PseCo

PseCo copied to clipboard

cls_head 计算特征时维度不匹配？

3

https://github.com/Hzzone/PseCo/blob/0192ee521cd3786ed5f2d89231fd74567ef87d9c/fsc147/4_1_train_roi_head.py#L277 我在跑这行代码的时候总是会遇到embeddings和prompts特征维度不匹配的问题，导致 https://github.com/Hzzone/PseCo/blob/0192ee521cd3786ed5f2d89231fd74567ef87d9c/models.py#L25 这里无法直接矩阵乘 比如： - 计算 https://github.com/Hzzone/PseCo/blob/0192ee521cd3786ed5f2d89231fd74567ef87d9c/fsc147/4_1_train_roi_head.py#L277 时，embeddings 和 prompts.unsqueeze(1) 的shape 分别是 torch.Size([8192, 3, 512]) torch.Size([8192, 1, 3, 512]) ，这样直接计算会显存爆炸，需要384G ** 我试过把prompts.unsqueeze(1) 去掉unsqueeze(1) ，这样不报显存OOM的bug了，但是下面的代码会报错： - 计算 https://github.com/Hzzone/PseCo/blob/0192ee521cd3786ed5f2d89231fd74567ef87d9c/fsc147/4_1_train_roi_head.py#L283 时，embeddings 和...

lichen14

PointDecoder

1

Hello, I noticed that in your paper, it is mentioned that the point decoder is based on HRNet. However, looking at the code, the structure of PointDecoder combines a Transformer,...

BlouseDong

pth file is

2

ld-xy

是否支持batch推理?

1

您好！请问您的工作是否支持batch推理？

longrunling

Inference speed

1

First of all, I would like to thank you for your great work on this project. I want to understand the average inference time per image for your entire pipeline...

jerpelhan