Fix rosetta pax result tables and cleaned up some dead code paths #496
Conversation
This file contains 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
|
Placed PR into draft until CI passes |
ashors1
previously approved these changes
Jan 25, 2024
|
presubmit timedout (6hr exceeded). Rerunning manually: https://github.com/NVIDIA/JAX-Toolbox/actions/runs/7687728046 |
hemildesai
previously approved these changes
Jan 28, 2024
|
The tables appear in different sections, but they are correctly rendered now. I believe we should merge this now since we can at least view failures correctly now. I'll follow up with a fix to the tables appearing in different sections. |
ashors1
approved these changes
Jan 29, 2024
hemildesai
approved these changes
Jan 29, 2024
terrykong
added a commit
that referenced
this pull request
Jan 31, 2024
…of metrics (#503) Follow up of #496 ## changes: 1. upstream MGMN metrics for pax and t5x were being combined so that if t5x finishes first, then paxml will print both t5x and paxml metrics. This is fixed by changing the artifact prefix so that t5x only prints t5x metrics and pax only prints pax metrics. 2. Updates the metric printing to the same as #496 to show everything in a markdown table instead of a bunch of json files which was difficult to visually inspect
ashors1
pushed a commit
that referenced
this pull request
Feb 3, 2024
This PR fixes some issues in reporting that were introduced in #440 . In particular, with the introduction of a non-empty ARTIFACT_NAME in the rosetta tests, the tables no longer were [presented correctly](https://github.com/NVIDIA/JAX-Toolbox/actions/runs/7638798396/attempts/1#summary-20820699653):  I couldn't find anywhere that used the non-default value of ARTIFACT_NAME, so I removed it in the `_test_*_rosetta.yaml` files. This change also cleans up some dead code (EXTRA_TEST_ARGS, EXTRA_GIN_ARGS) since they didn't appear to be used anywhere. Also the table for rosetta-pax metrics were quite long so I rewrote that to output as a markdown table.
ashors1
pushed a commit
that referenced
this pull request
Feb 3, 2024
…of metrics (#503) Follow up of #496 ## changes: 1. upstream MGMN metrics for pax and t5x were being combined so that if t5x finishes first, then paxml will print both t5x and paxml metrics. This is fixed by changing the artifact prefix so that t5x only prints t5x metrics and pax only prints pax metrics. 2. Updates the metric printing to the same as #496 to show everything in a markdown table instead of a bunch of json files which was difficult to visually inspect
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.
This PR fixes some issues in reporting that were introduced in #440 . In particular, with the introduction of a non-empty ARTIFACT_NAME in the rosetta tests, the tables no longer were presented correctly:
I couldn't find anywhere that used the non-default value of ARTIFACT_NAME, so I removed it in the
_test_*_rosetta.yamlfiles.This change also cleans up some dead code (EXTRA_TEST_ARGS, EXTRA_GIN_ARGS) since they didn't appear to be used anywhere.
Also the table for rosetta-pax metrics were quite long so I rewrote that to output as a markdown table.