Skip to content
Closed
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: 1 addition & 8 deletions .github/workflows/run-eval.yml
Original file line number Diff line number Diff line change
Expand Up @@ -61,11 +61,6 @@ on:
required: false
default: ''
type: string
push_to_index:
description: 'Push results to openhands-index-results (default: false)'
required: false
default: 'false'
type: string
enable_conversation_event_logging:
description: 'Enable Datadog persistence for conversation events (default: false)'
required: false
Expand Down Expand Up @@ -100,7 +95,6 @@ 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 }}
PUSH_TO_INDEX: ${{ github.event.inputs.push_to_index }}
ENABLE_CONVERSATION_EVENT_LOGGING: ${{ github.event.inputs.enable_conversation_event_logging }}
MAX_RETRIES: ${{ github.event.inputs.max_retries }}
run: |
Expand All @@ -122,10 +116,9 @@ jobs:
--arg instance_ids "$INSTANCE_IDS" \
--arg num_infer_workers "$NUM_INFER_WORKERS" \
--arg num_eval_workers "$NUM_EVAL_WORKERS" \
--arg push_to_index "$PUSH_TO_INDEX" \
--arg enable_conversation_event_logging "$ENABLE_CONVERSATION_EVENT_LOGGING" \
--arg max_retries "$MAX_RETRIES" \
'{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, push_to_index: $push_to_index, enable_conversation_event_logging: $enable_conversation_event_logging, max_retries: $max_retries}}')
'{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, enable_conversation_event_logging: $enable_conversation_event_logging, max_retries: $max_retries}}')

RESPONSE=$(curl -sS -o /tmp/dispatch.out -w "%{http_code}" -X POST \
-H "Authorization: token $PAT_TOKEN" \
Expand Down