Skip to content

Conversation

@ShadenSmith
Copy link
Contributor

This is a near rewrite of our testing configuration. We now use conda to matrix tests against various PyTorch and CUDA versions. The new configuration: runs unit tests with:

  1. python=3.6, torch=1.2, cudatoolkit=10.0
  2. python=3.7, torch=1.5, cudatoolkit=10.1

The more expensive model tests are only run on the second configuration. We can next extend our configuration to do nightly builds and include torch=1.3 and 1.4 and to also run model tests for all configurations. We should also include cudatoolkit=10.2 soon.

@ShadenSmith ShadenSmith added the enhancement New feature or request label May 7, 2020
@ShadenSmith ShadenSmith marked this pull request as draft May 7, 2020 15:43
@ShadenSmith ShadenSmith marked this pull request as ready for review May 12, 2020 04:38
@ShadenSmith ShadenSmith requested a review from jeffra May 12, 2020 04:38
@ShadenSmith ShadenSmith merged commit c7863d9 into deepspeedai:master May 12, 2020
@ShadenSmith ShadenSmith deleted the test_matrix branch May 12, 2020 23:17
ShadenSmith added a commit that referenced this pull request May 16, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants