Update benchmarks.yml

Author: juliagmt-google

.github/workflows/benchmarks.yml

@@ -59,17 +59,9 @@ jobs:
         with:
           halt-dispatch-input: ${{ inputs.halt-for-connection }}
 
-
-      - name: Build hlo_runner_main_gpu
-        working-directory: xla
-        run: bazel build -c opt --config=cuda --dynamic_mode=off //xla/tools/multihost_hlo_runner:hlo_runner_main_gpu
       - name: Build hlo_runner_main
         working-directory: xla
         run: bazel build -c opt --config=cuda --dynamic_mode=off //xla/tools/multihost_hlo_runner:hlo_runner_main
-
-      - name: Build test_gpu_profiler
-        working-directory: xla
-        run: bazel build -c opt --config=cuda --dynamic_mode=off //xla/tools/multihost_hlo_runner:test_gpu_profiler
 
       # - name: Wait For Connection
       #   uses: google-ml-infra/actions/ci_connection@main
@@ -88,44 +80,23 @@ jobs:
           }
           EOF
 
-      - name: Wait For Connection
-        uses: google-ml-infra/actions/ci_connection@main
-        with:
-          halt-dispatch-input: ${{ inputs.halt-for-connection }}
-
-      # - name: Run specific HLO file
-      #   working-directory: xla
-      #   run: |
-      #     nvidia-smi --query-gpu=utilization.gpu --format=csv -l 1 > gpu_utilization.log & bazel run -c opt --config=cuda --dynamic_mode=off //xla/tools/multihost_hlo_runner:hlo_runner_main_gpu -- --device_type=gpu --log_output=True --use_spmd_partitioning isolated_convolution.hlo &> results/isolated_convolution.log
-      # - name: Wait For Connection
-      #   uses: google-ml-infra/actions/ci_connection@main
-      #   with:
-      #     halt-dispatch-input: ${{ inputs.halt-for-connection }}
-      - name: Run specific HLO file hlo_runner_main_gpu
-        working-directory: xla
-        run: |
-           nvidia-smi --query-gpu=utilization.gpu --format=csv -l 1 > results.gpu_utilization.log & bazel run -c opt --config=cuda --dynamic_mode=off //xla/tools/multihost_hlo_runner:hlo_runner_main_gpu -- --device_type=gpu --log_output=True --use_spmd_partitioning isolated_convolution.hlo &> results/hlo_runner_main_gpu_isolated_convolution.log
-
       - name: Run specific HLO file hlo_runner_main
         working-directory: xla
         run: |
-           nvidia-smi --query-gpu=utilization.gpu --format=csv -l 1 > results/gpu_utilization_v2.log & bazel run -c opt --config=cuda --dynamic_mode=off //xla/tools/multihost_hlo_runner:hlo_runner_main -- --device_type=gpu --log_output=True --use_spmd_partitioning isolated_convolution.hlo &> results/hlo_runner_main_isolated_convolution.log
+           nvidia-smi --query-gpu=utilization.gpu --format=csv -l 1 > results/gpu_utilization_v2.log & ./bazel-bin/xla/tools/multihost_hlo_runner/hlo_runner_main --device_type=gpu --use_spmd_partitioning isolated_convolution.hlo &> results/hlo_runner_main_isolated_convolution.log
 
-      - name: Run test_gpu_profiler
-        working-directory: xla
-        run: |
-          nvidia-smi --query-gpu=utilization.gpu --format=csv -l 1 > results/gpu_utilization_profiler.log & ./bazel-bin/xla/tools/multihost_hlo_runner/test_gpu_profiler &> results/test_gpu_profiler.log
       - name: Wait For Connection
         uses: google-ml-infra/actions/ci_connection@main
         with:
           halt-dispatch-input: ${{ inputs.halt-for-connection }}
+
       - name: Download parse_xla_logs.py
         working-directory: xla
         run: wget https://raw.githubusercontent.com/juliagmt-google/xla/main/.github/workflows/parse_xla_logs.py  
 
       - name: Parse XLA logs
         working-directory: xla
-        run: python parse_xla_logs.py results/isolated_convolution.hlo.log
+        run: python parse_xla_logs.py results/hlo_runner_main_isolated_convolution.log
 
       - name: Upload Results
         uses: actions/upload-artifact@v4