Skip to content

[Bagel]: Support think mode in single stage deployment of Bagel#2650

Merged
lishunyang12 merged 5 commits into
vllm-project:mainfrom
princepride:single-stage-think-mode
Apr 13, 2026
Merged

[Bagel]: Support think mode in single stage deployment of Bagel#2650
lishunyang12 merged 5 commits into
vllm-project:mainfrom
princepride:single-stage-think-mode

Conversation

@princepride
Copy link
Copy Markdown
Collaborator

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.

Purpose

Support think mode in single stage deployment of Bagel.

Test Plan

python3 examples/offline_inference/bagel/end2end.py --model ByteDance-Seed/BAGEL-7B-MoT   --stage-configs-path vllm_omni/model_executor/stage_configs/bagel_single_stage.yaml   --think --modality text2img
python3 examples/offline_inference/bagel/end2end.py --model ByteDance-Seed/BAGEL-7B-MoT   --stage-configs-path vllm_omni/model_executor/stage_configs/bagel_single_stage.yaml   --think --modality img2img --image-path women.jpg   --prompts "Let the woman wear a blue dress"

Test Result

<think>
The model should generate an image of a small, fluffy cat with soft fur and a friendly expression.
Here’s the expanded prompt: A small, fluffy cat with soft, dense fur in warm tones like cream or light brown, featuring a gentle and friendly facial expression with large, expressive eyes and a slightly open mouth showing a hint of a smile. The setting is minimalistic and softly lit to emphasize the cat's delicate features, with a neutral background that enhances its charm without distraction. Ultra-realistic rendering with fine details capturing the texture of the fur and subtle shadows for depth.
</think>
image
<think>
The user wants the woman's dress color changed to blue while keeping the rest of the image unchanged. The structure, layout, and background should remain the same, ensuring only the dress color is altered to blue.
</think>
image

Signed-off-by: princepride <wangzhipeng628@gmail.com>
@chatgpt-codex-connector
Copy link
Copy Markdown

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

@princepride
Copy link
Copy Markdown
Collaborator Author

@natureofnature PTAL

@princepride
Copy link
Copy Markdown
Collaborator Author

I compared single stage deployment and multi stage deployment result. I found their totally different! Including think mode or non think mode, img2img or text2img.

Comment thread vllm_omni/diffusion/models/bagel/bagel_transformer.py
@lishunyang12 lishunyang12 enabled auto-merge (squash) April 13, 2026 08:26
@princepride princepride added the ready label to trigger buildkite CI label Apr 13, 2026
Signed-off-by: princepride <wangzhipeng628@gmail.com>
Signed-off-by: princepride <wangzhipeng628@gmail.com>
Signed-off-by: princepride <wangzhipeng628@gmail.com>
@lishunyang12 lishunyang12 merged commit 2c67c30 into vllm-project:main Apr 13, 2026
8 checks passed
Celeste-jq pushed a commit to IsleOfDawnlight/vllm-omni-voxcpm that referenced this pull request Apr 14, 2026
…lm-project#2650)

Signed-off-by: princepride <wangzhipeng628@gmail.com>
lengrongfu pushed a commit to lengrongfu/vllm-omni that referenced this pull request May 1, 2026
clodaghwalsh17 pushed a commit to clodaghwalsh17/nm-vllm-omni-ent that referenced this pull request May 12, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ready label to trigger buildkite CI

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants