about deepspeed

[ChrisXULC]

I have question about deepspeed. Assuming that we did not use DeepSpeed in our code initially, if we now want to use it in our fine-tuning process, do we just need to call it in the run_finetune_with_lora.sh script? Are there any other files that need to be modified? Thanks~

[research4pan]

Thanks for your interest in LMFlow! You may run ./run_finetune_with_lora.sh first, which uses alpaca dataset and galactica-1.3b by default. To use your own dataset, you may provide your own dataset in {path-to-your-dataset}/*.json, with "text_only" typed format for those json files. Then changing

--dataset_path {path-to-your-dataset}

in this line will specify the dataset.

If you would like to change model, you may change the base model with

--model_name_or_path {a-huggingface-model-or-path-to-your-own-model}

where a model name provided by huggingface models or a local path to a model (can be a previously pretrained model) is specified here.

Hope that answers your question. Thanks 😄

[shizhediao]

This issue has been marked as stale because it has not had recent activity. If you think this still needs to be addressed please feel free to reopen this issue. Thanks

[liqinga]

deepspeed: error: unrecognized arguments: --dataset_path