Llama-3_1-Nemotron-51B-Instruct / modeling_decilm.py

Commit History

fixed flash_attention backward_compat
c7f5725
verified

itlevy commited on

flash_attention_utils_backward_compat (#2)
186a08a
verified

itlevy commited on

transformers>=4.44.2
e9d7c68
verified

itlevy commited on