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

fix inputs

Browse Source 
Former-commit-id: 7d535bb8cdf7e81edda81152e63c8cfe6c9dcc9f
This commit is contained in:
hiyouga
2024-11-23 18:25:45 +00:00
parent cd2485f28d
commit 5003820a6a
14 changed files with 148 additions and 95 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/llamafactory/model/model_utils/packing.py
View File

@@ -45,7 +45,7 @@ from transformers.utils.versions import require_version
from ...extras import logging
from ...extras.constants import SUPPORTED_CLASS_FOR_BLOCK_DIAG_ATTN
from ...extras.packages import is_transformers_version_greater_than_4_43
from ...extras.packages import is_transformers_version_greater_than
if TYPE_CHECKING:
@@ -115,7 +115,7 @@ 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,<=4.46.1", "To fix: pip install transformers>=4.41.2,<=4.46.1")
if is_transformers_version_greater_than_4_43():
if is_transformers_version_greater_than("4.43.0"):
import transformers.modeling_flash_attention_utils
transformers.modeling_flash_attention_utils._get_unpad_data = get_unpad_data