Chen Weiyan
Chen Weiyan
源代码是不是将所有训练数据加载到内存中,还没加载完,内存直接爆了
I also have the same question for the authors.
Update dill to 0.3.9. It works for me.
It might be that the model uses with torch.no_grad() during the forward pass.
我看了一下 https://github.com/OpenGVLab/InternVL/blob/main/internvl_chat/eval/refcoco/evaluate_grounding.py,里面输出的 bounding_box 会 normalize 1000,然后再乘以原始图像的长宽
@mihirp1998