Pass in runner

Author: ahadnagy

.github/workflows/benchmark_v2.yml

@@ -3,6 +3,15 @@ name: Benchmark v2 Framework
 on:
   workflow_call:
     inputs:
+      runner_group:
+        description: 'GH Actions runner group to use'
+        required: false
+        type: string
+      runner_labels:
+        description: 'GH Actions runner label to use'
+        required: false
+        type: string
+        default: null
       model_id:
         description: 'Model ID to benchmark (e.g., meta-llama/Llama-2-7b-hf)'
         required: false
@@ -54,12 +63,9 @@ env:
 jobs:
   benchmark-v2:
     name: Benchmark v2
-    strategy:
-      matrix:
-        # Use GPU-enabled runners for accurate benchmarking
-        group: [aws-g5-4xlarge-cache]
     runs-on:
-      group: ${{ matrix.group }}
+      group: ${{ inputs.runner_group }}
+      labels: ${{ inputs.runner_labels }}
     container:
       image: huggingface/transformers-pytorch-gpu
       options: --gpus all --privileged --ipc host --shm-size "16gb"

.github/workflows/benchmark_v2_caller.yml

@@ -12,7 +12,6 @@ on:
       model_id:
         description: 'Model ID to benchmark (leave empty for default models)'
         required: false
-        type: string
         default: ''
       warmup_iterations:
         description: 'Number of warmup iterations'
@@ -47,6 +46,7 @@ jobs:
     name: Benchmark v2 - Default Models
     uses: ./.github/workflows/benchmark_v2.yml
     with:
+      runner_group: "aws-g5-4xlarge-cache"
       model_id: ${{ inputs.model_id || '' }}
       warmup_iterations: ${{ inputs.warmup_iterations || 3 }}
       measurement_iterations: ${{ inputs.measurement_iterations || 5 }}