Skip to content

fix(qwen-image): drop the txt_seq_lens kwarg removed by diffusers #190

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
Rockdu merged 2 commits into from
Aug 18, 2026
Merged

fix(qwen-image): drop the txt_seq_lens kwarg removed by diffusers #190

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8e1c0db
fix(qwen-image): drop the txt_seq_lens kwarg removed by diffusers
Rockdu Aug 18, 2026
2d05873
chore(deps): pin diffusers to the commit that merged MiniMax H3
Aug 18, 2026
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
3 changes: 1 addition & 2 deletions miles/backends/fsdp_utils/configs/qwen_image.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,7 +114,7 @@ def collate_cond_for_sample_batch(
) -> dict:
"""Pad+stack per-sample encoder_hidden_states to (M, max_len, D), build
the corresponding (M, max_len) bool mask from txt_seq_lens, and
list-concat img_shapes / txt_seq_lens. Mask isn't transmitted from
list-concat img_shapes. Mask isn't transmitted from
rollout — it is fully derivable from txt_seq_lens which is.

When ``pad_to_len`` is given, pad to ``max(max(seq_lens), pad_to_len)``
Expand Down Expand Up @@ -165,7 +165,6 @@ def collate_cond_for_sample_batch(
return {
"encoder_hidden_states": encoder_hidden_states,
"encoder_hidden_states_mask": mask,
"txt_seq_lens": seq_lens,
"img_shapes": img_shapes,
}

Expand Down
5 changes: 4 additions & 1 deletion requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,10 @@ accelerate==1.12.0
addict==2.4.0
blobfile==3.0.0
datasets==4.4.2
diffusers==0.38.0
# MiniMax H3 (PR #14355) is merged into diffusers main but ships in no release yet:
# 0.39.0 predates the merge. Pinned to the merge commit on main; switch to the next
# stable release once it is out.
diffusers @ git+https://github.com/huggingface/diffusers.git@f53d552036a0d1bd5570782a39cd40cfabf112bc
fastapi==0.128.8
httpx[http2]==0.28.1
ltx-core @ git+https://github.com/Lightricks/LTX-2.git@780984275fd47128b02bef9b5c085404276866ee#subdirectory=packages/ltx-core
Expand Down
Loading