davidluciolu

icon location Beijing

Results 14 comments of davidluciolu

llava-intern2-7b模型无法复现官方结果

您好，请问推理一遍mmbench dev en需要多久？ @PeterMao11

About removing abstract noun phrases in GRIT construction

Hi! Could you share the ChatGPT prompt? I am trying to use ChatGPT to generate a list, but results are awful. ChatGPT keeps generating repeated words. Many thanks!

About removing abstract noun phrases in GRIT construction

thank you!

【kosmos-2】The code for GRIT construction

Hi! I am also curious about the construction of the GRIT dataset. It is mentioned in the paper that > We eliminate certain abstract noun phrases that are challenging to...

AttributeError: 'str' object has no attribute 'impl'

try to use `pip install xformers==0.0.23.post1`

[Usage]: Qwen2-VL keyword argument `max_pixels` is not a valid argument for this processor and will be ignored.

mark

Serious misalignment in LLaVA implementation

I think mm_projector in not trained under this setting. You need to add a line `additional_target: multi_modal_projector` in your yaml to set multi_modal_projector trainable

QWEN2 VL forward function extremely slow for videos as short as 1.5s

@jbohnslav hello, I'm also facing the memory leak problem, any suggestions for this?

[Bug]: RuntimeError: Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing, you must use the 'spawn' start method

Same problem. - Successfully run with torch==2.5.1 and vllm==0.7.2, without adding `export VLLM_WORKER_MULTIPROC_METHOD=spawn` or removing device - The RuntimeError shows when update vllm to 0.8.1 and required torch==2.6.0. But fixes...

Would VLLM be used to accelerate model inference?

@YushunXiang Hello! Does this vLLM server support multi-card inference?