Closed zhuanask closed 10 months ago
The model parameters provided in drive.google are the LoRa weights after the first tuning stage. The adapter_model.bin is the final parameter of the LoRA weight for the first stage tuning and the Alpaca-LoRA-7B/checkpot-800(600/1000) is the checkpoint of step 800/1000 in the fine-tuning process of the first stage.
尊敬的作者您好,请问这个意思是说second finetune stage最终的lora weight没有在google drive里面吗?
尊敬的作者您好,请问如果想用您训练好的结果直接在evaluate上面做,base_model 和 output_dir 里应该放哪个文件呢?
What are the model parameters provided in drive.google? What I understand is that alpaca-lora-7B/adapter_model.bin is the parameter fine-tuned after the first stage instruction tuning, and alpaca-lora-7B/checkpot-800(600/1000) are the parameters after the second stage recommendation tuning, is that right?