ros/LLaMA-Factory
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix incorrect loss value for vlms

Browse Source 
Former-commit-id: 0aa29a71ce958343a2086090d647eb63b8f5f5be
This commit is contained in:
hiyouga
2024-10-30 08:56:46 +00:00
parent 03213984ec
commit 8185eb1890
12 changed files with 48 additions and 22 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,<=4.46.0", "To fix: pip install transformers>=4.41.2,<=4.46.0")
require_version("transformers>=4.41.2,<=4.46.1", "To fix: pip install transformers>=4.41.2,<=4.46.1")
if is_transformers_version_greater_than_4_43():
import transformers.modeling_flash_attention_utils