Fix processor kwargs qwen2 vl #36890
Conversation
|
Hi 👋, thank you for opening this pull request! The pull request is converted to draft by default. When it is ready for review, please click the |
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
|
Ah great, I was banging my head with this ! |
qubvel
left a comment
qubvel
left a comment
There was a problem hiding this comment.
Nice catch, thanks for fixing!
| min_pixels = size["shortest_edge"] | ||
| else: | ||
| size = self.size | ||
| size = {**self.size} |
There was a problem hiding this comment.
| size = {**self.size} | |
| size = self.size.copy() |
| min_pixels = size["shortest_edge"] | ||
| else: | ||
| size = self.size | ||
| size = {**self.size} |
There was a problem hiding this comment.
| size = {**self.size} | |
| size = self.size.copy() |
There was a problem hiding this comment.
Ah missed this comment sorry, will fix in a later PR
ArthurZucker
left a comment
ArthurZucker
left a comment
There was a problem hiding this comment.
LGTM waiting for comments to be adressed
yonigozlan
added
the
for patch
* Fix qwen2_vl and qwen2_5_vl processors cutom images kwargs * change version warning
|
This changes result in error in newest vllm 0.8.2 docker run --runtime nvidia --gpus all -d --name vllm-UI-Tars-7b --restart unless-stopped -v ~/.cache/huggingface:/root/.cache/huggingface -e VLLM_USE_V1=0 -p 8000:8000 vllm/vllm-openai:v0.8.2 --model bytedance-research/UI-TARS-7B-DPO --served-model-name llm --trust-remote-code error is: |
|
Arf, happy to fix cc @hmellor ! |
|
vLLM 0.8.2 uses Transformers 4.48.2, this PR is not present in that release. Your command doesn't indicate that you're using a custom image, so I'm not sure how you're seeing an error caused by this PR @ErykCh? |
* Fix qwen2_vl and qwen2_5_vl processors cutom images kwargs * change version warning
* Fix qwen2_vl and qwen2_5_vl processors cutom images kwargs * change version warning
7 participants
What does this PR do?
Add support for using custom images kwargs in Qwen2VL and Qwen2.5VL processors
Fix issue where
self.sizedict is mutated by a processor call in qwen2vl image processorsAdd tests for both these issues
Cc @hmellor @ArthurZucker