模型没merge进去

[brillianti]

Reminder

• I have read the README and searched the existing issues.

System Info

• llamafactory version: 0.9.1.dev0
• Platform: Linux-5.4.143.bsk.7-amd64-x86_64-with-glibc2.31
• Python version: 3.9.2
• PyTorch version: 2.3.1+cu121 (GPU)
• Transformers version: 4.46.1
• Datasets version: 2.19.1
• Accelerate version: 1.0.1
• PEFT version: 0.12.0
• TRL version: 0.9.6
• GPU type: NVIDIA A100-SXM4-80GB
• DeepSpeed version: 0.8.3

Reproduction

Note: DO NOT use quantized model or quantization_bit when merging lora adapters

model

model_name_or_path: /mnt/bn/lq-aigc/LLama_Factory_v2/LLaMA-Factory/checkpoint/Qwen/Qwen2-7B-Instruct
adapter_name_or_path: /mnt/bn/seed-aigc-aesthetic-lq/LLaMA-Factory-v2/LLaMA-Factory/outputs/outputs-1105/saves_rank64/simpo/simpo_8e-6_epoch_2_beta_10_gamma_5_align_test6/checkpoint-351
template: qwen
finetuning_type: lora

export

export_dir: /mnt/bn/seed-aigc-aesthetic-lq/lifanshi/LLaMA-Factory-v2/LLaMA-Factory/outputs/outputs-1105/Qwen/SFT/Qwen2-7B-Instruct—simpo_8e-6_epoch_2_beta_10_gamma_5_align_test6_ckpt1_test
export_size: 2
export_device: cpu
export_legacy_format: false

Expected behavior

我查看了output，发现lora都是正常load的，但是推理出来的时候还是保持了/mnt/bn/lq-aigc/LLama_Factory_v2/LLaMA-Factory/checkpoint/Qwen/Qwen2-7B-Instruct的输出 没有保持加lora的输出

Others

No response

[brillianti]

找到问题了，base model也就是【model_name_or_path】模型选错了，应该要找到adapter模型下面的config，直接索引到对应的base model