search

mli0603 / stereo-transformer

Revisiting Stereo Depth Estimation From a Sequence-to-Sequence Perspective with Transformers. (ICCV 2021 Oral)
Apache License 2.0
664 stars 106 forks source link

Unable to use `nn.DataParallel()` for multi-GPU training #87

Closed saveleaf closed 7 months ago

saveleaf commented 1 year ago

Hello, thank you very much for your outstanding work, but I have a very puzzling question. When I wanted to try multi-GPU parallel training, I used the following code as usual: model = nn.DataParallel(model) but I found that the runtime didn't find to get the expected results, can you please ask me if there's any particular reason for this?

mli0603 commented 1 year ago

Hi @saveleaf

Thank you for your interest in the project.

When I wrote the code, I only had one GPU and did not configure the code such that it supports parallel training. It may require some code changes to make it work. Feel free to submit a PR if you make it work!

XiandaGuo commented 10 months ago

We have seamlessly integrated the contributions of the talented author, STTR, into OpenStereo(https://github.com/XiandaGuo/OpenStereo), enabling support for multi-GPU training. We invite you to explore and experience it firsthand!