-
Notifications
You must be signed in to change notification settings - Fork 27.4k
Issue of "failed with initial frozen solve" #64906
Description
🐛 Bug
On Windows
conda install -yq pytorch=1.10.0.dev20210912 cudatoolkit=10.2 -c pytorch-nightly failed due to
Solving environment: ...working... failed with initial frozen solve. Retrying with flexible solve.
https://app.circleci.com/pipelines/github/pytorch/vision/10307/workflows/7b034fbd-1dcf-4f3e-b878-5a62d6e966f5/jobs/781089?invite=true#step-105-513
but 1.10.0.dev20210911 could be installed correctly
conda install -yq pytorch=1.10.0.dev20210911 cudatoolkit=10.2 -c pytorch-nightly works
I tried running conda install pytorch cudatoolkit=10.2 -c pytorch-nightly via ssh
the installed version is 1.10.dev20210911.
The conda version is 4.11 in vision Circle windows image.
Locally, conda install -yq pytorch=1.10.0.dev20210912 cudatoolkit=10.2 -c pytorch-nightly works while the conda version is 4.6
On Linux
1.10.0.dev20210912 works
PackageNotFound for 1.10.0.dev20210911
https://app.circleci.com/pipelines/github/pytorch/vision/10313/workflows/57af7785-6292-4d0e-b824-9b510f8ec962/jobs/781210?invite=true#step-104-355
It blocks vision CI now.
Is there anything wrong in pytorch-nightly channel?