[GH-ISSUE #45] cannot import name 'LlamaFlashAttention2' from 'transformers.models.llama.modeling_llama' #35

Open
opened 2026-10-26 21:10:36 +00:00 by bsakaguchi · 3 comments
Owner

Originally created by @panzhzh on GitHub (Jan 31, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-VL2/issues/45

Python version is above 3.10, patching the collections module.
Traceback (most recent call last):
File "/home/pzz/pythoncode/deepseek/vl2_load.py", line 4, in
from deepseek_vl2.models import DeepseekVLV2Processor, DeepseekVLV2ForCausalLM
File "/home/pzz/pythoncode/deepseek/DeepSeek-VL2-main/deepseek_vl2/models/init.py", line 21, in
from .modeling_deepseek_vl_v2 import DeepseekVLV2ForCausalLM
File "/home/pzz/pythoncode/deepseek/DeepSeek-VL2-main/deepseek_vl2/models/modeling_deepseek_vl_v2.py", line 28, in
from .modeling_deepseek import DeepseekV2ForCausalLM
File "/home/pzz/pythoncode/deepseek/DeepSeek-VL2-main/deepseek_vl2/models/modeling_deepseek.py", line 37, in
from transformers.models.llama.modeling_llama import (
ImportError: cannot import name 'LlamaFlashAttention2' from 'transformers.models.llama.modeling_llama' (/home/pzz/miniconda3/lib/python3.12/site-packages/transformers/models/llama/modeling_llama.py). Did you mean: 'LlamaAttention'?

Originally created by @panzhzh on GitHub (Jan 31, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-VL2/issues/45 Python version is above 3.10, patching the collections module. Traceback (most recent call last): File "/home/pzz/pythoncode/deepseek/vl2_load.py", line 4, in <module> from deepseek_vl2.models import DeepseekVLV2Processor, DeepseekVLV2ForCausalLM File "/home/pzz/pythoncode/deepseek/DeepSeek-VL2-main/deepseek_vl2/models/__init__.py", line 21, in <module> from .modeling_deepseek_vl_v2 import DeepseekVLV2ForCausalLM File "/home/pzz/pythoncode/deepseek/DeepSeek-VL2-main/deepseek_vl2/models/modeling_deepseek_vl_v2.py", line 28, in <module> from .modeling_deepseek import DeepseekV2ForCausalLM File "/home/pzz/pythoncode/deepseek/DeepSeek-VL2-main/deepseek_vl2/models/modeling_deepseek.py", line 37, in <module> from transformers.models.llama.modeling_llama import ( ImportError: cannot import name 'LlamaFlashAttention2' from 'transformers.models.llama.modeling_llama' (/home/pzz/miniconda3/lib/python3.12/site-packages/transformers/models/llama/modeling_llama.py). Did you mean: 'LlamaAttention'?
Author
Owner

@sgzqc commented on GitHub (Feb 18, 2025):

pip install transformers==4.47.1

<!-- gh-comment-id:2664827216 --> @sgzqc commented on GitHub (Feb 18, 2025): pip install transformers==4.47.1
Author
Owner

@mohayl commented on GitHub (Feb 18, 2025):

pip install transformers==4.47.1

does not work

<!-- gh-comment-id:2666382348 --> @mohayl commented on GitHub (Feb 18, 2025): > pip install transformers==4.47.1 does not work
Author
Owner

@DenceChen commented on GitHub (Mar 22, 2025):

pip install transformers==4.47.1

<!-- gh-comment-id:2744912731 --> @DenceChen commented on GitHub (Mar 22, 2025): pip install transformers==4.47.1
Sign in to join this conversation.
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepSeek-VL2#35
No description provided.