423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-29 10:10:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

refactor mm training

Browse Source 
Former-commit-id: 179c0558699e287cbf38a2d73bff47e86d589c5a
This commit is contained in:
hiyouga
2024-08-30 02:14:31 +08:00
parent 618e357740
commit 228f745235
32 changed files with 505 additions and 472 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/llamafactory/model/model_utils/packing.py
View File

@@ -114,7 +114,7 @@ def get_unpad_data(attention_mask: "torch.Tensor") -> Tuple["torch.Tensor", "tor
def _patch_for_block_diag_attn(model_type: str) -> None:
require_version("transformers>=4.41.2,<=4.43.4", "To fix: pip install transformers>=4.41.2,<=4.43.4")
require_version("transformers>=4.41.2,<=4.44.3", "To fix: pip install transformers>=4.41.2,<=4.44.3")
if is_transformers_version_greater_than_4_43():
import transformers.modeling_flash_attention_utils