Enable integration tests with the latest PyTorch by installing csprng from the source #407
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
… the broken release)
scripts/pytorch_install.sh
Outdated
| else | ||
| echo "No torchcsprng" | ||
| fi | ||
| pip install git+https://github.com/pytorch/csprng.git@cd8f2f670355f5a9b345dbe2cb58a52fb2c44d39#egg=torchcsprng |
Contributor
There was a problem hiding this comment.
Could you add a comment explaining why you chose this specific commit?
| if [[ $PYTORCH_NIGHTLY == true ]]; then | ||
| if [[ $CUDA == true ]]; then | ||
| pip install --upgrade --pre torch torchvision torchcsprng -f https://download.pytorch.org/whl/nightly/cu101/torch_nightly.html | ||
| pip install --upgrade --pre torch torchvision -f https://download.pytorch.org/whl/nightly/cu101/torch_nightly.html |
Contributor
There was a problem hiding this comment.
should this be cu102 since you are changing Cuda version to 10.2?
| # torchcsprng | ||
| if [ "$TORCH_VERSION" = "1.8.0" ] | ||
| then | ||
| pip install torchcsprng==${TORCHCSPRNG_VERSION} |
Contributor
There was a problem hiding this comment.
I suppose this means L22 is no longer required?
| pip install torch==$TORCH_VERSION+cpu torchvision==$TORCHVISION_VERSION+cpu -f https://download.pytorch.org/whl/torch_stable.html | ||
|
|
||
| If ($TORCH_VERSION -eq "1.8.0") { | ||
| pip install torchcsprng==$TORCHCSPRNG_VERSION+cpu -f https://download.pytorch.org/whl/torch_stable.html |
Contributor
There was a problem hiding this comment.
L19 is now irrelevant.
Contributor
|
What would be the oldest supported pytorch version after this change? |
added 3 commits
April 8, 2022 18:09
Contributor
|
Do we plan to land this or can we close this one and continue in #419 ? |
Contributor
Author
|
Closing this until csprng is fixed |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Problem
Integration tests depends on
csprng. The latest csprng release was made in Mar 2021 and has strict dependency onpytorch==1.8.0while the code actually works fine with the newer versions of pytorch.The release cycle is currently broken and there is no easy way to make a new release with the fix.
Temporary solution
Instead of installing
csprngfrom PyPI (having outdated broken release), for the purpose of integration tests we install it from the github source. Recommend doing the same in FAQ.This removes artificially created dependency on the earlier pytorch version and unblocks new features.
Further work
As agreed with the former csprng developer, the Opacus team is the new owner of
csprngrepository. We are going to update and fix the release cycle. After that csprng can be pip installed without wrong dependencies.