Remove redundant 'Apply workflow_dispatch overrides' step from build workflows

.github/workflows/build-commit0-images.yml

@@ -60,13 +60,16 @@ on:
         default: 'false'
         type: string
 
+# Resolve inputs with fallback defaults so both workflow_dispatch and
+# pull_request_target triggers work without a separate override step.
+# INSTANCE_IDS/SELECT_FILE must be initialized for set -euo pipefail.
 env:
-  DATASET: wentingzhao/commit0_combined
-  SPLIT: test
-  REPO_SPLIT: lite
-  MAX_WORKERS: '4'
-  N_LIMIT: ''
-  INSTANCE_IDS: ''
+  DATASET: ${{ inputs.dataset || 'wentingzhao/commit0_combined' }}
+  SPLIT: ${{ inputs.split || 'test' }}
+  REPO_SPLIT: ${{ inputs.repo-split || 'lite' }}
+  MAX_WORKERS: ${{ inputs.max-workers || '4' }}
+  N_LIMIT: ${{ inputs.n-limit || '' }}
+  INSTANCE_IDS: ${{ inputs.instance-ids || '' }}
   SELECT_FILE: ''
 
 concurrency:
@@ -108,17 +111,6 @@ jobs:
           ref: ${{ steps.checkout-ref.outputs.ref }}
           submodules: recursive
 
-      - name: Apply workflow_dispatch overrides (if any)
-        if: ${{ github.event_name == 'workflow_dispatch' }}
-        run: |
-          if [ -n "${{ inputs.dataset }}" ]; then echo "DATASET=${{ inputs.dataset }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.split }}" ]; then echo "SPLIT=${{ inputs.split }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.repo-split }}" ]; then echo "REPO_SPLIT=${{ inputs.repo-split }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-workers }}" ]; then echo "MAX_WORKERS=${{ inputs.max-workers }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.n-limit }}" ]; then echo "N_LIMIT=${{ inputs.n-limit }}" >> "$GITHUB_ENV"; else echo "N_LIMIT=" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.instance-ids }}" ]; then echo "INSTANCE_IDS=${{ inputs.instance-ids }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.sdk-commit }}" ]; then echo "SDK_COMMIT=${{ inputs.sdk-commit }}" >> "$GITHUB_ENV"; fi
-
       - uses: ./.github/actions/build-select-file
         with:
           instance-ids: ${{ env.INSTANCE_IDS }}

.github/workflows/build-multiswebench-images.yml

@@ -61,19 +61,18 @@ on:
         default: 'false'
         type: string
 
-# Defaults for automatic runs; keep INSTANCE_IDS/SELECT_FILE initialized so set -euo pipefail won't fail on unset vars.
+# Resolve inputs with fallback defaults so both workflow_dispatch and
+# pull_request_target triggers work without a separate override step.
+# INSTANCE_IDS/SELECT_FILE must be initialized for set -euo pipefail.
 env:
-  DATASET: bytedance-research/Multi-SWE-Bench
-  SPLIT: test
-  LANGUAGE: java
-  MAX_WORKERS: '12'
-  MAX_RETRIES: '2'
-  N_LIMIT: '50'
-  INSTANCE_IDS: ''
+  DATASET: ${{ inputs.dataset || 'bytedance-research/Multi-SWE-Bench' }}
+  SPLIT: ${{ inputs.split || 'test' }}
+  LANGUAGE: ${{ inputs.language || 'java' }}
+  MAX_WORKERS: ${{ inputs.max-workers || '12' }}
+  MAX_RETRIES: ${{ inputs.max-retries || '2' }}
+  N_LIMIT: ${{ inputs.n-limit || '50' }}
+  INSTANCE_IDS: ${{ inputs.instance-ids || '' }}
   SELECT_FILE: ''
-  BUILD_BATCH_SIZE: '15'
-  BUILDKIT_PRUNE_KEEP_GB: '60'
-  BUILDKIT_PRUNE_THRESHOLD_PCT: '60'
 
 concurrency:
   group: build-multiswebench-${{ github.ref }}
@@ -121,19 +120,6 @@ jobs:
           ref: ${{ steps.checkout-ref.outputs.ref }}
           submodules: recursive
 
-      # If this was a manual dispatch, override defaults with provided inputs.
-      - name: Apply workflow_dispatch overrides (if any)
-        if: ${{ github.event_name == 'workflow_dispatch' }}
-        run: |
-          if [ -n "${{ inputs.dataset }}" ]; then echo "DATASET=${{ inputs.dataset }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.split }}" ]; then echo "SPLIT=${{ inputs.split }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.language }}" ]; then echo "LANGUAGE=${{ inputs.language }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-workers }}" ]; then echo "MAX_WORKERS=${{ inputs.max-workers }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-retries }}" ]; then echo "MAX_RETRIES=${{ inputs.max-retries }}" >> "$GITHUB_ENV"; fi
-          # Empty string means "no limit"
-          if [ -n "${{ inputs.n-limit }}" ]; then echo "N_LIMIT=${{ inputs.n-limit }}" >> "$GITHUB_ENV"; else echo "N_LIMIT=" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.instance-ids }}" ]; then echo "INSTANCE_IDS=${{ inputs.instance-ids }}" >> "$GITHUB_ENV"; fi
-
       # Set N_LIMIT based on the label that triggered the workflow
       - name: Set N_LIMIT based on label
         if: ${{ github.event_name == 'pull_request_target' }}

.github/workflows/build-swebench-images.yml

@@ -65,14 +65,16 @@ on:
         default: 'default'
         type: string
 
-# Defaults for automatic runs; keep INSTANCE_IDS/SELECT_FILE initialized so set -euo pipefail won't fail on unset vars.
+# Resolve inputs with fallback defaults so both workflow_dispatch and
+# pull_request_target triggers work without a separate override step.
+# INSTANCE_IDS/SELECT_FILE must be initialized for set -euo pipefail.
 env:
-  DATASET: princeton-nlp/SWE-bench_Verified
-  SPLIT: test
-  MAX_WORKERS: '12'
-  MAX_RETRIES: '2'
-  N_LIMIT: '500'
-  INSTANCE_IDS: ''
+  DATASET: ${{ inputs.dataset || 'princeton-nlp/SWE-bench_Verified' }}
+  SPLIT: ${{ inputs.split || 'test' }}
+  MAX_WORKERS: ${{ inputs.max-workers || '12' }}
+  MAX_RETRIES: ${{ inputs.max-retries || '2' }}
+  N_LIMIT: ${{ inputs.n-limit || '500' }}
+  INSTANCE_IDS: ${{ inputs.instance-ids || '' }}
   SELECT_FILE: ''
 
 concurrency:
@@ -121,18 +123,6 @@ jobs:
           ref: ${{ steps.checkout-ref.outputs.ref }}
           submodules: recursive
 
-      # If this was a manual dispatch, override defaults with provided inputs.
-      - name: Apply workflow_dispatch overrides (if any)
-        if: ${{ github.event_name == 'workflow_dispatch' }}
-        run: |
-          if [ -n "${{ inputs.dataset }}" ]; then echo "DATASET=${{ inputs.dataset }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.split }}" ]; then echo "SPLIT=${{ inputs.split }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-workers }}" ]; then echo "MAX_WORKERS=${{ inputs.max-workers }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-retries }}" ]; then echo "MAX_RETRIES=${{ inputs.max-retries }}" >> "$GITHUB_ENV"; fi
-          # Empty string means "no limit"
-          if [ -n "${{ inputs.n-limit }}" ]; then echo "N_LIMIT=${{ inputs.n-limit }}" >> "$GITHUB_ENV"; else echo "N_LIMIT=" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.instance-ids }}" ]; then echo "INSTANCE_IDS=${{ inputs.instance-ids }}" >> "$GITHUB_ENV"; fi
-
       # Set N_LIMIT based on the label that triggered the workflow
       - name: Set N_LIMIT based on label
         if: ${{ github.event_name == 'pull_request_target' }}

.github/workflows/build-swebenchmultimodal-images.yml

@@ -60,14 +60,16 @@ on:
         default: 'false'
         type: string
 
-# Defaults for automatic runs; keep INSTANCE_IDS/SELECT_FILE initialized so set -euo pipefail won't fail on unset vars.
+# Resolve inputs with fallback defaults so both workflow_dispatch and
+# pull_request_target triggers work without a separate override step.
+# INSTANCE_IDS/SELECT_FILE must be initialized for set -euo pipefail.
 env:
-  DATASET: princeton-nlp/SWE-bench_Multimodal
-  SPLIT: dev
-  MAX_WORKERS: '12'
-  MAX_RETRIES: '2'
-  N_LIMIT: '500'
-  INSTANCE_IDS: ''
+  DATASET: ${{ inputs.dataset || 'princeton-nlp/SWE-bench_Multimodal' }}
+  SPLIT: ${{ inputs.split || 'dev' }}
+  MAX_WORKERS: ${{ inputs.max-workers || '12' }}
+  MAX_RETRIES: ${{ inputs.max-retries || '2' }}
+  N_LIMIT: ${{ inputs.n-limit || '500' }}
+  INSTANCE_IDS: ${{ inputs.instance-ids || '' }}
   SELECT_FILE: ''
 
 concurrency:
@@ -116,18 +118,6 @@ jobs:
           ref: ${{ steps.checkout-ref.outputs.ref }}
           submodules: recursive
 
-      # If this was a manual dispatch, override defaults with provided inputs.
-      - name: Apply workflow_dispatch overrides (if any)
-        if: ${{ github.event_name == 'workflow_dispatch' }}
-        run: |
-          if [ -n "${{ inputs.dataset }}" ]; then echo "DATASET=${{ inputs.dataset }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.split }}" ]; then echo "SPLIT=${{ inputs.split }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-workers }}" ]; then echo "MAX_WORKERS=${{ inputs.max-workers }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-retries }}" ]; then echo "MAX_RETRIES=${{ inputs.max-retries }}" >> "$GITHUB_ENV"; fi
-          # Empty string means "no limit"
-          if [ -n "${{ inputs.n-limit }}" ]; then echo "N_LIMIT=${{ inputs.n-limit }}" >> "$GITHUB_ENV"; else echo "N_LIMIT=" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.instance-ids }}" ]; then echo "INSTANCE_IDS=${{ inputs.instance-ids }}" >> "$GITHUB_ENV"; fi
-
       # Set N_LIMIT based on the label that triggered the workflow
       - name: Set N_LIMIT based on label
         if: ${{ github.event_name == 'pull_request_target' }}

.github/workflows/build-swegym-images.yml

@@ -56,18 +56,17 @@ on:
         default: 'false'
         type: string
 
-# Defaults for automatic runs; keep INSTANCE_IDS/SELECT_FILE initialized so set -euo pipefail won't fail on unset vars.
+# Resolve inputs with fallback defaults so both workflow_dispatch and
+# pull_request_target triggers work without a separate override step.
+# INSTANCE_IDS/SELECT_FILE must be initialized for set -euo pipefail.
 env:
-  DATASET: SWE-Gym/SWE-Gym
-  SPLIT: train
-  MAX_WORKERS: '12'
-  MAX_RETRIES: '2'
-  N_LIMIT: '2438'
-  INSTANCE_IDS: ''
+  DATASET: ${{ inputs.dataset || 'SWE-Gym/SWE-Gym' }}
+  SPLIT: ${{ inputs.split || 'train' }}
+  MAX_WORKERS: ${{ inputs.max-workers || '12' }}
+  MAX_RETRIES: ${{ inputs.max-retries || '2' }}
+  N_LIMIT: ${{ inputs.n-limit || '2438' }}
+  INSTANCE_IDS: ${{ inputs.instance-ids || '' }}
   SELECT_FILE: ''
-  BUILD_BATCH_SIZE: '15'
-  BUILDKIT_PRUNE_KEEP_GB: '60'
-  BUILDKIT_PRUNE_THRESHOLD_PCT: '60'
 
 concurrency:
   group: build-swe-gym-${{ github.ref }}
@@ -115,18 +114,6 @@ jobs:
           ref: ${{ steps.checkout-ref.outputs.ref }}
           submodules: recursive
 
-      # If this was a manual dispatch, override defaults with provided inputs.
-      - name: Apply workflow_dispatch overrides (if any)
-        if: ${{ github.event_name == 'workflow_dispatch' }}
-        run: |
-          if [ -n "${{ inputs.dataset }}" ]; then echo "DATASET=${{ inputs.dataset }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.split }}" ]; then echo "SPLIT=${{ inputs.split }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-workers }}" ]; then echo "MAX_WORKERS=${{ inputs.max-workers }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-retries }}" ]; then echo "MAX_RETRIES=${{ inputs.max-retries }}" >> "$GITHUB_ENV"; fi
-          # Empty string means "no limit"
-          if [ -n "${{ inputs.n-limit }}" ]; then echo "N_LIMIT=${{ inputs.n-limit }}" >> "$GITHUB_ENV"; else echo "N_LIMIT=" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.instance-ids }}" ]; then echo "INSTANCE_IDS=${{ inputs.instance-ids }}" >> "$GITHUB_ENV"; fi
-
       # Set N_LIMIT based on the label that triggered the workflow
       - name: Set N_LIMIT based on label
         if: ${{ github.event_name == 'pull_request_target' }}

.github/workflows/build-swesmith-images.yml

@@ -56,18 +56,18 @@ on:
         default: 'false'
         type: string
 
-# Defaults for automatic runs; keep INSTANCE_IDS/SELECT_FILE initialized so set -euo pipefail won't fail on unset vars.
+# Resolve inputs with fallback defaults so both workflow_dispatch and
+# pull_request_target triggers work without a separate override step.
+# INSTANCE_IDS/SELECT_FILE must be initialized for set -euo pipefail.
+# Note: N_LIMIT default of 51000 is effectively "all" since total unique base images <200.
 env:
-  DATASET: SWE-bench/SWE-smith-py
-  SPLIT: train
-  MAX_WORKERS: '12'
-  MAX_RETRIES: '2'
-  N_LIMIT: '51000' # Total number of unique base images in SWE-Smith is <200
-  INSTANCE_IDS: ''
+  DATASET: ${{ inputs.dataset || 'SWE-bench/SWE-smith-py' }}
+  SPLIT: ${{ inputs.split || 'train' }}
+  MAX_WORKERS: ${{ inputs.max-workers || '12' }}
+  MAX_RETRIES: ${{ inputs.max-retries || '2' }}
+  N_LIMIT: ${{ inputs.n-limit || '51000' }}
+  INSTANCE_IDS: ${{ inputs.instance-ids || '' }}
   SELECT_FILE: ''
-  BUILD_BATCH_SIZE: '15'
-  BUILDKIT_PRUNE_KEEP_GB: '60'
-  BUILDKIT_PRUNE_THRESHOLD_PCT: '60'
 
 concurrency:
   group: build-swe-smith-${{ github.ref }}
@@ -115,18 +115,6 @@ jobs:
           ref: ${{ steps.checkout-ref.outputs.ref }}
           submodules: recursive
 
-      # If this was a manual dispatch, override defaults with provided inputs.
-      - name: Apply workflow_dispatch overrides (if any)
-        if: ${{ github.event_name == 'workflow_dispatch' }}
-        run: |
-          if [ -n "${{ inputs.dataset }}" ]; then echo "DATASET=${{ inputs.dataset }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.split }}" ]; then echo "SPLIT=${{ inputs.split }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-workers }}" ]; then echo "MAX_WORKERS=${{ inputs.max-workers }}" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.max-retries }}" ]; then echo "MAX_RETRIES=${{ inputs.max-retries }}" >> "$GITHUB_ENV"; fi
-          # Empty string means "no limit"
-          if [ -n "${{ inputs.n-limit }}" ]; then echo "N_LIMIT=${{ inputs.n-limit }}" >> "$GITHUB_ENV"; else echo "N_LIMIT=" >> "$GITHUB_ENV"; fi
-          if [ -n "${{ inputs.instance-ids }}" ]; then echo "INSTANCE_IDS=${{ inputs.instance-ids }}" >> "$GITHUB_ENV"; fi
-
       # Set N_LIMIT based on the label that triggered the workflow
       - name: Set N_LIMIT based on label
         if: ${{ github.event_name == 'pull_request_target' }}