Repository navigation
feat(evaluation): add topology filter chips on Recipe Comparison - #11
Merged
Merged
Conversation
Mirrors the existing 'Filter by knob' chip group with a 'Filter by topology' row. Lets the user isolate a single deployment shape (e.g. 1× (TP=1) vs 2× (TP=2)) and read the table without mixing rows from different hardware layouts. Chips are derived from rows: one per distinct r.topology string, sorted by total GPU count ascending. Only renders when more than one topology is present (matching the knob-filter chip's auto-hide pattern). Filter chain in `rows`: topology AND (category | baseline). Baseline rows still show through any non-"all" category filter to provide a reference variant. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Adds a "Filter by topology" chip row to /evaluation, alongside the existing "Filter by knob" row:
```
Filter by knob: [All] [Spec decoding] [Batch size] [Baseline only]
Filter by topology: [All] [1× (TP=1)] [2× (TP=2)]
```
Lets the user isolate a single deployment shape without scanning the Topology column.
Chips are auto-derived from rows (one per distinct `r.topology` string), sorted by total GPU count ascending. The row renders only when there's more than one topology in the data (matches the knob-filter's auto-hide pattern).
Filter chain: topology AND (category | baseline). Baseline rows still show through any non-"all" category filter as a reference variant.
Test plan
🤖 Generated with Claude Code