Megatron-DeepSpeed
4179f2a6 - HACK: in order to evaluate gpt on prefix lm loss

Commit

4 years ago

HACK: in order to evaluate gpt on prefix lm loss

References

thomas/evaluate_gpt_on_prefix_lm_loss

Author

thomasw21

thomasw21

Committer

thomasw21

thomasw21

Parents

Loading