Andy Wong comments

Repositories
Issues
Comments

Results 1 comments of


                                            Andy Wong

fix: the source part should not participate in loss calculation in SFT stage

1. The function is now moved to "DeepSpeedExamples/applications/DeepSpeed-Chat/dschat/utils/data/data_utils.py" 2. This solution seems to assume single-turn conversation, please consider cases for multi-turn conversation.