Skip to content

Conversation

@ervteng
Copy link
Contributor

@ervteng ervteng commented May 12, 2021

Proposed change(s)

In the ResNet encoder, we were grabbing the batch size from the actual observations themselves and using that to reshape the final output of the conv layer. This was not compatible with current versions of the Barracuda importer. Now we just cache it first and use that value.

Useful links (Github issues, JIRA tickets, ML-Agents forum threads etc.)

https://forum.unity.com/threads/sac-training-very-unstable.1104580/#post-7131539

Types of change(s)

  • Bug fix
  • New feature
  • Code refactor
  • Breaking change
  • Documentation update
  • Other (please describe)

Checklist

  • Added tests that prove my fix is effective or that my feature works
  • Updated the changelog (if applicable)
  • Updated the documentation (if applicable)
  • Updated the migration guide (if applicable)

Other comments

@surfnerd surfnerd merged commit 894bffa into main May 13, 2021
@delete-merged-branch delete-merged-branch bot deleted the develop-fix-resnet branch May 13, 2021 18:35
surfnerd added a commit that referenced this pull request May 13, 2021
@github-actions github-actions bot locked as resolved and limited conversation to collaborators May 13, 2022
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants