feat(cli): add geniex eval - compare models on eval packs - #1260
Open
MAN$I VERMA (mansiverma897993) wants to merge 1 commit into
Open
feat(cli): add geniex eval - compare models on eval packs#1260MAN$I VERMA (mansiverma897993) wants to merge 1 commit into
MAN$I VERMA (mansiverma897993) wants to merge 1 commit into
Conversation
1 task
MAN$I VERMA (mansiverma897993)
force-pushed
the
feat/eval-harness
branch
from
July 31, 2026 11:01
8b103ce to
d9b061b
Compare
Contributor
Author
|
Test report (go1.26.5, windows/amd64 - logic tests; no Snapdragon hardware on my side) 5 tests, all passing: pack parsing (valid + 9 rejection cases), pack resolution (file > built-in > error), prompt rendering, scoring (letter extraction, stripping, choice-text match, normalization), reports (accuracy/tok-s math, table, JSON round-trip). Not covered here: |
Run one or more models over an eval pack (built-in or custom JSON; task types multiple_choice / exact_match / contains) and compare accuracy, per-category breakdown, and decode tok/s, with optional JSON export. Greedy decoding, a fixed seed, thinking disabled, and a context reset per task keep runs repeatable. LLM models only. Closes qualcomm#1070 Signed-off-by: mansiverma897993 <vmansi756@gmail.com>
MAN$I VERMA (mansiverma897993)
force-pushed
the
feat/eval-harness
branch
from
July 31, 2026 11:25
d9b061b to
f517120
Compare
Contributor
|
Alan Zhu (@alanzhuly) Could you take a look? |
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #1070.
Adds
geniex eval: run one or more models over an eval pack and compare accuracy.Custom packs are JSON (task types:
multiple_choice,exact_match,contains); one built-in smoke pack ships embedded. Runs are repeatable: greedy decoding, fixed seed, thinking off, context reset per task. LLM only for now; wiring reuses the infer plumbing.Alan Zhu (@alanzhuly) happy to rework schema/scoring/UX to fit your direction. Test report in the comment below.