Split benchmark workflows to publish metrics on postsubmit

.github/workflows/nightly_benchmarks.yml

@@ -0,0 +1,32 @@
+# Copyright 2026 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+name: Nightly Benchmarks
+
+on:
+  schedule:
+    - cron: "0 0 * * *"  # Runs at 00:00 UTC every day
+
+permissions:
+  contents: read
+  pull-requests: write
+
+jobs: 
+  run_benchmarks:
+    name: Run Single Host Benchmarks
+    uses: google-ml-infra/bap/.github/workflows/run-benchmarks.yaml@fd37d1f65a06f0124d9d2022d97efa8dc8812557 # main
+    with:
+      registry_file: "tpu_sync/benchmarks/benchmark_registry.pbtxt"
+      bap_ref: "fd37d1f65a06f0124d9d2022d97efa8dc8812557"
+      runner: "linux-x86-n2-8"

.github/workflows/postsubmit_benchmarks.yml

@@ -0,0 +1,66 @@
+# Copyright 2026 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+name: Postsubmit Benchmarks
+
+on:
+  # Required for internal services to invoke workflow runs.
+  # Required inputs: benchmark_filter, environment_filter, custom_metadata.
+  workflow_dispatch:
+    inputs:
+      benchmark_filter:
+        description: "Regex to filter by benchmark name (e.g., 'h2d_d2h_gating')."
+        required: false
+        type: string
+        default: ""
+      environment_filter:
+        description: "Regex to filter by environment configuration ID (e.g., 'tpu-v5e-single-node')."
+        required: false
+        type: string
+        default: ""
+      custom_metadata:
+        description: "JSON string of metadata to append to the benchmark result. Overwrites static metadata from the registry but is overwritten by dynamic metadata from the workload."
+        required: false
+        type: string
+        default: "{}"
+
+  push:
+    branches:
+      - main
+
+permissions:
+  contents: read
+  pull-requests: write
+
+concurrency:
+  # Group by workflow name and branch to ensure runs execute sequentially.
+  # This prevents parallel postsubmit runs from clashing and ensures accurate 
+  # reporting history.
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: false
+
+jobs: 
+  run_benchmarks:
+    name: Run Single Host Benchmarks
+    uses: google-ml-infra/bap/.github/workflows/run-benchmarks.yaml@fd37d1f65a06f0124d9d2022d97efa8dc8812557 # main
+    with:
+      registry_file: "tpu_sync/benchmarks/benchmark_registry.pbtxt"
+      bap_ref: "fd37d1f65a06f0124d9d2022d97efa8dc8812557"
+      runner: "linux-x86-n2-8"
+      publish_metrics: true
+      pub_sub_gcp_project_id: "ml-oss-benchmarking-production"
+      pub_sub_gcp_topic_id: "public-results-prod"
+      benchmark_filter: ${{ inputs.benchmark_filter || '' }}
+      environment_filter: ${{ inputs.environment_filter || '' }}
+      custom_metadata: ${{ inputs.custom_metadata || '{}' }}

.github/workflows/presubmit_benchmarks.yml

@@ -0,0 +1,31 @@
+# Copyright 2026 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+name: Presubmit Benchmarks
+
+on:
+  pull_request:
+
+permissions:
+  contents: read
+  pull-requests: write
+
+jobs: 
+  run_benchmarks:
+    name: Run Single Host Benchmarks
+    uses: google-ml-infra/bap/.github/workflows/run-benchmarks.yaml@fd37d1f65a06f0124d9d2022d97efa8dc8812557 # main
+    with:
+      registry_file: "tpu_sync/benchmarks/benchmark_registry.pbtxt"
+      bap_ref: "fd37d1f65a06f0124d9d2022d97efa8dc8812557"
+      runner: "linux-x86-n2-8"

.github/workflows/run_benchmarks.yml

@@ -15,15 +15,39 @@
 name: Run Benchmarks
 
 on:
+  # Allow manual runs from the Actions tab for testing
   workflow_dispatch:
-
-  push:
-    branches:
-      - main
-
-  pull_request:
-  schedule:
-    - cron: "0 0 * * *"  # Runs at 00:00 UTC every day
+    inputs:
+      ab_mode:
+        description: "Run A/B comparison (baseline vs experiment)."
+        required: false
+        type: boolean
+        default: false
+      baseline_ref:
+        description: "Git ref for the baseline in A/B mode. Defaults to PR base or main."
+        required: false
+        type: string
+        default: ""
+      experiment_ref:
+        description: "Git ref for the experiment in A/B mode. Defaults to current commit SHA."
+        required: false
+        type: string
+        default: ""
+      benchmark_filter:
+        description: "Regex to filter by benchmark name (e.g., 'h2d_d2h_gating')."
+        required: false
+        type: string
+        default: ""
+      environment_filter:
+        description: "Regex to filter by environment configuration ID (e.g., 'tpu-v5e-single-node')."
+        required: false
+        type: string
+        default: ""
+      custom_metadata:
+        description: "JSON string of metadata to append to the benchmark result. Overwrites static metadata from the registry but is overwritten by dynamic metadata from the workload."
+        required: false
+        type: string
+        default: "{}"
 
 permissions:
   contents: read
@@ -32,11 +56,14 @@ permissions:
 jobs: 
   run_benchmarks:
     name: Run Single Host Benchmarks
-    uses: google-ml-infra/bap/.github/workflows/run-benchmarks.yaml@affd28b492f67c242afa8b119e590b25816b7b88
+    uses: google-ml-infra/bap/.github/workflows/run-benchmarks.yaml@fd37d1f65a06f0124d9d2022d97efa8dc8812557 # main
     with:
       registry_file: "tpu_sync/benchmarks/benchmark_registry.pbtxt"
-      ml_actions_ref: "affd28b492f67c242afa8b119e590b25816b7b88"
+      bap_ref: "fd37d1f65a06f0124d9d2022d97efa8dc8812557"
       runner: "linux-x86-n2-8"
-      publish_metrics: false
-      pub_sub_gcp_project_id: "ml-oss-benchmarking-production"
-      pub_sub_gcp_topic_id: "public-results-prod"
+      ab_mode: ${{ inputs.ab_mode }}
+      baseline_ref: ${{ inputs.baseline_ref }}
+      experiment_ref: ${{ inputs.experiment_ref }}
+      benchmark_filter: ${{ inputs.benchmark_filter }}
+      environment_filter: ${{ inputs.environment_filter }}
+      custom_metadata: ${{ inputs.custom_metadata }}

.github/workflows/update_d2h_h2d_record.yml

@@ -1,16 +1,16 @@
 name: Record H2D/D2H Baselines
+
 on:
   workflow_dispatch:
 
-permissions: { contents: read, pull-requests: write }
+permissions:
+  contents: read
+  pull-requests: write
 
 jobs:
   record:
-    uses: google-ml-infra/bap/.github/workflows/run-benchmarks.yaml@affd28b492f67c242afa8b119e590b25816b7b88
+    uses: google-ml-infra/bap/.github/workflows/run-benchmarks.yaml@fd37d1f65a06f0124d9d2022d97efa8dc8812557 # main
     with:
       registry_file: "tpu_sync/benchmarks/benchmark_registry_record.pbtxt"
-      ml_actions_ref: "affd28b492f67c242afa8b119e590b25816b7b88"
+      ml_actions_ref: "fd37d1f65a06f0124d9d2022d97efa8dc8812557"
       runner: "linux-x86-n2-32"
-      publish_metrics: false
-      pub_sub_gcp_project_id: "ml-oss-benchmarking-production"
-      pub_sub_gcp_topic_id: "public-results-prod"