NER-Flair-ONNX

This code is used to convert and infer sequence-tagger model in ONNX type.
Move model to /checkpoints/best-model.pt
Save and move tag dictionary to /tag_dictionary/tag_dictionary.pkl via:

tag_dictionary.save("./tag_dictionary/tag_dictionary.pkl")

when traning

Environment

To create virtual environment venv:

virtualenv venv
source venv/bin/activate
pip install -r requirements.txt

Export ONNX

To convert embedding to onnx (if finetune embedding):

python3 main.py --scenario embedding

To convert sequence tagger model to onnx:

python3 main.py --scenario tagger

Testing

To test onnx:

python3 main.py --scenario test

Inference

To infer:

python3 onnx_infer.py

TODO

Check cuda providers
Deploy on triton