Skip to content

Qwen2VL fix cos,sin dtypes to float when used with deepspeed #36188

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
zucchini-nlp merged 5 commits into from
Feb 18, 2025
Merged

Qwen2VL fix cos,sin dtypes to float when used with deepspeed #36188

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
6019576
fix dtype of cos,sin when used with deepspeed
Feb 14, 2025
ffaac84
Merge branch 'main' into fix_qwen2vl
ArdalanM Feb 14, 2025
0f0a7f1
move sin,cos casting withing flash attention functions
Feb 14, 2025
1583b64
fix cos,sin float casting in modular
Feb 18, 2025
033a3b8
Merge branch 'main' into fix_qwen2vl
ArdalanM Feb 18, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions src/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -198,6 +198,8 @@ def forward(
sin = emb.sin().float()
else:
cos, sin = position_embeddings
cos = cos.to(torch.float)
sin = sin.to(torch.float)
q, k = apply_rotary_pos_emb_flashatt(q.unsqueeze(0), k.unsqueeze(0), cos, sin)
q = q.squeeze(0)
k = k.squeeze(0)
Expand Down
2 changes: 2 additions & 0 deletions src/transformers/models/qwen2_vl/modeling_qwen2_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -322,6 +322,8 @@ def forward(
sin = emb.sin().float()
else:
cos, sin = position_embeddings
cos = cos.to(torch.float)
sin = sin.to(torch.float)
q, k = apply_rotary_pos_emb_vision(q, k, cos, sin)

attention_mask = torch.full(
Expand Down