34, 38], "statement": 5, "get_lib": [5, 7, 8, 12, 24, 26, 31], "get_book": 5, "reset": 5, "alloc": [5, 11, 22, 24, 30], "timeit": 5, "input_data_dir": [5, 7, 8, 9, 10, 11, 12, 27, 28, 30], "output_data_dir": [5, 6, 26, 27, 28, 29, 33, 34], "num_epoch": [5, 6], "25": [5, 6, 24, 26, 27, 34, 35, 36, 37, 42, 43, 44], "697025": 5, "26": [5, 11, 24, 26, 27, 29, 30, 35, 42], "988036": 5, "xla": [5, 6], "stream_executor": [5, 6, 11, 26, 33, 34], "cuda": [5, 6, 9, 10, 11, 24, 26, 28, 29, 30, 33, 34, 35, 36, 37, 42, 43, 44], "cuda_gpu_executor": [5, 6, 11, 26, 33, 34], "998": [5, 6], "success": [5, 6, 11, 26, 33, 34], "numa": [5, 6, 11, 26, 29, 33, 34], "sysf": [5, 6, 11, 26, 33, 34], "had": [5, 6, 11, 20, 26, 33, 34], "neg": [5, 6, 11, 16, 19, 26, 33, 34, 38], "zero": [5, 6, 11, 26, 33, 34], "988386": 5, "988518": 5, "28": [5, 24, 26, 34, 35, 36, 37, 42, 43, 44], "519868": 5, "520815": 5, "520999": 5, "521129": 5, "591345": 5, "591534": 5, "591665": 5, "591770": 5, "591778": 5, "gpu_process_st": [5, 11, 24, 30], "222": [5, 11, 30], "malloc": [5, 11, 24, 30], "async": [5, 11, 24, 30], "591860": 5, "08": [5, 6, 11, 21, 26, 30, 34, 35, 36, 37, 42, 43, 44], "149": 5, "compound": [5, 8, 26], "atom": [5, 8, 26], "session_id": [5, 19, 38], "0x7fe90a7fce80": 5, "0x7fe90a7f7820": 5, "vist": 5, "sever": [5, 6, 18, 33, 34, 40], "destin": 5, "displai": [5, 9, 24, 29], "otherwis": [5, 8, 24, 26, 30], "hidden": 5, "read_csv": [5, 6, 7], "train_set": 5, "parse_d": 5, "checkin": 5, "device_class": 5, "affiliate_id": 5, "1000027": 5, "2016": 5, "13": [5, 6, 24, 26, 27, 28, 29, 30, 34, 35, 36, 37, 42, 43, 44], "8183": 5, "desktop": 5, "7168": 5, "15626": 5, "60902": 5, "30628": 5, "253": 5, "1000033": 5, "04": [5, 7, 21, 24, 27, 30, 35, 36, 37, 42, 43, 44], "09": [5, 21, 24, 26, 27, 33, 35, 36, 37, 42, 43, 44], "38677": 5, "mobil": 5, "359": 5, "booker_countri": 5, "hotel_countri": 5, "elbonia": 5, "gondal": 5, "1000027_1": 5, "cobra": 5, "island": 5, "1000033_1": 5, "sequenc": [5, 20, 38], "attempt": [5, 26], "hop": 5, "journei": [5, 6], "length": [5, 12, 18, 22, 24, 26], "batch": [5, 6, 9, 10, 12, 16, 18, 19, 20, 22, 24, 26, 31], "box": [5, 8, 21], "doesn": [5, 6, 19, 26, 30], "t": [5, 6, 8, 9, 12, 16, 19, 20, 26, 27, 28, 30], "thank": 5, "pad": 5, "trim": 5, "particular": [5, 6, 8, 16, 20, 22], "effect": 5, "mask": [5, 6, 38], "discard": 5, "shorter": 5, "sens": [5, 20, 22], "noth": 5, "could": [5, 19, 20, 22, 28, 38], "present": [5, 20], "begin": [5, 6, 24, 33, 34, 41], "split": [5, 16, 18, 20, 22, 28], "mani": [5, 7, 10, 11, 14, 16, 19, 20, 22, 38], "shuffl": [5, 6, 7, 8, 22, 28, 38], "along": [5, 6, 11, 33, 34], "frac": 5, "number": [5, 7, 8, 10, 16, 19, 20, 21, 22, 24, 27], "len": [5, 6, 7, 26, 28], "217686": 5, "assign": [5, 8, 19, 22], "furthermor": 5, "proper": 5, "citi": [5, 19], "thei": [5, 7, 8, 15, 17, 19, 20, 21, 22, 27, 38], "cannot": [5, 24, 29], "from_panda": 5, "to_panda": 5, "groupbi": [5, 6], "num_exampl": [5, 6], "train_set_utrip_id": 5, "shape": [5, 10, 11, 12, 26, 30, 31], "validation_set_utrip_id": 5, "isin": 5, "sort_valu": [5, 33], "validation_set": 5, "too": [5, 8], "short": 5, "calcul": [5, 8, 10, 28, 38], "total": [5, 8, 10, 12, 24, 27, 28, 30], "schema_select_by_tag": 5, "addtag": [5, 8, 34], "train_set_dataset": 5, "validation_set_dataset": 5, "weekday_checkin": 5, "to_datetim": 5, "dt": 5, "weekdai": 5, "weekday_checkout": 5, "categorical_featur": [5, 16, 19], "groupby_featur": [5, 6], "groupby_col": [5, 6], "agg": [5, 6], "count": [5, 6, 19, 21], "sort_col": [5, 6], "list_featur": 5, "city_id_list": 5, "booker_country_list": 5, "hotel_country_list": 5, "weekday_checkin_list": 5, "weekday_checkout_list": 5, "less": [5, 16, 17, 19, 22, 27, 28, 38], "minimum_session_length": [5, 6], "city_id_count": 5, "filtered_sess": [5, 6], "wf": [5, 6], "train_process": 5, "validation_process": 5, "integ": [5, 8, 24, 28], "8238": 5, "156": [5, 6], "2278": 5, "2097": 5, "63": 5, "1160": 5, "87": [5, 26], "618": 5, "1050": 5, "65": [5, 26], "1032": 5, "757": 5, "140": [5, 16], "3603": 5, "262": [5, 34], "662": 5, "250": 5, "seq_schema": 5, "is_list": [5, 8, 18, 24], "is_rag": [5, 8, 18, 24], "properti": [5, 8, 10, 19, 24], "num_bucket": [5, 8, 24, 28], "freq_threshold": [5, 8, 24, 34], "max_siz": [5, 8, 24, 28], "start_index": [5, 8], "cat_path": [5, 8, 24], "domain": [5, 8, 24, 38, 39], "min": [5, 8, 16, 18, 19, 21, 24], "max": [5, 8, 16, 19, 21, 24], "embedding_s": [5, 8, 10, 24, 30], "cardin": [5, 8, 10, 24], "dimens": [5, 8, 10, 16, 22, 24], "value_count": 5, "int64": [5, 6, 8, 9, 12, 24, 28], "element_typ": [5, 8, 18, 24], "lt": [5, 8, 18, 24, 35, 36, 37, 42, 43, 44], "elementtyp": [5, 8, 18, 24], "none": [5, 6, 11, 12, 18, 19, 24, 26, 27, 28, 30, 31], "37202": 5, "37203": 5, "512": [5, 8, 10, 24, 26, 29], "195": 5, "variou": 5, "hyperparamet": [5, 15, 22, 26, 30], "portion": 5, "xlnet": [5, 38], "masking_prob": [5, 6], "paramet": [5, 6, 9, 10, 19, 21, 22, 26, 28, 29, 33, 34], "timestep": 5, "probabl": [5, 12, 18, 23, 26, 38], "summar": [5, 20, 38], "sequencemaskrandom": [5, 6], "pre": [5, 6, 22, 38], "randomli": [5, 10, 19, 22], "necessari": [5, 16, 17, 19], "configur": [5, 6, 11, 16, 17, 20, 21, 22, 24, 25, 26, 28, 29], "huggingfac": [5, 38], "its": [5, 6, 8, 9, 10, 11, 16, 19, 21, 22, 26, 28, 29], "replac": [5, 6, 7, 12, 19, 22, 27], "dummi": 5, "trainabl": [5, 22], "leakag": 5, "inputblockv2": [5, 6], "xlnetblock": [5, 6], "categoricaloutput": [5, 6], "select_by_nam": [5, 6, 10], "sequencemasklast": 5, "dmodel": [5, 6], "48": [5, 9, 27], "mlp_block": [5, 6], "activ": [5, 6, 10, 19, 21, 22, 33, 34], "relu": [5, 6, 9, 10, 21, 22, 29], "transformer_block": [5, 6], "d_model": [5, 6], "n_head": [5, 6], "n_layer": [5, 6], "sequence_combin": [5, 6], "default_loss": [5, 6], "44": [5, 18, 37, 44], "034041": 5, "cuda_dnn": [5, 6], "428": [5, 6], "cudnn": [5, 6, 29, 35, 36, 37, 42, 43, 44], "8700": [5, 6], "gradient": [5, 6], "mask_emb": [5, 6], "layer_": [5, 6], "_0": [5, 6, 24], "rel_attn": [5, 6], "r_s_bia": [5, 6], "seg_emb": [5, 6], "_1": [5, 6], "minim": [5, 6], "forget": [5, 6, 26], "54": [5, 26, 30], "541024": 5, "grappler": [5, 6], "loop_optim": [5, 6], "907": [5, 6], "loop": [5, 6, 8, 28], "control": [5, 6, 12, 20, 22, 26, 27], "xl_net_block": [5, 6], "sequential_block_5": 5, "replace_masked_embed": [5, 6], "raggedwher": [5, 6], "assert": [5, 6, 34], "assertguard": [5, 6], "branch_execut": [5, 6], "_95": [5, 6], "2720": [5, 29], "81": 5, "25m": [5, 7, 10, 11], "3315": 5, "1973": 5, "mrr_at_10": [5, 6], "0863": 5, "1123": 5, "map_at_10": [5, 6], "precision_at_10": [5, 6], "0197": 5, "3306": 5, "70": [5, 26, 33], "0979": 5, "3633": 5, "1707": 5, "2161": 5, "0363": 5, "0950": 5, "71": [5, 10], "26m": 5, "5827": 5, "4306": 5, "2056": 5, "2588": 5, "0431": 5, "72": 5, "3211": 5, "4627": 5, "2213": 5, "2784": 5, "0463": 5, "3194": 5, "1920": 5, "4787": 5, "2306": 5, "2892": 5, "0479": 5, "1903": 5, "0x7fe67105a7f0": 5, "But": [5, 6, 20, 22], "come": 5, "realiti": 5, "care": [5, 11, 17], "mimic": 5, "scenario": [5, 22], "measur": [5, 8], "return_dict": [5, 30], "968982": 5, "_74": 5, "340": 5, "20m": [5, 26], "7151": 5, "5533": 5, "3083": 5, "3665": 5, "0553": 5, "7149": 5, "715089797973633": 5, "5533444881439209": 5, "30831339955329895": 5, "36654922366142273": 5, "055334459990262985": 5, "635858535766602": 5, "deliv": [5, 17, 18], "fast": [5, 8, 17, 18], "ai": [5, 17, 18, 21, 22, 24], "inf_op": [5, 33, 34], "serial": [5, 6, 11, 27, 30], "tfsharedembed": [5, 6], "_feature_shap": [5, 6], "tensorshap": [5, 6], "_feature_dtyp": [5, 6], "dropout": [5, 6, 10, 16, 20, 21, 22], "model_context_layer_call_fn": [5, 6, 11, 24, 26, 30], "model_context_layer_call_and_return_conditional_loss": [5, 6, 11, 24, 26, 30], "sequence_mask_random_layer_call_fn": [5, 6], "sequence_mask_random_layer_call_and_return_conditional_loss": [5, 6], "sequence_mask_last_layer_call_fn": 5, "108": [5, 28], "tmp1sakw940": 5, "101": [5, 6], "custommaskwarn": [5, 6], "get_config": [5, 6], "pass": [5, 6, 11, 26], "custom_object": [5, 6], "serialize_keras_object": [5, 6], "maybe_valu": [5, 6], "288": [5, 6], "legaci": [5, 6, 11], "layer_seri": [5, 6], "134": [5, 6], "obj": [5, 6], "manual": [5, 6, 11, 24, 26], "produc": [5, 16, 22, 26], "downstream": 5, "wait": [5, 6, 7, 17, 18, 24, 33, 34], "until": [5, 6, 17, 18, 21, 24, 33, 34], "i0414": [5, 6, 17, 18, 24, 33, 34], "741833": [5, 6, 17, 18, 24, 33, 34], "4067": [5, 6, 17, 18, 24, 33, 34], "grpc_server": [5, 6, 17, 18, 24, 33, 34], "4421": [5, 6, 17, 18, 24, 33, 34], "grpcinferenceservic": [5, 6, 17, 18, 24, 33, 34], "742197": [5, 6, 17, 18, 24, 33, 34], "http_server": [5, 6, 17, 18, 24, 33, 34], "3113": [5, 6, 17, 18, 24, 33, 34], "httpservic": [5, 6, 17, 18, 24, 33, 34], "783470": [5, 6, 17, 18, 24, 33, 34], "178": [5, 6, 17, 18, 24, 33, 34], "servic": [5, 6, 17, 18, 19, 22, 24, 33, 34], "correspond": [5, 6, 8, 19, 24, 33, 34], "subsequent": [5, 6], "stop": [5, 10], "convert_df_to_triton_input": [5, 6, 12, 24, 26, 31, 33, 34], "iloc": [5, 6, 18], "tritoncli": [5, 6, 12, 18, 24, 26, 31, 33, 34], "grpc": [5, 6, 12, 18, 24, 31, 33, 34], "grpcclient": [5, 6, 12, 18, 24, 31, 33, 34], "inferenceservercli": [5, 6, 12, 18, 24, 26, 31, 33, 34], "logit": [5, 6], "as_numpi": [5, 6, 12, 24, 26, 31, 33, 34], "categorical_output": [5, 6], "8206294": 5, "3849059": 5, "9042726": 5, "851537": 5, "4237087": 5, "73849726": 5, "anoth": [5, 6, 22, 24, 28, 31, 38], "goe": [5, 6], "deeper": [5, 6], "cover": [5, 6], "weight": [5, 6, 16, 20], "ty": [5, 6], "fashion": [5, 6, 38], "e": [5, 6, 12, 16, 17, 18, 19, 20, 21, 22, 23, 24, 26, 28, 38], "commerc": [5, 6, 19, 22, 38], "chose": 6, "sigir": 6, "ecom": 6, "2021": [6, 7, 8, 9, 10, 11, 24, 27, 28, 29, 30], "36m": 6, "event": [6, 19, 20, 22], "brows": 6, "action": 6, "taken": [6, 11, 17, 20, 33], "22": [6, 7, 8, 9, 10, 11, 26, 33, 34, 41], "58": [6, 26], "667322": 6, "026020": 6, "026445": 6, "026622": 6, "place": [6, 16, 28], "alongsid": 6, "sigir_dataset": 6, "extract": [6, 10, 19, 22, 27], "By": [6, 8, 19, 22, 27], "boolean": 6, "run_on_synthetic_data": 6, "deal": [6, 16, 20], "nan": [6, 8, 19], "pars": [6, 24, 26], "string": [6, 7, 8, 19, 27], "browsing_train": 6, "sku": 6, "sku_to_cont": 6, "description_vector": 6, "image_vector": 6, "x": [6, 11, 20, 26, 27, 28], "els": [6, 28], "eval": [6, 16, 18, 21, 22, 24, 26, 29], "counterpart": 6, "further": 6, "product_sku_hash": 6, "essenti": 6, "descript": 6, "piec": 6, "sure": [6, 8, 17, 26, 28], "addition": [6, 11, 25], "group": [6, 19, 20, 38], "session_id_hash": 6, "arrang": 6, "industri": 6, "consult": [6, 9, 10, 33, 34], "retain": 6, "cat_op": 6, "event_typ": 6, "product_act": 6, "hashed_url": 6, "server_timestamp_epoch_m": 6, "normalizeminmax": 6, "hashed_url_count": 6, "won": 6, "nor": 6, "longer": 6, "product_sku_hash_list": 6, "event_type_list": 6, "product_action_list": 6, "hashed_url_list": 6, "train_transform": [6, 33, 34], "coupl": 6, "578": 6, "972": 6, "420": 6, "328": 6, "126": 6, "233": 6, "925": 6, "410": 6, "766": 6, "955": 6, "745": 6, "210": 6, "940": 6, "688": 6, "986": 6, "524": 6, "425": 6, "298": 6, "304": [6, 35, 36, 37, 42, 43, 44], "393": 6, "697": 6, "706": 6, "313": 6, "834": 6, "83": [6, 26, 27, 35, 42], "502": 6, "221": 6, "915": 6, "658": 6, "802": 6, "180": [6, 20, 34], "580": 6, "702": 6, "339": 6, "645": 6, "436": 6, "358": 6, "84": [6, 34, 35, 36, 37, 42, 43, 44], "35": [6, 26, 27, 36, 43], "271": 6, "562": [6, 29], "498": [6, 26], "172": 6, "239": 6, "270": 6, "215": 6, "489": [6, 26], "278": 6, "153": 6, "717": 6, "540": [6, 29], "219": 6, "200": [6, 9, 12, 18, 20, 21, 24, 26], "169": [6, 26], "419": 6, "875": 6, "725": 6, "926": 6, "770": 6, "160": [6, 9], "554": 6, "763": 6, "922": 6, "914": 6, "592": [6, 26], "842": 6, "916": 6, "137": 6, "928": 6, "615": 6, "318": 6, "506": 6, "281": 6, "191": [6, 18, 24], "480": 6, "965": 6, "399": 6, "761": 6, "preserv": 6, "grab": 6, "just": [6, 16, 19, 22, 28], "category_hash": 6, "price_bucket": 6, "07939800762120258": 6, "3465797761609977": 6, "186690": 6, "4275482879608162": 6, "30569476366666": 6, "1440": 6, "951997": 6, "31035419787213536": 6, "18070481533058008": 6, "973384": 6, "31319783485940356": 6, "11623980504981396": 6, "138": 6, "146260": 6, "25091279302969943": 6, "33473442518442525": 6, "119": 6, "808252": 6, "wf_sku": 6, "skus_d": 6, "836": 6, "979": 6, "469": 6, "118": 6, "emebed": 6, "to_npi": 6, "npy": 6, "know": [6, 24, 28, 31], "associ": [6, 10], "07939801": 6, "34657978": 6, "38269496": 6, "56307004": 6, "10142923": 6, "03702352": 6, "11606304": 6, "10070879": 6, "21879928": 6, "06107687": 6, "20743195": 6, "01330719": 6, "60182867": 6, "0920322": 6, "2648726": 6, "56061561": 6, "48643498": 6, "39045152": 6, "40012162": 6, "09153962": 6, "38351605": 6, "57134731": 6, "59986226": 6, "40321368": 6, "32984972": 6, "37559494": 6, "1554353": 6, "0413067": 6, "33814398": 6, "30678041": 6, "24001132": 6, "42737922": 6, "41554601": 6, "40451691": 6, "50428902": 6, "2004803": 6, "38297056": 6, "06580838": 6, "48285745": 6, "51406472": 6, "02268894": 6, "36343324": 6, "32497967": 6, "29736346": 6, "00538915": 6, "12329302": 6, "04998194": 6, "27843002": 6, "20212714": 6, "39019503": 6, "embeddingoper": 6, "lookup": 6, "hardwar": [6, 11, 28], "eas": 6, "npartit": 6, "part_mem_fract": [6, 27, 30], "broad": [6, 22], "lot": [6, 27], "permit": 6, "significantli": [6, 38], "increas": [6, 22], "embedding_oper": 6, "id_lookup_t": 6, "lookup_kei": 6, "embedding_nam": 6, "product_embed": 6, "835162": 6, "836068": 6, "836268": 6, "836425": 6, "836673": 6, "836849": 6, "837009": 6, "837114": 6, "837130": 6, "insruct": 6, "sample_batch": 6, "include_target": 6, "prepare_featur": 6, "dict_kei": 6, "input_block": 6, "pretrained_embed": 6, "pretrainedembed": 6, "normal": [6, 8, 11, 19, 26, 28], "norm": 6, "output_dim": 6, "amen": 6, "subsequ": [6, 11, 18], "everyth": [6, 11], "input_batch": 6, "remain": [6, 25, 27, 28, 33, 34], "itself": 6, "And": [6, 10, 22], "950175": 6, "59": 6, "285571": 6, "sequential_block_7": 6, "9800": 6, "0106": 6, "0033": 6, "0050": 6, "0011": 6, "9689": 6, "34": [6, 26], "9591": 6, "0031": 6, "0048": 6, "9363": 6, "9471": 6, "0107": 6, "0028": 6, "0046": 6, "9206": 6, "9398": 6, "0103": 6, "0030": 6, "0047": 6, "0010": 6, "9015": 6, "9375": 6, "0104": 6, "9095": 6, "0x7f55081d17c0": 6, "inference_oper": 6, "prepare_list_features_1_layer_call_fn": 6, "110": [6, 26], "tmpi3g8g7q7": 6, "five": 6, "2332087": 6, "1218574": 6, "390479": 6, "7735352": 6, "1954267": 6, "34523243": 6, "arriv": 6, "hash": [6, 19, 22], "predicted_hashed_url_id": 6, "argmax": 6, "academ": [7, 10, 11, 13, 14], "public": [7, 10, 11, 13, 14, 15, 27, 30, 38], "movi": [7, 8, 10, 11, 12, 33, 34], "rate": [7, 8, 9, 10, 11, 12, 16, 19, 20, 22, 33, 34, 38], "62": [7, 8, 10, 11, 35, 36, 37, 42, 43, 44], "000": [7, 10, 11], "162": [7, 10, 11, 24], "project": [7, 10, 11, 21, 22], "genr": [7, 8, 9, 10, 11, 34], "improv": [7, 10, 11, 16, 21, 22, 38], "state": [7, 10, 18, 19, 24, 33, 34], "art": [7, 10], "hot": [7, 8, 10, 13, 14, 19, 20, 22], "download_fil": [7, 27], "df_lib": [7, 8, 12, 24, 26, 31], "expandus": [7, 8, 9, 10, 11, 12], "unzip": [7, 34], "grouplen": 7, "ml": [7, 17, 21, 23, 24, 26, 33, 34], "zip": [7, 16, 19, 26, 34], "262mb": 7, "18mb": 7, "00file": 7, "movieid": [7, 8, 9, 10, 11, 12, 33, 34], "titl": [7, 34], "toi": [7, 8, 34], "stori": [7, 8, 34], "adventur": [7, 8], "anim": [7, 8], "children": [7, 8], "comedi": [7, 8], "fantasi": [7, 8], "jumanji": [7, 8], "grumpier": 7, "men": 7, "romanc": [7, 8], "exhal": 7, "drama": [7, 8], "father": [7, 8], "bride": [7, 8], "ii": [7, 8], "str": [7, 12, 18, 24, 26, 27, 28], "axi": [7, 20], "movies_convert": [7, 8], "userid": [7, 8, 9, 10, 11, 12, 33, 34], "296": 7, "1147880044": 7, "306": 7, "1147868817": 7, "307": 7, "1147868828": 7, "665": 7, "1147878820": 7, "899": 7, "1147868510": 7, "train_df": 7, "num_valid": 7, "c": [8, 9, 21, 26, 27, 28, 29], "One": [8, 10, 18, 19, 22, 28, 38], "vari": [8, 20, 21, 24], "often": [8, 14, 22, 38], "frequent": [8, 19, 20], "ones": [8, 19, 20, 38], "howev": [8, 31, 33, 34, 35, 36, 37, 38, 42, 43, 44], "better": [8, 16, 19, 20, 21, 22], "datapoint": [8, 10], "ad": [8, 19, 22, 23, 25, 26, 39], "v0": [8, 41], "familiar": [8, 13, 14, 25, 26, 28, 29, 30], "joinextern": 8, "wise": 8, "shutil": [8, 28], "columngroup": 8, "overload": 8, "turn": [8, 26], "op1": 8, "op2": 8, "sound": 8, "complic": 8, "left": [8, 38], "road": 8, "categorical_column": [8, 28], "label_column": [8, 24, 26, 28, 31], "acycl": 8, "visual": [8, 21, 38], "graphviz": 8, "contigu": [8, 19, 22, 28], "increment": 8, "fulfil": 8, "cat_featur": [8, 28], "int8": [8, 9, 11, 18], "mean": [8, 16, 19, 22, 24, 27, 38], "deviat": 8, "account": [8, 20, 21, 22, 26], "manifest": 8, "matter": 8, "solv": [8, 38], "break": 8, "chunk": 8, "main": [8, 16, 29], "purpos": [8, 10, 17, 18, 33, 34], "reliabl": 8, "dask_cudf": [8, 16], "hood": 8, "decomposit": 8, "lazili": 8, "evalu": [8, 15, 16, 19, 21, 23], "handl": [8, 26], "global": [8, 19, 27, 29], "partit": [8, 16, 19], "instanti": [8, 28], "couldn": [8, 26, 28], "label": [8, 9, 24, 26, 27, 28, 29], "enforc": [8, 28], "dictionari": [8, 28], "give": [8, 10, 11, 16, 24, 26, 28, 38], "dict_dtyp": [8, 28], "train_dataset": [8, 10, 24, 26, 28], "valid_dataset": [8, 24, 26, 28], "ll": [8, 24, 28, 30], "scikit": 8, "148": [8, 26, 29], "692": 8, "m": [8, 10, 12, 16, 19, 21, 22, 26, 28], "sy": [8, 10, 12, 24, 26, 27, 28, 30], "354": 8, "wall": [8, 10, 12, 24, 27, 28, 30], "0x7fb2d7fc42b0": 8, "clear": [8, 9], "clean": [8, 28, 29], "rmtree": [8, 28], "_metadata": [8, 24], "per_partit": [8, 28], "cat": [8, 28], "write_hugectr_keyset": 8, "speed": [8, 9, 10, 28, 29], "1e": [8, 10, 16, 21, 22, 26], "03": [8, 11, 21, 24, 26, 27, 29, 34], "86": [8, 12, 33], "\u00b5": 8, "restor": [8, 28], "162541": 8, "162542": [8, 9, 10], "56658": 8, "56659": [8, 10], "glob": [8, 10, 11, 27, 30, 31], "train_path": [8, 10, 24, 26, 28], "valid_path": [8, 10, 24, 26, 28], "part_0": [8, 24, 30], "1691": [8, 10], "332": [8, 10], "1001": [8, 10], "154": [8, 10], "967": [8, 10], "245": [8, 10], "150851": 8, "622": 8, "39553": 8, "1146": 8, "benefit": [9, 16, 20], "basic": [9, 13, 14, 19, 20, 21, 25, 26], "object": [9, 19, 24], "adopt": 9, "estim": [9, 19, 26, 29], "ctr": [9, 19, 29], "experienc": [9, 29], "dgx": [9, 29, 35, 36, 37, 42, 43, 44], "a100": [9, 28, 29, 35, 36, 37, 42, 43, 44], "prove": [9, 29], "fastest": [9, 29], "commerci": [9, 29], "parallel": [9, 11, 20, 21, 29], "dcn": [9, 11, 16, 20, 29, 38], "deepfm": [9, 16, 20, 29], "profici": 9, "oversubscrib": 9, "don": [9, 16, 19, 20, 26], "prefetch": 9, "asynchron": 9, "multithread": 9, "via": [9, 10, 21, 29, 31], "output_dtyp": 9, "numer": [9, 19, 26, 27], "previou": [9, 10, 11, 20, 21, 22, 26, 28, 31], "model_dir": [9, 11, 26], "movielens_hugectr": 9, "mkdir": [9, 28, 29], "dedic": 9, "commandlin": [9, 29], "train_hugectr": 9, "afterward": [9, 28, 31], "writefil": [9, 26], "get_embedding_s": 9, "mpi4pi": [9, 29], "mpi": [9, 29], "solver": [9, 29], "createsolv": [9, 29], "vvgpu": [9, 29], "batchsiz": [9, 29], "2048": 9, "batchsize_ev": [9, 29], "max_eval_batch": [9, 29], "i64_input_kei": [9, 29], "use_mixed_precis": [9, 29], "repeat_dataset": [9, 29], "createoptim": [9, 29], "optimizer_typ": [9, 29], "optimizer_t": [9, 29], "reader": [9, 25, 26, 28, 29], "datareaderparam": [9, 29], "data_reader_typ": [9, 29], "datareadertype_t": [9, 29], "_file_list": [9, 24, 29], "txt": [9, 16, 21, 24, 29], "eval_sourc": [9, 29], "check_typ": [9, 29], "check_t": [9, 29], "non": [9, 12, 29], "slot_size_arrai": [9, 29], "56586": 9, "label_dim": [9, 29], "label_nam": [9, 29], "dense_dim": [9, 29], "dense_nam": [9, 29], "data_reader_sparse_param_arrai": [9, 29], "datareadersparseparam": [9, 29], "data1": [9, 29], "nnz_per_slot": 9, "is_fixed_length": 9, "slot_num": 9, "sparseembed": [9, 29], "embedding_typ": [9, 29], "embedding_t": [9, 29], "localizedslotsparseembeddinghash": [9, 29], "workspace_size_per_gpu_in_mb": [9, 29], "embedding_vec_s": [9, 29], "sum": [9, 10, 29], "sparse_embedding_nam": [9, 29], "sparse_embedding1": [9, 29], "bottom_nam": [9, 29], "denselay": [9, 29], "layer_typ": [9, 29], "layer_t": [9, 29], "top_nam": [9, 29], "reshape1": 9, "leading_dim": 9, "innerproduct": [9, 29], "fc1": [9, 29], "num_output": [9, 29], "relu1": [9, 29], "fc2": [9, 29], "relu2": [9, 29], "fc3": [9, 29], "binarycrossentropyloss": [9, 29], "summari": [9, 29, 39], "max_it": [9, 29], "2000": 9, "eval_interv": [9, 29], "snapshot": [9, 29], "1900": 9, "graph_to_json": [9, 29], "graph_config_fil": [9, 29], "overwrit": [9, 26], "l": [9, 28, 29, 30], "0_opt_sparse_1900": 9, "_dense_1900": 9, "_opt_dense_1900": 9, "0_sparse_1900": 9, "emb_vector": 9, "slot_id": 9, "ask": 9, "mv": 9, "observ": [10, 20, 22], "slow": 10, "bottleneck": 10, "nativ": 10, "veri": [10, 16, 19, 22, 24], "experi": [10, 20, 22, 24, 33, 34, 38], "tutori": [10, 16, 20, 21, 33, 34], "explain": [10, 11, 17, 19, 20, 21, 22, 30, 38], "gc": 10, "framework_util": 10, "process_epoch": 10, "tensor": [10, 24, 29], "recogn": 10, "accordingli": [10, 19, 22], "validation_dataset": 10, "train_load": 10, "valid_load": 10, "32768": [10, 21], "usual": [10, 38], "tupl": 10, "actual": [10, 17, 22], "reason": [10, 16], "second": [10, 26], "3rd": 10, "4th": 10, "5th": 10, "third": 10, "6th": 10, "88555": 10, "88556": 10, "88557": 10, "848": 10, "1847": 10, "5456": 10, "3095": 10, "1062": 10, "3705": 10, "flat": 10, "del": 10, "concaten": [10, 11, 16, 22, 38], "forward": [10, 26], "batchnorm": 10, "detail": [10, 14, 16, 20, 22, 25, 26, 30, 31, 38], "captur": [10, 22], "def": [10, 26], "extract_info": 10, "col_nam": 10, "single_hot_embedding_tables_shap": 10, "mutli_hot_embedding_tables_shap": 10, "embedding_table_shap": 10, "num_continu": 10, "emb_dropout": 10, "layer_hidden_dim": 10, "layer_dropout_r": 10, "initial_cat_lay": 10, "concatenatedembed": 10, "embedding_lay": 10, "modulelist": 10, "inplac": 10, "mh_cat_lay": 10, "multihotembed": 10, "embeddingbag": 10, "mode": [10, 12, 16, 26], "initial_cont_lay": 10, "batchnorm1d": 10, "ep": 10, "momentum": 10, "affin": 10, "track_running_stat": 10, "linear": [10, 19, 22], "in_featur": 10, "1040": 10, "out_featur": 10, "bia": 10, "output_lay": 10, "lr": [10, 11, 16, 20, 21, 22, 30], "train_loss": 10, "y_pred": 10, "y": [10, 20, 21, 26], "loss_func": 10, "nn": 10, "bcewithlogitsloss": 10, "valid_loss": 10, "02d": 10, "4f": 10, "610": 10, "152": 10, "5204": 10, "2798": 10, "17": [10, 26, 30, 34], "323": 10, "cross": [11, 22], "streamlin": 11, "thu": [11, 19], "despit": 11, "fairli": 11, "elabor": 11, "great": 11, "fullest": 11, "datalod": 11, "optmiz": 11, "maximum": [11, 22, 27], "example_util": [11, 24], "workflow_fit_transform": [11, 24], "save_result": 11, "373230": 11, "40": [11, 24, 26, 27, 34, 35, 36, 37, 42, 43, 44], "514976": 11, "996": 11, "515413": 11, "515595": 11, "727125": 11, "728141": 11, "728351": 11, "728508": 11, "41": [11, 24, 26, 34], "475009": 11, "475231": 11, "475394": 11, "475508": 11, "475570": 11, "1637": 11, "train_d": 11, "valid_d": 11, "am": 11, "modifi": [11, 33, 34], "remove_col": 11, "extrem": 11, "facilit": 11, "carri": 11, "dcnmodel": [11, 22], "depth": 11, "deep_block": 11, "binaryoutput": [11, 24], "intern": 11, "v2": [11, 12, 16, 18, 20, 24], "upon": 11, "explicit": [11, 12, 16, 19, 27], "complementari": 11, "implicit": [11, 19, 23, 38], "overal": 11, "depict": 11, "stack": [11, 22], "embbed": 11, "emb": 11, "\ud835\udc5b": 11, "\ud835\udc65": 11, "imag": [11, 16, 19, 21, 22, 41], "paper": [11, 22, 38], "opt": [11, 26], "adagrad": [11, 22], "learning_r": [11, 30], "1221": 11, "6m": 11, "6609": 11, "5281": 11, "6588": 11, "5626": 11, "6537": 11, "0x7f7b5959ed30": 11, "serving_oper": [11, 24, 26, 31], "remove_by_tag": 11, "ens_conf": [11, 24, 31], "node_conf": [11, 24, 31], "unsupport": [11, 30], "signatur": [11, 30], "binary_output": [11, 12, 18, 24], "tensorspec": [11, 30], "sample_weight": [11, 30], "negative_candidate_id": 11, "func_graph": [11, 30], "unknownargu": [11, 30], "0x7f7b58732c40": 11, "_wrapped_model": [11, 30], "charact": [11, 19, 30], "train_compute_metr": [11, 26, 30], "dense_6_layer_call_fn": 11, "dense_6_layer_call_and_return_conditional_loss": 11, "47": [11, 24, 26, 35, 42], "tmp5eyrbewk": 11, "0_predicttensorflowtriton": [11, 12], "2020": 12, "sinc": [12, 17, 18, 26], "yet": 12, "path_to_model": 12, "hasn": 12, "merlin_triton": 12, "movielen": [12, 33, 34, 38], "deactiv": 12, "filterwarn": 12, "ignor": [12, 19, 22, 24, 33, 34], "endpoint": [12, 31], "verifi": [12, 24, 33, 34], "tritonhttpcli": 12, "try": [12, 14, 18, 23, 24, 31, 33, 38], "triton_cli": [12, 18, 24], "verbos": [12, 18, 24, 26], "channel": [12, 18, 24], "fail": [12, 18, 24], "aliv": 12, "is_server_l": [12, 18, 24], "health": [12, 18, 24], "live": [12, 18, 24], "header": [12, 16, 18, 24, 26], "httpsocketpoolrespons": [12, 18, 24], "plain": [12, 18, 24], "get_model_repository_index": [12, 18, 24], "bytearrai": [12, 18, 24], "b": [12, 18, 20, 21, 24, 26], "ensemble_model": 12, "load_model": [12, 24, 31], "model_nam": [12, 26], "623": 12, "craft": 12, "9098613": 12, "59266": 12, "553": 12, "2556225": 12, "17037": 12, "608": 12, "20514728": 12, "133412": 12, "40583": 12, "inferinput": [12, 24, 26, 31, 33, 34], "inferrequestedoutput": [12, 24, 31, 33, 34], "request_id": [12, 24, 31], "decod": 12, "60931313": 12, "6242866": 12, "6068166": 12, "reflect": 12, "movielens25m": [13, 14], "teach": [13, 16], "becom": 13, "convert": [13, 19, 24, 26, 28, 29, 31, 32], "adapt": 14, "primarili": 14, "largest": [14, 26, 27, 30, 32], "publicli": [14, 32], "1tb": [14, 28, 30, 32], "uncompress": [14, 16, 27, 28, 30, 32], "billion": [14, 27, 28, 30, 32, 38], "amazon": [14, 26], "sdk": 14, "ecr": [14, 26], "boto3": [14, 26], "8797": [14, 23, 25], "8787": [14, 23, 25, 28, 34], "8796": [14, 23, 25], "8786": [14, 23, 25], "guid": [15, 16, 18, 29, 38], "power": [15, 16, 19, 22], "best": [15, 16], "practic": [15, 16], "typic": [15, 16, 19, 22, 38], "scienc": [15, 16], "involv": [15, 16, 17], "tune": 15, "meant": [15, 24], "effort": [15, 17], "possibl": [16, 19, 20, 22], "accuraci": [16, 21, 22], "problem": [16, 22, 38, 39], "hypertun": [16, 20], "feedback": [16, 19], "expos": [16, 22], "accord": [16, 19, 22], "doc": [16, 21, 26], "switch": 16, "english": 16, "prefer": [16, 28], "term": [16, 26], "mail": [16, 38], "gb": [16, 28, 35, 36, 37, 42, 43, 44], "qk": [16, 19], "video": [16, 19], "contextu": [16, 19], "g": [16, 19, 21, 22, 23, 26, 38], "watch": 16, "493": [16, 29], "feel": 16, "free": [16, 21, 22, 26, 28], "reduc": [16, 20, 22, 29, 38], "v100": [16, 28, 29, 30], "truncat": 16, "10000001": 16, "10m": 16, "input_data_path": [16, 18], "insid": [16, 21, 24, 26], "quick_start": [16, 17, 19, 21, 22], "r": [16, 18, 21, 26], "prototyp": 16, "dev": [16, 26, 28], "option": [16, 17, 22, 27, 28, 38, 39], "explan": 16, "discret": [16, 19, 28], "space": [16, 19, 21, 27, 28, 34], "binary_classif_target": [16, 19], "regression_target": [16, 19], "user_id_featur": [16, 19], "item_id_featur": [16, 19], "to_int32": [16, 19], "to_int16": [16, 19], "to_int8": [16, 19], "cast": 16, "lower": [16, 19, 20], "precis": [16, 30], "dataset_split_strategi": [16, 19], "random_by_us": [16, 19], "chosen": [16, 19], "random_split_eval_perc": [16, 19], "percentag": [16, 19], "filter_queri": [16, 19], "equal": [16, 20], "min_item_freq": [16, 19], "min_user_freq": [16, 19], "max_user_freq": [16, 19], "frequenc": [16, 18, 22], "threshold": [16, 19], "larger": [16, 19, 20, 22, 28, 29], "enable_dask_cuda_clust": [16, 19], "localcudaclust": [16, 19, 27, 28], "persist_intermediate_fil": [16, 19], "persist": [16, 19, 22], "cach": [16, 19, 26], "intermedi": [16, 19, 26], "paricular": 16, "data_path": [16, 19], "out_dataset_path": [16, 19, 22], "preproc": [16, 18, 19], "input_data_format": [16, 19], "csv_na_valu": [16, 19], "150": [16, 19], "num_max_rounds_filt": [16, 19], "video_categori": [16, 18, 19], "gender": [16, 18, 19, 34], "ag": [16, 18, 19, 34], "watching_tim": [16, 19], "2nd": [16, 22], "embeddings_dim": [16, 21, 22], "mlp_layer": [16, 21, 22], "decai": [16, 22], "lr_decay_r": [16, 21, 22], "lr_decay_step": [16, 21, 22], "regular": [16, 20], "l2_reg": [16, 21, 22], "embeddings_l2_reg": [16, 21, 22], "among": [16, 19, 21, 22], "train_batch_s": [16, 21, 22], "eval_batch_s": [16, 21, 22], "65536": [16, 21, 22], "faster": [16, 19, 21, 38], "train_steps_per_epoch": [16, 21, 22], "There": [16, 19, 22, 38], "lead": [16, 20, 22], "unbal": 16, "stl_positive_class_weight": [16, 21, 22], "rarer": [16, 22], "cuda_visible_devic": [16, 21, 22, 27, 28], "train_data_path": [16, 21, 22], "eval_data_path": [16, 19, 21, 22], "99": [16, 21, 22, 37, 44], "joint": 16, "expert": [16, 20], "independ": [16, 22], "sub": [16, 18, 22, 33, 34], "mmoe_num_mlp_expert": [16, 22], "expert_mlp_lay": [16, 22], "gate": [16, 20, 22], "gate_dim": [16, 22], "averag": [16, 19, 20, 22], "harvest": 16, "tower_lay": [16, 22], "balanc": [16, 20, 22], "mtl_loss_weight_": [16, 20, 22], "mtl_pos_class_weight_": [16, 20, 22], "use_task_tow": [16, 22], "embedding_sizes_multipli": [16, 21, 22], "mtl_pos_class_weight_click": 16, "mtl_pos_class_weight_lik": 16, "mtl_pos_class_weight_shar": 16, "mtl_pos_class_weight_follow": 16, "mtl_loss_weight_click": 16, "mtl_loss_weight_lik": 16, "mtl_loss_weight_follow": 16, "mtl_loss_weight_shar": 16, "bias": [16, 20, 22], "sweep": [16, 20], "benchmark": [16, 28], "compar": [16, 20, 22, 33, 34], "empir": [16, 22], "curat": [16, 24, 26], "moreov": 16, "quick": [17, 18, 20, 22], "individu": [17, 20], "phase": 17, "skew": 17, "context": [17, 22, 27, 38], "encapsul": 17, "procedur": 17, "postprocess": 17, "section": [17, 19, 20, 22, 26], "br": [17, 18], "nvt_workflow_path": 17, "tf_saved_model_path": 17, "output_ensemble_path": 17, "load_model_path": [17, 22], "exact": 17, "save_model_path": [17, 22], "broken": [18, 33, 34], "preprocesss": 18, "locat": [18, 19, 26], "23": [18, 23, 26, 27, 41], "57": [18, 26, 35, 42], "737454": 18, "input_fold": [18, 31], "workflow_stored_path": 18, "output_targets_path": 18, "outfil": 18, "whose": [18, 20, 22], "thing": [18, 28], "_cach": 18, "16794": 18, "221049": 18, "23542": 18, "61962": 18, "85886": 18, "281786": 18, "6016": 18, "26929": 18, "66043": 18, "30710": 18, "39752": 18, "222908": 18, "8365": 18, "273888": 18, "73739": 18, "280425": 18, "27552": 18, "28110": 18, "17866": 18, "69910": 18, "good": [18, 19, 20, 22, 24], "50231797": 18, "50405663": 18, "50262684": 18, "5003805": 18, "50613105": 18, "4995402": 18, "5027875": 18, "5036676": 18, "4998571": 18, "5052081": 18, "49693626": 18, "49303743": 18, "49347958": 18, "49609515": 18, "4981295": 18, "49890146": 18, "49202597": 18, "49149314": 18, "5004128": 18, "49684843": 18, "congratul": 18, "templat": [19, 22], "portal": 19, "social": [19, 22], "stream": [19, 22], "media": [19, 22], "dislik": 19, "cart": [19, 22], "purchas": [19, 22], "shown": 19, "desir": [19, 24, 27], "tenrec": [19, 20, 21, 22], "static": 19, "dynam": [19, 22, 28], "price": 19, "distinct": 19, "continuous_featur": 19, "natur": [19, 38], "approach": [19, 22, 38], "z": 19, "imput": 19, "constant": [19, 21], "float": [19, 22, 26], "median": [19, 20], "accept": [19, 22, 26], "nomin": 19, "meaning": 19, "represent": [19, 22, 24, 28, 38], "higher": [19, 20, 22], "infrequ": [19, 22, 28], "choic": [19, 20, 28, 38], "shard": [19, 22], "low": [19, 22, 26], "categ_min_freq_cap": 19, "cold": [19, 38], "fresh": 19, "major": 19, "long": [19, 28], "bucket": [19, 26], "collis": 19, "trade": 19, "off": [19, 26], "side": [19, 38], "respect": [19, 21], "udf": 19, "tempor": 19, "geograph": 19, "tip": 19, "inherit": 19, "preprocessingrunn": 19, "generate_nvt_featur": 19, "smooth": 19, "compress": [19, 26, 27, 35, 36, 37, 42, 43, 44], "small": [19, 20, 22, 38], "expand": [19, 20], "binar": 19, "ordin": 19, "period": 19, "dai": [19, 27, 28, 30], "besid": [19, 20], "whole": [19, 22], "segment": 19, "likelihood": [19, 22, 24, 38], "win": 19, "competit": [19, 27], "target_encoding_featur": 19, "target_encoding_target": 19, "week": 19, "month": 19, "year": 19, "quarter": 19, "hour": 19, "subtract": 19, "seen": [19, 20], "trend": 19, "accumul": 19, "engag": 19, "semest": 19, "treat": 19, "countri": 19, "latitud": 19, "longitud": 19, "airport": 19, "tourist": 19, "landmark": 19, "enrich": 19, "geoloc": 19, "censu": 19, "stratifi": 19, "dataset_split_temporal_timestamp": 19, "tsv": [19, 22, 28], "latter": 19, "columnar": 19, "setup": [19, 22, 26], "splitmust": 19, "predict_data_path": [19, 22], "csv_sep": 19, "tab": 19, "saveddefault": 19, "output_num_partit": 19, "filesdefault": 19, "whether": 19, "control_featur": 19, "comma": [19, 22], "kept": [19, 22], "sep": [19, 22, 27], "properli": [19, 21], "continuous_features_fillna": 19, "fill": [19, 28], "max_item_freq": 19, "timestamp_featur": 19, "session_id_featur": 19, "target_encoding_kfold": 19, "fold": 19, "consid": 19, "caus": 19, "overfit": [19, 22, 28], "target_encoding_smooth": 19, "factor": [19, 22, 38], "noisi": 19, "formula": 19, "sum_target_per_categ_valu": 19, "global_target_avg": 19, "categ_value_count": 19, "int16": 19, "to_float32": 19, "greater": [19, 22], "round": [19, 26], "interleav": 19, "proport": 19, "dask_cuda_visible_gpu_devic": 19, "visible_gpu_devic": 19, "dask_cuda_gpu_device_spill_frac": 19, "spill": [19, 28], "rais": [19, 27], "decid": [20, 22], "divid": 20, "mmoe": 20, "ple": 20, "bottom": [20, 22], "special": [20, 38], "fair": 20, "comparison": [20, 28], "bayesian": [20, 21], "bay": [20, 21], "reus": 20, "trial": [20, 21], "mayb": 20, "todo": 20, "maxim": [20, 21], "baselin": 20, "avg": 20, "highlight": 20, "bold": 20, "blue": [20, 38], "highest": 20, "sparser": [20, 22], "densiti": 20, "denser": 20, "necessarili": 20, "mantain": 20, "simpler": 20, "cost": [20, 26, 38], "evolv": [20, 21], "sensit": 20, "7023": 20, "7828": 20, "105": 20, "7826": 20, "chanc": 20, "accur": 20, "diminish": 20, "budget": 20, "bar": [20, 21], "chat": 20, "seach": 20, "spot": 20, "hotter": 20, "color": [20, 21], "smaller": [20, 27], "16k": 20, "dim": [20, 22], "reg": [20, 22], "analyz": [20, 24, 33, 34], "plot": [20, 21], "boxplot": 20, "hpo": 20, "curv": 20, "dot": [20, 38], "achiv": 20, "adjust": [21, 26], "interest": [21, 28, 30, 31], "cli": [21, 25, 26], "googl": [21, 26, 38], "vertex": 21, "azur": 21, "program": [21, 31], "toward": 21, "int_uniform": 21, "uniform": 21, "log_uniform": 21, "256": [21, 28, 29], "16384": 21, "preproc_data": 21, "easiest": 21, "preproc_input_data_path": 21, "reinstal": 21, "pypi": 21, "nee": 21, "author": [21, 22], "me": 21, "profil": 21, "hit": 21, "enter": 21, "press": [21, 28], "ctrl": 21, "quit": 21, "preproc_data_path": 21, "log_to_wandb": [21, 22], "wandb_ent": [21, 22], "wandb_project": [21, 22], "sucessfulli": 21, "websit": [21, 40], "rememb": 21, "sweep_nam": 21, "wandb_sweep": 21, "stl_mlp": 21, "stl_click_mlp": 21, "agent": 21, "sweep_id": 21, "team_nam": 21, "project_nam": 21, "identif": 21, "env": [21, 26], "background": 21, "nohup": 21, "nohup0": 21, "nice": [21, 22], "chart": 21, "hparam": 21, "inter": 22, "compani": 22, "heurist": 22, "matrix": [22, 38], "youtubednn": 22, "somewhat": 22, "correl": 22, "spend": 22, "resourc": [22, 39], "abil": 22, "capac": 22, "memor": 22, "wideanddeepmodel": 22, "deepfmmodel": 22, "drlm": 22, "bound": 22, "mixtur": 22, "mmoeblock": 22, "cgcblock": 22, "progress": [22, 26], "Their": 22, "pleblock": 22, "1d": 22, "cap": 22, "fraction": [22, 26], "unbalanc": 22, "alwai": 22, "in_batch_negatives_train": 22, "in_batch_negatives_ev": 22, "domin": 22, "tasks_sample_spac": 22, "match": [22, 26, 28, 30], "empti": 22, "On": 22, "hand": 22, "prior": 22, "predict_output_path": 22, "keep_column": 22, "ignore_column": 22, "wide_n_deep": 22, "three": [22, 26], "selu": 22, "elu": 22, "tanh": 22, "sigmoid": [22, 26], "mlp_init": 22, "glorot_uniform": 22, "decreas": 22, "dcn_interacted_layer_num": 22, "wnd_hashed_cross_num_bin": 22, "10000": [22, 29], "wnd_wide_l2_reg": 22, "wnd_ignore_combin": 22, "_categori": 22, "multihot_max_seq_length": 22, "cgc_num_task_expert": 22, "cgc_num_shared_expert": 22, "ple_num_lay": 22, "mtl_gates_softmax_temperatur": 22, "train_metrics_step": 22, "bit": 22, "throughput": [22, 38], "validation_step": 22, "random_se": 22, "reproduc": 22, "shuffled_train": 22, "metrics_log_frequ": 22, "tensorboard": 22, "log_to_tensorboard": 22, "sign": 22, "wandb": 22, "consol": 22, "wandb_exp_group": 22, "Not": [22, 35, 36, 37, 42, 43, 44], "home": [22, 25, 26], "predict_output_format": 22, "predict_output_keep_col": 22, "tradit": 23, "xgboost": [23, 38], "soon": [23, 39], "xx": [23, 41], "600621": 24, "183": [24, 33], "070258": 24, "070303": 24, "226": [24, 26], "070448": 24, "1638": 24, "16249": 24, "gv100": 24, "tianchi": [24, 26], "aliyun": [24, 26], "1000000": 24, "synthetic_data": [24, 26], "category_temp_directori": 24, "out_path": [24, 28], "61": [24, 27], "68": [24, 33, 36, 43], "meta": 24, "transit": 24, "828993": 24, "executor": 24, "1209": 24, "abort": 24, "indic": [24, 27, 28], "invalid_argu": 24, "placehold": 24, "1563": [24, 30], "eta": [24, 26], "4998": [24, 26], "190967": 24, "69": 24, "38m": 24, "5000": [24, 26], "1min": 24, "2min": 24, "0x7f74b2a4b1c0": 24, "prepare_list_features_layer_call_fn": 24, "prepare_list_features_layer_call_and_return_conditional_loss": 24, "dense_9_layer_call_fn": 24, "96": [24, 33, 34, 35, 36, 37, 42, 43, 44], "wokflow": 24, "dl": 24, "std": 24, "guarante": 24, "edg": 24, "inferenc": 24, "remove_input": [24, 26, 31, 33, 34], "0x7f74b290f550": 24, "subclass": 24, "tf_model_path": 24, "illustr": 24, "meet": 24, "hard": 24, "someth": 24, "model_context_2_layer_call_fn": 24, "model_context_2_layer_call_and_return_conditional_loss": 24, "prepare_list_features_2_layer_call_fn": 24, "prepare_list_features_2_layer_call_and_return_conditional_loss": 24, "tmpomjyo5xq": 24, "772": 24, "773": 24, "789": 24, "790": 24, "67": [24, 26, 36, 43], "73": 24, "74": 24, "0_transformworkflow": [24, 26], "1_predicttensorflow": [24, 26], "compos": [24, 33, 34], "__null_dask_index__": [24, 26], "800000": 24, "85": [24, 30, 33, 34], "5936": 24, "2045": 24, "1670": 24, "484": 24, "830": 24, "88": 24, "800001": 24, "2850": 24, "982": 24, "1879": 24, "544": 24, "934": 24, "800002": 24, "238": 24, "82": 24, "557": 24, "277": 24, "isol": 24, "declar": 24, "output_col": [24, 31], "earlier": 24, "5002032": 24, "5001995": 24, "put": 24, "credenti": [25, 26], "mount": [25, 26], "restart": 26, "wtth": 26, "116": 26, "tar": 26, "gz": [26, 27], "kb": 26, "25hrequir": 26, "attr": 26, "pasta": 26, "importlib": 26, "importlib_metadata": 26, "whl": 26, "patho": 26, "25hcollect": 26, "protobuf3": 26, "py2": 26, "smdebug_rulesconfig": 26, "s3transfer": 26, "botocor": 26, "jmespath": 26, "six": 26, "zipp": 26, "pypars": 26, "pytz": 26, "2017": 26, "dateutil": 26, "pox": 26, "dill": 26, "multiprocess": 26, "py38": 26, "ppft": 26, "contextlib2": 26, "urllib3": [26, 30], "wheel": 26, "25ldone": 26, "25h": 26, "filenam": 26, "809052": 26, "sha256": 26, "f446dd6eed6d268b7f3f2709f8f11c1ba153e382fbea9b2caedd517c1fb71215": 26, "3e": 26, "cb": 26, "b1": 26, "5b13ff7b150aa151e4a11030a6c41b1e457c31a52ea1ef11b0": 26, "protobuf3_to_dict": 26, "4029": 26, "8f99baaa875ba544d54f624f95dfbf4fd52ca96d52ce8af6d05c1ff2bb8435b2": 26, "fc": 26, "2d1e23d8b9a9013a83fbb418a0b17b1e6f81c8db8f53b53934": 26, "smdebug": 26, "rulesconfig": 26, "gather": 26, "traffic": 26, "taobao": 26, "retail": 26, "1_000_000": 26, "archiv": 26, "made": 26, "s3": 26, "upload": 26, "argpars": 26, "tempfil": 26, "runtim": 26, "tf_memory_alloc": 26, "logger": 26, "getlogg": 26, "__name__": 26, "setlevel": 26, "addhandl": 26, "streamhandl": 26, "stdout": 26, "parse_arg": 26, "parser": 26, "argumentpars": 26, "add_argu": 26, "train_dir": 26, "sm_channel_train": 26, "valid_dir": 26, "sm_channel_valid": 26, "sm_model_dir": 26, "parse_known_arg": 26, "create_nvtabular_workflow": 26, "create_ensembl": 26, "arg": [26, 28], "mkdtemp": 26, "workflow_path": 26, "train_data": 26, "valid_data": 26, "model_path": 26, "ensemble_path": 26, "__main__": 26, "_": [26, 33, 34], "pip3": 26, "push": 26, "build_and_push_imag": 26, "sh": 26, "outsid": 26, "region": 26, "euo": 26, "pipefail": 26, "algorithm": [26, 29], "algorithm_nam": 26, "east": 26, "st": 26, "caller": 26, "west": 26, "dkr": 26, "amazonaw": 26, "image_uri": 26, "login": 26, "usernam": 26, "stdin": 26, "ne": 26, "fi": [26, 29], "role": 26, "sess": 26, "prefix": 26, "get_execution_rol": 26, "get_rol": 26, "arn": 26, "awso": 26, "iam": 26, "843263297212": 26, "data_loc": 26, "upload_data": 26, "key_prefix": 26, "sts_client": 26, "get_caller_ident": 26, "my_sess": 26, "region_nam": 26, "ecr_imag": 26, "g4dn": 26, "equip": [26, 33, 34], "t4": [26, 35, 36, 37, 42, 43, 44], "entry_point": 26, "behind": 26, "scene": 26, "sagemaker_program": 26, "training_instance_typ": 26, "xlarg": 26, "instance_count": 26, "instance_typ": 26, "1_024": 26, "trainingprofilerreport": 26, "1667989110": 26, "inprogress": 26, "34m": 26, "34mnvidia": 26, "42766143": 26, "34mcopyright": 26, "2018": 26, "affili": 26, "34mvariou": 26, "modif": 26, "34mthi": 26, "34mby": 26, "34mhttp": 26, "34mnote": 26, "driver": [26, 35, 36, 37, 42, 43, 44], "515": 26, "kernel": 26, "510": 26, "34m2022": 26, "405": 26, "neuron": 26, "detect": [26, 27], "438": 26, "473": 26, "485": 26, "invok": 26, "34mtrain": 26, "additional_framework_paramet": 26, "channel_input_dir": 26, "current_host": 26, "algo": 26, "current_instance_group": 26, "homogeneousclust": 26, "current_instance_group_host": 26, "current_instance_typ": 26, "distribution_host": 26, "distribution_instance_group": 26, "framework_modul": 26, "input_config_dir": 26, "input_data_config": 26, "traininginputmod": 26, "s3distributiontyp": 26, "fullyrepl": 26, "recordwrappertyp": 26, "input_dir": 26, "instance_group": 26, "instance_groups_dict": 26, "instance_group_nam": 26, "is_hetero": 26, "is_mast": 26, "is_modelparallel_en": 26, "is_smddpmprun_instal": 26, "job_nam": 26, "376": [26, 29], "log_level": 26, "master_hostnam": 26, "module_dir": 26, "sourcedir": 26, "module_nam": 26, "network_interface_nam": 26, "eth0": 26, "num_cpu": 26, "num_gpu": [26, 28], "num_neuron": 26, "output_dir": 26, "output_intermediate_dir": 26, "resource_config": 26, "current_group_nam": 26, "user_entry_point": 26, "34menviron": 26, "34msm_host": 26, "34msm_network_interface_nam": 26, "34msm_hp": 26, "34msm_user_entry_point": 26, "34msm_framework_param": 26, "34msm_resource_config": 26, "34msm_input_data_config": 26, "34msm_output_data_dir": 26, "34msm_channel": 26, "34msm_current_host": 26, "34msm_current_instance_typ": 26, "34msm_current_instance_group": 26, "34msm_current_instance_group_host": 26, "34msm_instance_group": 26, "34msm_instance_groups_dict": 26, "34msm_distribution_instance_group": 26, "34msm_is_hetero": 26, "34msm_module_nam": 26, "34msm_log_level": 26, "34msm_framework_modul": 26, "34msm_input_dir": 26, "34msm_input_config_dir": 26, "34msm_output_dir": 26, "34msm_num_cpu": 26, "34msm_num_gpu": 26, "34msm_num_neuron": 26, "34msm_model_dir": 26, "34msm_module_dir": 26, "34msm_training_env": 26, "34msm_user_arg": 26, "34msm_output_intermediate_dir": 26, "34msm_channel_train": 26, "34msm_channel_valid": 26, "34msm_hp_batch_siz": 26, "34msm_hp_epoch": 26, "34mpythonpath": 26, "python38": 26, "dynload": 26, "merlin_hp": 26, "34minvok": 26, "486": 26, "debugg": 26, "761711": 26, "slightli": 26, "tf_enable_onednn_opt": 26, "818302": 26, "991": [26, 33, 34], "819693": 26, "819906": 26, "894084": 26, "34mto": 26, "895367": 26, "895631": 26, "895807": 26, "651703": 26, "651981": 26, "652183": 26, "653025": 26, "1532": [26, 30], "10752": 26, "tesla": [26, 28, 29, 30, 35, 36, 37, 42, 43, 44], "34mworkflow": 26, "tmp5fpdavsc": 26, "34mbatch_siz": 26, "34mepoch": 26, "34m684": 26, "684": 26, "5026": 26, "4990": 26, "11m": 26, "6922": 26, "5222": 26, "6941": 26, "4989": 26, "6858": 26, "5509": 26, "6991": 26, "4994": 26, "6790": 26, "5660": 26, "7052": 26, "4993": 26, "6751": 26, "5722": 26, "7096": 26, "6722": 26, "5755": 26, "7184": 26, "4991": 26, "6700": 26, "5777": 26, "7289": 26, "6687": 26, "5792": 26, "7404": 26, "6678": 26, "5801": 26, "7393": 26, "4988": 26, "89": [26, 30, 34], "requestsdependencywarn": [26, 30], "chardet": [26, 30], "34mwarn": 26, "output_layer_layer_call_fn": [26, 30], "output_layer_layer_call_and_return_conditional_loss": [26, 30], "97": [26, 37, 44], "34minfo": 26, "34mmodel": 26, "34mensembl": 26, "589": 26, "billabl": 26, "model_data": 26, "s3download": 26, "s3down": 26, "xvzf": 26, "artificat": 26, "sagemaker_triton_default_model_nam": 26, "sm_client": 26, "service_nam": 26, "modeldataurl": 26, "sagemaker_triton_tensorflow_vers": 26, "strftime": 26, "d": 26, "h": 26, "gmtime": 26, "create_model_respons": 26, "create_model": 26, "modelnam": 26, "executionrolearn": 26, "primarycontain": 26, "model_arn": 26, "modelarn": 26, "endpoint_instance_typ": 26, "endpoint_config_nam": 26, "create_endpoint_config_respons": 26, "create_endpoint_config": 26, "endpointconfignam": 26, "productionvari": 26, "instancetyp": 26, "initialvariantweight": 26, "initialinstancecount": 26, "variantnam": 26, "alltraff": 26, "endpoint_config_arn": 26, "endpointconfigarn": 26, "endpoint_nam": 26, "create_endpoint_respons": 26, "create_endpoint": 26, "endpointnam": 26, "endpoint_arn": 26, "endpointarn": 26, "describe_endpoint": 26, "endpointstatu": 26, "sleep": 26, "60": [26, 34], "rv": 26, "inservic": 26, "700000": 26, "700001": 26, "80": [26, 34, 35, 36, 37, 42, 43, 44], "5621": 26, "1936": 26, "700002": 26, "776": 26, "700003": 26, "970": 26, "334": 26, "700004": 26, "1939": 26, "668": 26, "700005": 26, "15893": 26, "5474": 26, "700006": 26, "102": 26, "7172": 26, "2470": 26, "700007": 26, "1163": 26, "401": 26, "700008": 26, "582": 26, "201": 26, "700009": 26, "4458": 26, "1536": 26, "636": 26, "983": 26, "1965": 26, "2890": 26, "1214": 26, "694": 26, "521": 26, "174": 26, "184": 26, "316": 26, "285": 26, "569": 26, "977": [26, 29], "837": 26, "1436": 26, "352": 26, "604": 26, "345": 26, "151": 26, "259": 26, "103": 26, "payload": 26, "httpclient": 26, "request_bodi": 26, "header_length": 26, "generate_request_bodi": 26, "datatyp": [26, 28], "binary_data_s": 26, "binary_data_output": 26, "x0c": 26, "x00": 26, "x12": 26, "x003": 26, "x16": 26, "x04": 26, "x02": 26, "x1e": 26, "x05": 26, "x06": 26, "x0b": 26, "x07": 26, "x18": 26, "x03": 26, "x00p": 26, "x0e": 26, "x1c": 26, "xe2": 26, "x00f": 26, "x11": 26, "xc2": 26, "xf5": 26, "x15": 26, "x08": 26, "xca": 26, "x93": 26, "x8b": 26, "x00j": 26, "x00c": 26, "x90": 26, "x01": 26, "x00n": 26, "x9c": 26, "x00b": 26, "xa6": 26, "x91": 26, "xc9": 26, "xd7": 26, "xad": 26, "xbe": 26, "xb6": 26, "xae": 26, "xb8": 26, "x1d": 26, "x009": 26, "x00e": 26, "x97": 26, "xe9": 26, "xd1": 26, "x00y": 26, "x00w": 26, "x004": 26, "x00g": 26, "kserv": 26, "protocol": [26, 28], "runtime_sm_cli": 26, "invoke_endpoint": 26, "contenttyp": 26, "vnd": 26, "bodi": 26, "header_length_prefix": 26, "header_length_str": 26, "parse_response_bodi": 26, "output_data": 26, "48595208": 26, "4647554": 26, "50048226": 26, "53553176": 26, "5209902": 26, "54944164": 26, "5032344": 26, "475241": 26, "5077254": 26, "5009623": 26, "incur": 26, "delete_model": 26, "delete_endpoint_config": 26, "delete_endpoint": 26, "responsemetadata": 26, "requestid": 26, "6ad24616": 26, "5c7c": 26, "4525": 26, "a63c": 26, "62d1b06ee8ad": 26, "httpstatuscod": 26, "httpheader": 26, "amzn": 26, "amz": 26, "wed": 26, "nov": 26, "gmt": 26, "retryattempt": 26, "tb": [27, 28, 30], "span": [27, 28, 30], "record": 27, "criteolab": 27, "kaggl": 27, "15gb": 27, "47gb": 27, "30min": 27, "download_criteo": 27, "input_path": [27, 30, 31], "crit_orig": 27, "number_dai": 27, "valueerror": 27, "minimum": 27, "day_": [27, 28], "storag": 27, "googleapi": 27, "cail": 27, "day_0": [27, 28], "3gb": 27, "0mb": 27, "day_1": [27, 28], "7gb": 27, "3mb": 27, "15min": 27, "3min": [27, 30], "19min": 27, "23min": 27, "orient": 27, "dask_cuda": [27, 28, 34], "device_mem_s": [27, 28], "get_rmm_siz": 27, "frac_siz": 27, "rmm_pool_siz": [27, 28], "local_directori": [27, 28], "564": 27, "preload": [27, 28, 34], "cont_nam": 27, "cat_nam": 27, "hex": 27, "hexadecim": 27, "file_list": 27, "384": [27, 33, 34], "futurewarn": 27, "set_dask_cli": 27, "preserve_fil": 27, "prerequisit": 28, "8min": [28, 30], "1x": 28, "9min": 28, "8x": 28, "100x": 28, "10000x": 28, "clip": 28, "publish": 28, "strong": 28, "altern": 28, "numba": 28, "fillmiss": 28, "pynvml_mem_s": 28, "test_dask": [28, 29, 30, 31], "use_hugectr": 28, "bool": 28, "stats_path": 28, "stat": 28, "dask_workdir": 28, "workdir": 28, "worker": [28, 29, 34], "fname": 28, "num_dai": 28, "listdir": 28, "dashboard": 28, "dashboard_port": 28, "tcp": [28, 34], "ucx": 28, "is_avail": 28, "visible_devic": 28, "device_limit_frac": 28, "device_pool_frac": 28, "part_mem_frac": 28, "device_s": 28, "device_limit": 28, "device_pool_s": 28, "occupi": 28, "fmem": 28, "1e9": 28, "bewar": 28, "schedul": 28, "port": [28, 34], "n_worker": 28, "device_memory_limit": 28, "dashboard_address": 28, "242": 28, "16739757": 28, "7554": 28, "11ed": 28, "88cc": 28, "2a33bb9638f6": 28, "c4f394c8": 28, "thread": 28, "gib": 28, "a605479f": 28, "02de": 28, "4ca6": 28, "8661": 28, "fd02b84655ac": 28, "comm": 28, "44059": 28, "45833": 28, "38541": 28, "nanni": 28, "43849": 28, "9hxldmq1": 28, "sxm2": [28, 29, 30], "32gb": [28, 29, 30], "75": [28, 36, 43], "35261": 28, "46685": 28, "37015": 28, "nsltf61z": 28, "still": 28, "isn": 28, "ideal": 28, "consumpt": 28, "fact": 28, "magnitud": 28, "uncent": 28, "concis": 28, "worth": 28, "handi": 28, "ve": 28, "occurr": 28, "signal": 28, "continuous_column": 28, "10000000": [28, 29], "categorify_op": 28, "cont_featur": 28, "min_valu": 28, "out_dtyp": 28, "label_featur": 28, "output_train_dir": 28, "output_valid_dir": 28, "260": 28, "0x7fdacec4fdc0": 28, "output_fil": 28, "cont": 28, "980": 28, "202": 28, "801": 28, "203": 28, "exploit": 29, "cubla": 29, "nccl": 29, "criteo_hugectr": 29, "file_to_writ": 29, "noqa": [29, 31], "sgd": [29, 30], "3014529": 29, "400781": 29, "2209": 29, "11869": 29, "38713": 29, "584616": 29, "12883": 29, "109": [29, 33], "17177": 29, "7425": 29, "20266": 29, "7085": 29, "1535": 29, "6000": 29, "relu3": 29, "interaction1": 29, "fc4": 29, "relu4": 29, "fc5": 29, "relu5": 29, "fc6": 29, "relu6": 29, "fc7": 29, "relu7": 29, "fc8": 29, "3200": 29, "1000": 29, "utf": 29, "run_tim": 29, "hctr": 29, "374": 29, "rk0": 29, "2831956451": 29, "490": 29, "peer": 29, "fulli": [29, 38], "all2al": 29, "warmup": 29, "491": 29, "num": 29, "dataread": 29, "vocabulari": 29, "54120457": 29, "541": 29, "max_vocabulary_size_per_gpu_": 29, "12288000": 29, "analysi": 29, "resolv": 29, "055": 29, "gpu0": 29, "107": 29, "111": 29, "warm": 29, "113": 29, "125": 29, "9122965335846": 29, "942873": 30, "579306": 30, "579506": 30, "16255": 30, "582211": 30, "582328": 30, "30655": 30, "8a": 30, "path_to_train_data": 30, "path_to_valid_data": 30, "part_1": 30, "2867": 30, "224": 30, "72m": 30, "0378": 30, "recal": 30, "0745e": 30, "binary_accuraci": 30, "9677": 30, "5673": 30, "1385": 30, "val_precis": 30, "val_recal": 30, "val_binary_accuraci": 30, "9680": 30, "6279": 30, "1405": 30, "13min": 30, "21min": 30, "45": 30, "0x7f58457b1c40": 30, "eval_metr": 30, "2787": 30, "31m": 30, "13853003084659576": 30, "9679937958717346": 30, "6279259324073792": 30, "14048266410827637": 30, "predictionoutput": 30, "positive_item_id": 30, "label_relevant_count": 30, "valid_negatives_mask": 30, "negative_item_id": 30, "0x7f57c5f8ecd0": 30, "c1": 30, "c10": 30, "c11": 30, "c12": 30, "c13": 30, "c14": 30, "c15": 30, "c16": 30, "c17": 30, "c18": 30, "c19": 30, "c2": 30, "c20": 30, "c21": 30, "c22": 30, "c23": 30, "c24": 30, "c25": 30, "c26": 30, "c3": 30, "c4": 30, "c5": 30, "c6": 30, "c7": 30, "c8": 30, "c9": 30, "i1": 30, "i10": 30, "i11": 30, "i12": 30, "i13": 30, "i2": 30, "i3": 30, "i4": 30, "i5": 30, "i6": 30, "i7": 30, "i8": 30, "i9": 30, "161": 30, "though": 31, "affect": 31, "f401": 31, "original_data_path": 31, "input_col": 31, "na": 31, "workaround": 31, "fillna": 31, "extend": [33, 34], "friendli": [33, 34], "accustom": [33, 34], "accommod": [33, 34], "100k": [33, 34], "proce": [33, 34], "breez": [33, 34], "bayesianpersonalizedrank": 33, "entertain": [33, 34], "get_movielen": [33, 34], "variant": [33, 34], "henc": 33, "train_workflow": [33, 34], "681291": 33, "681740": 33, "681877": 33, "frame": [33, 34], "777": 33, "52it": 33, "train_auc": 33, "incom": [33, 34], "lifecycl": [33, 34], "redefin": [33, 34], "subtl": [33, 34], "inf_workflow": [33, 34], "wrap": [33, 34], "predictimplicit": 33, "implicit_model": 33, "ten_exampl": [33, 34], "to_fram": 33, "143": 33, "609": 33, "858": 33, "predictions_from_triton": [33, 34], "local_predict": [33, 34], "assert_allclos": 33, "xgb": 34, "preprocess_categori": 34, "preprocess_r": 34, "rating_binari": 34, "num_boost_round": 34, "verbose_ev": 34, "446602": 34, "447091": 34, "447227": 34, "94mb": 34, "45mb": 34, "32file": 34, "947": 34, "diskutil": 34, "stale": 34, "lock": 34, "oqemvhkv": 34, "purg": 34, "41809": 34, "got": 34, "rmse": 34, "36952": 34, "95316": 34, "92447": 34, "90741": 34, "89437": 34, "89138": 34, "fil": 34, "predictforest": 34, "te_movieid_r": 34, "userid_count": 34, "zip_cod": 34, "77": 34, "779876": 34, "572154": 34, "231": 34, "896619": 34, "goldeney": 34, "366": 34, "954632": 34, "room": 34, "093809": 34, "shorti": 34, "383": 34, "539376": 34, "copycat": 34, "perhap": 34, "35647": 34, "allclos": 34, "station": [35, 36, 37, 42, 43, 44], "ubuntu": [35, 36, 37, 42, 43, 44], "certifi": [35, 36, 37, 42, 43, 44], "465": [35, 36, 37, 42, 43, 44], "center": [35, 36, 37, 42, 43, 44], "formerli": [35, 36, 37, 42, 43, 44], "418": [35, 36, 37, 42, 43, 44], "r418": [35, 36, 37, 42, 43, 44], "440": [35, 36, 37, 42, 43, 44], "r440": [35, 36, 37, 42, 43, 44], "450": [35, 36, 37, 42, 43, 44], "r450": [35, 36, 37, 42, 43, 44], "r460": [35, 36, 37, 42, 43, 44], "amper": [35, 36, 37, 42, 43, 44], "ture": [35, 36, 37, 42, 43, 44], "volta": [35, 36, 37, 42, 43, 44], "pascal": [35, 36, 37, 42, 43, 44], "023": [35, 36, 37, 42, 43, 44], "011": [35, 36, 37, 42, 43, 44], "065": [35, 36, 37, 42, 43, 44], "rmm": [35, 36, 37, 42, 43, 44], "00a": [35, 36, 37, 42, 43, 44], "gf6bf047": [35, 36, 37, 42, 43, 44], "dirti": [35, 36, 37, 42, 43, 44], "g6ca81bbc78": [35, 36, 37, 42, 43, 44], "131": [35, 36, 37, 42, 43, 44], "cuda11": [35, 36, 37, 42, 43, 44], "95": [35, 42], "017": [35, 36, 37, 42, 43, 44], "014": [35, 36, 37, 42, 43, 44], "022": [35, 36, 37, 42, 43, 44], "76": [35, 36, 37, 42, 43, 44], "g185c18e6": [35, 36, 37, 42, 43, 44], "319": [35, 36, 37, 42, 43, 44], "g97422602b8": [35, 36, 37, 42, 43, 44], "g1354dcf": [35, 36, 37, 42, 43, 44], "g042ee198": [35, 37, 42, 44], "g886cf6de": [35, 36, 37, 42, 43, 44], "gf1e2bf5": [35, 37, 42, 44], "g712b04d": [35, 36, 37, 42, 43, 44], "g32ed5fa12": [35, 36, 37, 42, 43, 44], "g3367d725": [35, 37, 42, 44], "g046077b8": [36, 43], "g07bf4ab": [36, 43], "1a": 38, "1b": 38, "ncf": 38, "criteo": 38, "guidanc": 38, "2b": 38, "yellow": 38, "cosin": 38, "latenc": 38, "2a": 38, "mf": 38, "delai": 38, "retrain": 38, "hi": 38, "her": 38, "intent": 38, "viewed_item_id": 38, "4a": 38, "nlp": 38, "rnn": 38, "gru": 38, "lstm": 38, "4b": 38, "othersequenti": 38, "casual": 38, "clm": 38, "mlm": 38, "gru4rec": 38, "lstm4rec": 38, "gpt": 38, "bert": 38, "xl": 38, "electra": 38, "introduct": [38, 40], "bi": 38, "flexibl": 38, "divers": 39, "bring": 40, "older": 41}, "objects": {}, "objtypes": {}, "objnames": {}, "titleterms": {"nvidia": [0, 14], "merlin": [0, 1, 2, 3, 10, 13, 14, 15, 16, 18, 23, 24, 25, 26, 30, 31, 33, 34, 40, 41, 42, 43, 44], "benefit": 0, "compon": 0, "instal": 0, "exampl": [0, 4, 14, 23, 25], "notebook": [0, 4, 14, 23, 25], "tutori": 0, "i": 0, "built": 0, "On": 0, "feedback": 0, "support": [0, 22, 41, 42, 43, 44], "contain": [1, 16, 26, 35, 36, 37, 42, 43, 44], "structur": 1, "chang": 1, "begin": 1, "22": [1, 35, 36, 37, 42, 43, 44], "06": 1, "releas": 1, "build": [2, 14, 24, 26], "intellig": 2, "recommend": [2, 3, 4, 12, 24, 26, 33, 34, 38, 39], "system": [2, 3, 4, 18, 24, 33, 34, 39], "overview": [2, 5, 6, 8, 9, 10, 11, 28, 29, 30, 31, 33, 34], "learn": [2, 5, 6, 8, 10, 11, 16, 18, 20, 22, 28, 29, 30, 31, 33, 34], "object": [2, 3, 5, 6, 8, 10, 11, 18, 28, 29, 30, 31, 33, 34], "import": [2, 3, 20, 24], "requir": [2, 3], "librari": [2, 3, 24], "function": [2, 3], "set": [2, 3, 19, 22], "up": [2, 3, 26], "featur": [2, 3, 9, 19, 22, 24], "store": [2, 3], "feast": 2, "export": [2, 3, 24], "user": 2, "item": [2, 5, 6], "engin": [2, 19, 24], "nvtabular": [2, 5, 8, 24, 28, 31], "train": [2, 5, 6, 7, 9, 10, 11, 14, 16, 19, 22, 23, 24, 25, 26, 29, 30, 33, 34], "retriev": [2, 3, 24, 26, 33, 34, 38], "model": [2, 5, 6, 9, 11, 12, 16, 17, 18, 20, 22, 23, 24, 26, 30, 31, 33, 34, 38], "two": 2, "tower": 2, "queri": 2, "rank": [2, 16, 17, 18, 20, 22, 23, 24, 38], "dlrm": [2, 22, 24, 30], "extract": 2, "save": [2, 11, 24, 30, 31], "embed": [2, 6, 22], "creat": [2, 3, 6, 17, 21, 24, 26, 33, 34], "definit": [2, 11, 19], "next": [2, 5, 6, 8, 30, 31], "step": [2, 8, 30, 31], "deploi": [3, 4, 14, 17, 18, 23, 24, 28, 31], "multi": [3, 4, 14, 16, 20, 22], "stage": [3, 4, 14, 22], "recsi": [3, 14, 22], "product": 3, "triton": [3, 5, 12, 16, 17, 18, 24, 26, 31, 33, 34], "infer": [3, 5, 11, 12, 16, 17, 18, 24, 26, 31, 33, 34], "server": [3, 5, 12, 16, 17, 18, 24, 26, 31, 33, 34], "regist": [3, 26], "our": [3, 8, 9, 11, 28], "load": [3, 12], "from": [3, 12, 24, 26, 31, 33, 34], "offlin": 3, "an": [3, 16, 33, 34], "onlin": 3, "faiss": 3, "index": 3, "client": [3, 21], "ensembl": [3, 17, 24, 31, 33, 34], "graph": [3, 17, 24, 33, 34], "start": [3, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 18, 21, 24, 33, 34], "run": [4, 8, 14, 20, 23, 25], "transform": [5, 6, 26], "base": [5, 6, 22, 38], "architectur": [5, 6, 10, 22], "predict": [5, 6, 12, 31], "task": [5, 6, 16, 19, 20, 22], "download": [5, 6, 7, 16, 27], "prepar": [5, 6, 11, 16, 33, 34], "dataset": [5, 6, 7, 13, 14, 16, 19, 26, 27, 32, 33, 34], "preprocess": [5, 8, 16, 19, 28], "construct": [5, 6], "evalu": [5, 22, 30], "serv": [5, 6, 12, 14, 25, 26, 33, 34], "us": [5, 9, 26], "summari": [5, 6, 18, 24, 30, 31], "pretrain": 6, "clean": [6, 26], "data": [6, 11, 16, 19, 28, 33, 34], "gener": [6, 26], "synthet": 6, "workflow": [6, 31], "get": [7, 8, 9, 10, 11, 13, 14, 31], "movielen": [7, 8, 9, 10, 11, 13, 14], "convert": [7, 27], "movielens25m": [7, 10, 11], "split": [7, 19], "valid": 7, "etl": [8, 28], "defin": [8, 9, 10, 19, 22, 28], "pipelin": [8, 28], "check": 8, "pre": 8, "process": [8, 20, 21], "output": [8, 19, 22], "hugectr": [9, 29, 42], "why": 9, "other": 9, "scale": [9, 14, 27, 28, 29, 30, 31, 32], "acceler": 9, "let": 9, "": 9, "pytorch": [10, 43], "initi": 10, "dataload": 10, "neural": [10, 20], "network": 10, "tensorflow": [11, 12, 30, 31, 44], "specifi": 11, "hyperparamet": [11, 16, 20, 21], "launch": [12, 17], "inventori": 14, "larg": [14, 32], "criteo": [14, 27, 28, 29, 30, 31, 32], "aw": [14, 25, 26], "sagemak": [14, 25, 26], "quick": [15, 16, 21], "setup": [16, 20, 21], "tenrec": 16, "docker": [16, 26], "mmoe": [16, 22], "tune": [16, 20, 21], "deploy": 16, "send": [17, 26], "request": [17, 26], "command": [17, 19, 22], "line": [17, 19, 22], "argument": [17, 19, 22], "input": [17, 19, 22], "script": [19, 22, 26, 27], "best": [19, 20, 22], "practic": [19, 22], "deal": [19, 22], "high": [19, 22], "cardin": [19, 22], "target": 19, "encod": 19, "cast": 19, "filter": 19, "frequenc": [19, 28], "cap": 19, "eval": 19, "cuda": 19, "cluster": [19, 28], "option": 19, "benchmark": 20, "singl": 20, "stl": 20, "result": [20, 21], "most": 20, "mtl": [20, 22], "analysi": 20, "varianc": 20, "accuraci": 20, "its": 20, "improv": 20, "over": 20, "refin": 20, "search": 20, "space": [20, 22], "weight": [21, 22], "bias": 21, "sweep": 21, "1": [21, 26], "config": 21, "file": 21, "2": [21, 26], "environ": 21, "wandb": 21, "3": [21, 26], "test": 21, "4": 21, "5": 21, "hypertun": 21, "monitor": 21, "analys": 21, "baselin": 22, "categor": 22, "size": 22, "regular": 22, "class": 22, "neg": 22, "sampl": 22, "loss": 22, "dcn": 22, "v2": 22, "deepfm": 22, "wide": 22, "deep": 22, "cgc": 22, "ple": 22, "expert": 22, "log": [22, 27], "verif": 24, "artifact": [24, 26], "part": 26, "imag": [26, 35, 36, 37, 42, 43, 44], "dockerfil": 26, "your": 26, "python": 26, "sdk": 26, "raw": 26, "termin": 26, "endpoint": 26, "1tb": 27, "click": 27, "convers": 27, "csv": 27, "parquet": 27, "prep": 28, "distribut": 28, "dask": 28, "threshold": 28, "implicit": 33, "tool": [33, 34], "prerequisit": [33, 34], "xgboost": 34, "23": [35, 36, 37, 42, 43, 44], "xx": [35, 36, 37, 42, 43, 44], "resourc": [38, 40], "sequenti": 38, "session": 38, "guid": 39, "relat": 40, "matrix": [41, 42, 43, 44]}, "envversion": {"sphinx.domains.c": 2, "sphinx.domains.changeset": 1, "sphinx.domains.citation": 1, "sphinx.domains.cpp": 8, "sphinx.domains.index": 1, "sphinx.domains.javascript": 2, "sphinx.domains.math": 2, "sphinx.domains.python": 3, "sphinx.domains.rst": 2, "sphinx.domains.std": 2, "sphinx.ext.viewcode": 1, "sphinx.ext.intersphinx": 1, "sphinx": 57}, "alltitles": {"NVIDIA Merlin": [[0, "nvidia-merlin"]], "Benefits": [[0, "benefits"]], "Components of NVIDIA Merlin": [[0, "components-of-nvidia-merlin"]], "Installation": [[0, "installation"]], "Example Notebooks and Tutorials": [[0, "example-notebooks-and-tutorials"]], "Merlin Is Built On": [[0, "merlin-is-built-on"]], "Feedback and Support": [[0, "feedback-and-support"]], "Merlin Containers": [[1, "merlin-containers"]], "Structural Changes Beginning with the 22.06 Releases": [[1, "structural-changes-beginning-with-the-22-06-releases"]], "Building Intelligent Recommender Systems with Merlin": [[2, "building-intelligent-recommender-systems-with-merlin"]], "Overview": [[2, "overview"], [5, "overview"], [6, "overview"], [8, "overview"], [9, "overview"], [10, "overview"], [11, "overview"], [28, "overview"], [29, "overview"], [30, "overview"], [31, "overview"], [33, "overview"], [34, "overview"]], "Learning objectives": [[2, "learning-objectives"], [5, "learning-objectives"], [6, "learning-objectives"], [8, "learning-objectives"], [10, "learning-objectives"], [11, "learning-objectives"], [28, "learning-objectives"], [29, "learning-objectives"], [30, "learning-objectives"], [31, "learning-objectives"], [33, "learning-objectives"], [34, "learning-objectives"]], "Import required libraries and functions": [[2, "import-required-libraries-and-functions"], [3, "import-required-libraries-and-functions"]], "Set up a feature store with Feast": [[2, "set-up-a-feature-store-with-feast"]], "Exporting user and item features": [[2, "exporting-user-and-item-features"]], "Feature Engineering with NVTabular": [[2, "feature-engineering-with-nvtabular"], [2, "id1"], [24, "feature-engineering-with-nvtabular"]], "Training a Retrieval Model with Two-Tower Model": [[2, "training-a-retrieval-model-with-two-tower-model"]], "Exporting query (user) model": [[2, "exporting-query-user-model"]], "Training a Ranking Model with DLRM": [[2, "training-a-ranking-model-with-dlrm"]], "Extract and save Item embeddings": [[2, "extract-and-save-item-embeddings"]], "Create feature definitions": [[2, "create-feature-definitions"]], "Next Steps": [[2, "next-steps"], [8, "next-steps"]], "Deploying a Multi-Stage RecSys into Production with Merlin Systems and Triton Inference Server": [[3, "deploying-a-multi-stage-recsys-into-production-with-merlin-systems-and-triton-inference-server"]], "Register our features on feature store": [[3, "register-our-features-on-feature-store"]], "Loading features from offline store into an online store": [[3, "loading-features-from-offline-store-into-an-online-store"]], "Set up Faiss index, create feature store client and objects for the Triton ensemble": [[3, "set-up-faiss-index-create-feature-store-client-and-objects-for-the-triton-ensemble"]], "Export Graph as Ensemble": [[3, "export-graph-as-ensemble"], [24, "export-graph-as-ensemble"]], "Starting Triton Server": [[3, "starting-triton-server"], [24, "starting-triton-server"]], "Retrieving Recommendations from Triton": [[3, "retrieving-recommendations-from-triton"]], "Deploying a Multi-Stage Recommender System": [[4, "deploying-a-multi-stage-recommender-system"]], "Running the Example Notebooks": [[4, "running-the-example-notebooks"], [14, "running-the-example-notebooks"], [23, "running-the-example-notebooks"]], "Transformer-based architecture for next-item prediction task": [[5, "transformer-based-architecture-for-next-item-prediction-task"]], "Downloading and preparing the dataset": [[5, "downloading-and-preparing-the-dataset"], [6, "downloading-and-preparing-the-dataset"]], "Preprocessing with NVTabular": [[5, "preprocessing-with-nvtabular"]], "Constructing the model": [[5, "constructing-the-model"]], "Model training": [[5, "model-training"]], "Model evaluation": [[5, "model-evaluation"]], "Serving predictions using the Triton Inference Server": [[5, "serving-predictions-using-the-triton-inference-server"]], "Summary": [[5, "summary"], [6, "summary"], [18, "summary"], [24, "summary"], [30, "summary"], [31, "summary"]], "Transformer-based architecture for next-item prediction task with pretrained embeddings": [[6, "transformer-based-architecture-for-next-item-prediction-task-with-pretrained-embeddings"]], "Clean downloaded data": [[6, "clean-downloaded-data"]], "Generate synthetic data": [[6, "generate-synthetic-data"]], "Constructing a workflow": [[6, "constructing-a-workflow"]], "Creating and training the model": [[6, "creating-and-training-the-model"]], "Serving predictions": [[6, "serving-predictions"]], "Getting Started MovieLens: Download and Convert": [[7, "getting-started-movielens-download-and-convert"]], "MovieLens25M": [[7, "movielens25m"], [10, "movielens25m"], [11, "movielens25m"]], "Download the dataset": [[7, "download-the-dataset"]], "Convert the dataset": [[7, "convert-the-dataset"]], "Splitting into train and validation dataset": [[7, "splitting-into-train-and-validation-dataset"]], "Getting Started MovieLens: ETL with NVTabular": [[8, "getting-started-movielens-etl-with-nvtabular"]], "ETL with NVTabular": [[8, "etl-with-nvtabular"], [28, "etl-with-nvtabular"]], "Defining our Preprocessing Pipeline": [[8, "defining-our-preprocessing-pipeline"], [28, "defining-our-preprocessing-pipeline"]], "Running the pipeline": [[8, "running-the-pipeline"]], "Checking the pre-processing outputs": [[8, "checking-the-pre-processing-outputs"]], "Getting Started MovieLens: Training with HugeCTR": [[9, "getting-started-movielens-training-with-hugectr"]], "Why use HugeCTR?": [[9, "why-use-hugectr"]], "Other Features of HugeCTR": [[9, "other-features-of-hugectr"]], "Getting Started": [[9, "getting-started"]], "Scaling Accelerated training with HugeCTR": [[9, "scaling-accelerated-training-with-hugectr"]], "Let\u2019s define our model": [[9, "let-s-define-our-model"]], "Getting Started MovieLens: Training with PyTorch": [[10, "getting-started-movielens-training-with-pytorch"]], "Initializing the Merlin Dataloader for PyTorch": [[10, "initializing-the-merlin-dataloader-for-pytorch"]], "Defining Neural Network Architecture": [[10, "defining-neural-network-architecture"]], "Getting Started MovieLens: Training with TensorFlow": [[11, "getting-started-movielens-training-with-tensorflow"]], "Data Preparation": [[11, "data-preparation"]], "Training our model": [[11, "training-our-model"], [11, "id1"]], "Model definition": [[11, "model-definition"]], "Specifying Hyperparameters": [[11, "specifying-hyperparameters"]], "Saving the model for inference": [[11, "saving-the-model-for-inference"]], "Serve Recommendations from the TensorFlow Model": [[12, "serve-recommendations-from-the-tensorflow-model"]], "Launching and Starting the Triton Server": [[12, "launching-and-starting-the-triton-server"]], "Loading the Tensorflow Model with Triton Inference Server": [[12, "loading-the-tensorflow-model-with-triton-inference-server"]], "Predicting": [[12, "predicting"]], "Getting Started with Merlin and the MovieLens Dataset": [[13, "getting-started-with-merlin-and-the-movielens-dataset"]], "NVIDIA Merlin Example Notebooks": [[14, "nvidia-merlin-example-notebooks"]], "Inventory": [[14, "inventory"]], "Building and Deploying a multi-stage RecSys": [[14, "building-and-deploying-a-multi-stage-recsys"]], "Getting Started with MovieLens": [[14, "getting-started-with-movielens"]], "Scaling Large Datasets with Criteo": [[14, "scaling-large-datasets-with-criteo"], [32, "scaling-large-datasets-with-criteo"]], "Training and Serving with Merlin on AWS SageMaker": [[14, "training-and-serving-with-merlin-on-aws-sagemaker"]], "Quick-start for Merlin": [[15, "quick-start-for-merlin"]], "Quick-start for ranking models with Merlin": [[16, "quick-start-for-ranking-models-with-merlin"]], "Setup": [[16, "setup"]], "Download the TenRec dataset": [[16, "download-the-tenrec-dataset"]], "Preparing the data": [[16, "preparing-the-data"]], "Start Docker container": [[16, "start-docker-container"]], "Preprocessing": [[16, "preprocessing"]], "Training a ranking model": [[16, "training-a-ranking-model"]], "Training a ranking model with multi-task learning": [[16, "training-a-ranking-model-with-multi-task-learning"]], "Training an MMOE model": [[16, "training-an-mmoe-model"]], "Hyperparameter tuning": [[16, "hyperparameter-tuning"]], "Model Deployment on Triton Inference Server": [[16, "model-deployment-on-triton-inference-server"]], "Deploying a Ranking model on Triton Inference Server": [[17, "deploying-a-ranking-model-on-triton-inference-server"]], "Creating the Ensemble Graph": [[17, "creating-the-ensemble-graph"]], "Launching Triton Inference Server": [[17, "launching-triton-inference-server"]], "Sending request to Triton": [[17, "sending-request-to-triton"]], "Command line arguments": [[17, "command-line-arguments"], [19, "command-line-arguments"], [22, "command-line-arguments"]], "Inputs": [[17, "inputs"], [19, "inputs"], [22, "inputs"]], "Deploying Ranking Models with Merlin Systems": [[18, "deploying-ranking-models-with-merlin-systems"]], "Learning Objectives": [[18, "learning-objectives"]], "Starting Triton Inference Server": [[18, "starting-triton-inference-server"]], "Deploy models on Triton Inference Server": [[18, "deploy-models-on-triton-inference-server"]], "Preprocessing script": [[19, "preprocessing-script"]], "Best practices": [[19, "best-practices"], [22, "best-practices"]], "Dataset": [[19, "dataset"], [33, "dataset"], [34, "dataset"]], "Defining the task": [[19, "defining-the-task"]], "Preprocessing features": [[19, "preprocessing-features"]], "Dealing with high-cardinality data": [[19, "dealing-with-high-cardinality-data"]], "Feature Engineering": [[19, "feature-engineering"]], "Data set splitting": [[19, "data-set-splitting"]], "Outputs": [[19, "outputs"], [22, "outputs"]], "Features and targets definition": [[19, "features-and-targets-definition"]], "Target encoding features": [[19, "target-encoding-features"]], "Data casting and filtering": [[19, "data-casting-and-filtering"]], "Filtering and frequency capping": [[19, "filtering-and-frequency-capping"]], "Dataset splitting (train and eval sets)": [[19, "dataset-splitting-train-and-eval-sets"]], "CUDA cluster options": [[19, "cuda-cluster-options"]], "Benchmark of ranking models": [[20, "benchmark-of-ranking-models"]], "Neural ranking models.": [[20, "neural-ranking-models"]], "Hyperparameter tuning setup": [[20, "hyperparameter-tuning-setup"]], "Single-task learning": [[20, "single-task-learning"]], "STL Benchmark results": [[20, "stl-benchmark-results"]], "Most important hyperparameters": [[20, "most-important-hyperparameters"], [20, "id1"]], "Best runs hyperparameters": [[20, "best-runs-hyperparameters"], [20, "id2"]], "Multi-task learning": [[20, "multi-task-learning"], [22, "multi-task-learning"]], "MTL Benchmark results": [[20, "mtl-benchmark-results"]], "Analysis of the hyperparameter tuning process": [[20, "analysis-of-the-hyperparameter-tuning-process"]], "Variance of accuracy and its improvement over runs": [[20, "variance-of-accuracy-and-its-improvement-over-runs"]], "Refine the search space": [[20, "refine-the-search-space"]], "Hyperparameter tuning with Quick-start and Weights&Biases Sweeps": [[21, "hyperparameter-tuning-with-quick-start-and-weights-biases-sweeps"]], "1. Creating the Sweeps config file": [[21, "creating-the-sweeps-config-file"]], "2. Environment setup": [[21, "environment-setup"]], "2.1 Setup wandb client": [[21, "setup-wandb-client"]], "3. Testing the environment": [[21, "testing-the-environment"]], "4. Creating the Sweep at Weights&Biases": [[21, "creating-the-sweep-at-weights-biases"]], "5. Starting the hypertuning process": [[21, "starting-the-hypertuning-process"]], "5. Monitoring the hypertuning process & analysing results": [[21, "monitoring-the-hypertuning-process-analysing-results"]], "Ranking script": [[22, "ranking-script"]], "Ranking in multi-stage RecSys": [[22, "ranking-in-multi-stage-recsys"]], "Multi-task learning for ranking models": [[22, "multi-task-learning-for-ranking-models"]], "Supported models": [[22, "supported-models"]], "Baseline ranking architectures": [[22, "baseline-ranking-architectures"]], "Multi-task learning architectures": [[22, "multi-task-learning-architectures"]], "Modeling inputs features": [[22, "modeling-inputs-features"]], "Dealing with high-cardinality categorical features": [[22, "dealing-with-high-cardinality-categorical-features"]], "Defining the embedding size": [[22, "defining-the-embedding-size"]], "Regularization": [[22, "regularization"]], "Classes weights": [[22, "classes-weights"]], "Negative sampling": [[22, "negative-sampling"], [22, "id1"]], "Losses weights": [[22, "losses-weights"]], "Setting tasks sample space": [[22, "setting-tasks-sample-space"]], "Tasks": [[22, "tasks"]], "Model": [[22, "model"]], "DCN-v2": [[22, "dcn-v2"]], "DLRM and DeepFM": [[22, "dlrm-and-deepfm"]], "Wide&Deep": [[22, "wide-deep"]], "Wide&Deep and DeepFM": [[22, "wide-deep-and-deepfm"]], "MMOE": [[22, "mmoe"]], "CGC and PLE": [[22, "cgc-and-ple"]], "Expert-based MTL models": [[22, "expert-based-mtl-models"]], "Multi-task learning models": [[22, "multi-task-learning-models"]], "Training and evaluation": [[22, "training-and-evaluation"]], "Logging": [[22, "logging"]], "Training and Deploying Ranking models with Merlin": [[23, "training-and-deploying-ranking-models-with-merlin"]], "Exporting Ranking Models": [[24, "exporting-ranking-models"]], "Importing Libraries": [[24, "importing-libraries"]], "Build and Train a DLRM model": [[24, "build-and-train-a-dlrm-model"]], "Save model": [[24, "save-model"]], "Deploying the model with Merlin Systems": [[24, "deploying-the-model-with-merlin-systems"]], "Create the Ensemble Graph": [[24, "create-the-ensemble-graph"], [33, "create-the-ensemble-graph"], [34, "create-the-ensemble-graph"]], "Verification of Ensemble Artifacts": [[24, "verification-of-ensemble-artifacts"]], "Retrieving Recommendations from Triton Inference Server": [[24, "retrieving-recommendations-from-triton-inference-server"], [33, "retrieving-recommendations-from-triton-inference-server"], [34, "retrieving-recommendations-from-triton-inference-server"]], "Training and Serving Merlin on AWS SageMaker": [[25, "training-and-serving-merlin-on-aws-sagemaker"], [26, "training-and-serving-merlin-on-aws-sagemaker"]], "Running the Example Notebook": [[25, "running-the-example-notebook"]], "Part 1: Generating Dataset and Docker image": [[26, "part-1-generating-dataset-and-docker-image"]], "Generating Dataset": [[26, "generating-dataset"]], "Training Script": [[26, "training-script"]], "Create the Dockerfile": [[26, "create-the-dockerfile"]], "Building and registering the container": [[26, "building-and-registering-the-container"]], "Part 2: Training your Merlin model on Sagemaker": [[26, "part-2-training-your-merlin-model-on-sagemaker"]], "Training on Sagemaker using the Python SDK": [[26, "training-on-sagemaker-using-the-python-sdk"]], "Part 3: Retrieving Recommendations from Triton Inference Server": [[26, "part-3-retrieving-recommendations-from-triton-inference-server"]], "Send a Request to Triton Inference Server to Transform a Raw Dataset": [[26, "send-a-request-to-triton-inference-server-to-transform-a-raw-dataset"]], "Terminate endpoint and clean up artifacts": [[26, "terminate-endpoint-and-clean-up-artifacts"]], "Scaling Criteo: Download and Convert": [[27, "scaling-criteo-download-and-convert"]], "Criteo 1TB Click Logs dataset": [[27, "criteo-1tb-click-logs-dataset"]], "Conversion Script for Criteo Dataset (CSV-to-Parquet)": [[27, "conversion-script-for-criteo-dataset-csv-to-parquet"]], "Scaling Criteo: ETL with NVTabular": [[28, "scaling-criteo-etl-with-nvtabular"]], "Data Prep": [[28, "data-prep"]], "Deploy a Distributed-Dask Cluster": [[28, "deploy-a-distributed-dask-cluster"]], "Frequency Thresholding": [[28, "frequency-thresholding"]], "Scaling Criteo: Training with HugeCTR": [[29, "scaling-criteo-training-with-hugectr"]], "Training with HugeCTR": [[29, "training-with-hugectr"]], "Scaling Criteo: Training with Merlin Models TensorFlow": [[30, "scaling-criteo-training-with-merlin-models-tensorflow"]], "Training a DLRM model": [[30, "training-a-dlrm-model"]], "Evaluate the model": [[30, "evaluate-the-model"]], "Save the model": [[30, "save-the-model"]], "Next steps": [[30, "next-steps"], [31, "next-steps"]], "Scaling Criteo: Triton Inference with Merlin Models TensorFlow": [[31, "scaling-criteo-triton-inference-with-merlin-models-tensorflow"]], "Saved NVTabular workflow and Merlin Models": [[31, "saved-nvtabular-workflow-and-merlin-models"]], "Deploying Ensemble to Triton Inference Server": [[31, "deploying-ensemble-to-triton-inference-server"]], "Get prediction from Triton Inference Server": [[31, "get-prediction-from-triton-inference-server"]], "Serving an Implicit Model with Merlin Systems": [[33, "serving-an-implicit-model-with-merlin-systems"]], "Tools": [[33, "tools"], [34, "tools"]], "Prerequisite: Preparing the data and Training Implicit": [[33, "prerequisite-preparing-the-data-and-training-implicit"]], "Starting the Triton Inference Server": [[33, "starting-the-triton-inference-server"], [34, "starting-the-triton-inference-server"]], "Serving an XGBoost Model with Merlin Systems": [[34, "serving-an-xgboost-model-with-merlin-systems"]], "Prerequisite: Preparing the data and Training XGBoost": [[34, "prerequisite-preparing-the-data-and-training-xgboost"]], "23.xx Container Images": [[35, "xx-container-images"], [36, "xx-container-images"], [37, "xx-container-images"], [42, "xx-container-images"], [43, "xx-container-images"], [44, "xx-container-images"]], "22.xx Container Images": [[35, "id13"], [36, "id13"], [37, "id13"], [42, "id13"], [43, "id13"], [44, "id13"]], "Recommender Models": [[38, "recommender-models"]], "Ranking Models": [[38, "ranking-models"]], "Resources for Ranking Models:": [[38, "resources-for-ranking-models"]], "Retrieval Models": [[38, "retrieval-models"]], "Resources for Retrieval Models:": [[38, "resources-for-retrieval-models"]], "Sequential and Session-Based Models": [[38, "sequential-and-session-based-models"]], "Resources for Session-Based Models:": [[38, "resources-for-session-based-models"]], "Recommender Systems Guide": [[39, "recommender-systems-guide"]], "Merlin": [[40, "merlin"]], "Related Resources": [[40, "related-resources"]], "Merlin Support Matrix": [[41, "merlin-support-matrix"]], "Merlin HugeCTR Support Matrix": [[42, "merlin-hugectr-support-matrix"]], "Merlin PyTorch Support Matrix": [[43, "merlin-pytorch-support-matrix"]], "Merlin TensorFlow Support Matrix": [[44, "merlin-tensorflow-support-matrix"]]}, "indexentries": {}}) \ No newline at end of file diff --git a/main/support_matrix/index.html b/main/support_matrix/index.html index 3f10df335..f0c724daa 100644 --- a/main/support_matrix/index.html +++ b/main/support_matrix/index.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -513,8 +512,8 @@

    Merlin Support Matrix - - + +
    diff --git a/main/support_matrix/support_matrix_merlin_hugectr.html b/main/support_matrix/support_matrix_merlin_hugectr.html index 0d3724ff4..22d6bfd5b 100644 --- a/main/support_matrix/support_matrix_merlin_hugectr.html +++ b/main/support_matrix/support_matrix_merlin_hugectr.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -1111,8 +1110,8 @@

    22.xx Container Images - + +
    diff --git a/main/support_matrix/support_matrix_merlin_pytorch.html b/main/support_matrix/support_matrix_merlin_pytorch.html index 9ea18cc10..0a7b2a70f 100644 --- a/main/support_matrix/support_matrix_merlin_pytorch.html +++ b/main/support_matrix/support_matrix_merlin_pytorch.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -148,6 +148,7 @@ +

    NVIDIA Merlin

    @@ -186,7 +187,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -205,7 +205,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -1116,8 +1115,8 @@

    22.xx Container Images - + +
    diff --git a/main/support_matrix/support_matrix_merlin_tensorflow.html b/main/support_matrix/support_matrix_merlin_tensorflow.html index 7d2cc85da..f93291596 100644 --- a/main/support_matrix/support_matrix_merlin_tensorflow.html +++ b/main/support_matrix/support_matrix_merlin_tensorflow.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -187,7 +188,6 @@
  • Training with PyTorch
  • -
  • Serving the HugeCTR Model with Triton
  • Serve Recommendations from the TensorFlow Model
  • @@ -206,7 +206,6 @@
  • Feature Engineering with NVTabular
  • Training with HugeCTR
  • Training with Merlin Models and TensorFlow
  • -
  • Deploy the HugeCTR Model with Triton
  • Deploy the TensorFlow Model with Triton
  • @@ -1126,8 +1125,8 @@

    22.xx Container Images - + +
    diff --git a/stable/README.html b/stable/README.html index 160c6032a..99b0c8b06 100644 --- a/stable/README.html +++ b/stable/README.html @@ -19,12 +19,12 @@ - - - + + + - + @@ -36,9 +36,9 @@ - - - + + + @@ -149,6 +149,7 @@ +

    NVIDIA Merlin

    @@ -583,7 +584,7 @@

    Installation

    Example Notebooks and Tutorials#

    -

    A collection of end-to-end examples are available in the form of Jupyter notebooks. +

    A collection of end-to-end examples are available in the form of Jupyter notebooks. The example notebooks demonstrate how to: