CUDA out of memory for video understanding

#41
by luweigen - opened

Like in https://github.com/QwenLM/Qwen2.5-VL/blob/main/cookbooks/video_understanding.ipynb , it takes almost 1TB memory to inference 30 seconds of 720 video with CPU. With GPU it's always CUDA out of memory.
Anyone has successful record with how many VRAM?

luweigen changed discussion status to closed

Sign up or log in to comment