Skip to content

Commit bc0d33a

Browse files
committed
infra: evaluate complete frozen V121 shard set only
1 parent cffde88 commit bc0d33a

1 file changed

Lines changed: 65 additions & 0 deletions

File tree

Lines changed: 65 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,65 @@
1+
name: Trace Ace V121 Eval Only
2+
3+
on:
4+
push:
5+
branches: [agent/v121-cache-batch2]
6+
paths:
7+
- '.github/workflows/trace-ace-v121-eval-only.yml'
8+
workflow_dispatch:
9+
10+
concurrency:
11+
group: trace-ace-v121-eval-only
12+
cancel-in-progress: false
13+
14+
env:
15+
PREPARED_RUN_ID: '32400309220'
16+
EMBEDDING_RUN_ID: '32402681183'
17+
18+
jobs:
19+
evaluate:
20+
runs-on: ubuntu-24.04
21+
timeout-minutes: 20
22+
permissions:
23+
actions: read
24+
contents: read
25+
steps:
26+
- uses: actions/checkout@v4
27+
- uses: actions/setup-python@v5
28+
with:
29+
python-version: '3.12'
30+
cache: pip
31+
- name: Install evaluation dependencies
32+
run: python -m pip install --disable-pip-version-check numpy pandas scipy scikit-learn fastembed==0.8.0
33+
- name: Download exact frozen V121 prepared artifact
34+
uses: actions/download-artifact@v4
35+
with:
36+
name: v121-prepared
37+
path: v121_prepared
38+
repository: heathsanchez/mathgraph
39+
run-id: ${{ env.PREPARED_RUN_ID }}
40+
github-token: ${{ secrets.GITHUB_TOKEN }}
41+
- name: Verify frozen manifest
42+
run: |
43+
test -f v121_prepared/manifest.json
44+
grep -q 'b1612f9fe4558680e468afb2a2452b75c603c244934fe62f7345feee68a61bc1' v121_prepared/manifest.json
45+
- name: Download all 128 frozen embedding shards from original run
46+
env:
47+
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
48+
run: |
49+
mkdir -p v121_embedding_shards
50+
for i in $(seq 0 127); do
51+
gh run download "$EMBEDDING_RUN_ID" -n "v121-embeddings-$i" -D "v121_embedding_shards/$i"
52+
done
53+
test "$(find v121_embedding_shards -name 'v121_embeddings_shard_*.npz' | wc -l)" -eq 128
54+
- name: Evaluate unchanged frozen V121 precommit
55+
run: |
56+
cd competitions/trace_the_ace
57+
python v121_staged_transport.py evaluate --dir ../../v121_prepared \
58+
--embeddings ../../v121_embedding_shards --out ../../v121_pretrained_semantic_residual.json
59+
- name: Show decision
60+
run: cat v121_pretrained_semantic_residual.json
61+
- uses: actions/upload-artifact@v4
62+
with:
63+
name: trace-ace-v121-pretrained-semantic-residual
64+
path: v121_pretrained_semantic_residual.json
65+
retention-days: 14

0 commit comments

Comments
 (0)