Skip to content

[trainer] fix: fallback vision tower to flash_attention_2 for Qwen2.5-VL when u… #4670

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
wuxibin89 merged 1 commit into from
Dec 29, 2025
Merged

[trainer] fix: fallback vision tower to flash_attention_2 for Qwen2.5-VL when u… #4670

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
9 changes: 9 additions & 0 deletions verl/workers/fsdp_workers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -332,6 +332,15 @@ def _build_model_optimizer(
if self.ulysses_sequence_parallel_size > 1 and hasattr(actor_model_config, "vision_config"):
actor_model_config.vision_config._attn_implementation = "eager"

# patch for qwen2.5-vl: when using flash_attention_3, set vision tower to use flash_attention_2
# because the vision tower does not support flash_attention_3
if (
getattr(actor_model_config, "model_type", None) == "qwen2_5_vl"
and attn_implementation == "flash_attention_3"
and hasattr(actor_model_config, "vision_config")
):
actor_model_config.vision_config._attn_implementation = "flash_attention_2"

# patch for kimi-vl
if getattr(actor_model_config, "model_type", None) == "kimi_vl":
actor_model_config.text_config.topk_method = "greedy"
Expand Down
Loading