Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion .github/workflows/run-eval.yml
Original file line number Diff line number Diff line change
Expand Up @@ -60,6 +60,11 @@ on:
required: false
default: ''
type: string
output_jsonl_gcs:
description: Optional GCS path to output.jsonl or results tar.gz to skip inference
required: false
default: ''
type: string

jobs:
dispatch-run-eval:
Expand All @@ -84,6 +89,7 @@ jobs:
INSTANCE_IDS: ${{ github.event.inputs.instance_ids }}
NUM_INFER_WORKERS: ${{ github.event.inputs.num_infer_workers }}
NUM_EVAL_WORKERS: ${{ github.event.inputs.num_eval_workers }}
OUTPUT_JSONL_GCS: ${{ github.event.inputs.output_jsonl_gcs }}
run: |
set -euo pipefail
if [ -z "$PAT_TOKEN" ]; then
Expand All @@ -103,7 +109,8 @@ jobs:
--arg instance_ids "$INSTANCE_IDS" \
--arg num_infer_workers "$NUM_INFER_WORKERS" \
--arg num_eval_workers "$NUM_EVAL_WORKERS" \
'{ref: $ref, inputs: {benchmark: $benchmark, sdk_ref: $sdk_ref, eval_limit: $eval_limit, model_ids: $model_ids, reason: $reason, eval_branch: $eval_branch, benchmarks_branch: $benchmarks_branch, instance_ids: $instance_ids, num_infer_workers: $num_infer_workers, num_eval_workers: $num_eval_workers}}')
--arg output_jsonl_gcs "$OUTPUT_JSONL_GCS" \
'{ref: $ref, inputs: {benchmark: $benchmark, sdk_ref: $sdk_ref, eval_limit: $eval_limit, model_ids: $model_ids, reason: $reason, eval_branch: $eval_branch, benchmarks_branch: $benchmarks_branch, instance_ids: $instance_ids, num_infer_workers: $num_infer_workers, num_eval_workers: $num_eval_workers, output_jsonl_gcs: $output_jsonl_gcs}}')

RESPONSE=$(curl -sS -o /tmp/dispatch.out -w "%{http_code}" -X POST \
-H "Authorization: token $PAT_TOKEN" \
Expand Down
2 changes: 1 addition & 1 deletion pyproject.toml
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@ dependencies = [
"fastmcp>=2.11.3",
"httpx>=0.27.0",
"litellm>=1.77.7.dev9",
"pydantic>=2.11.7",
"pydantic>=2.12.0",
"python-frontmatter>=1.1.0",
"python-json-logger>=3.3.0",
"tenacity>=9.1.2",
Expand Down
Loading