test 70 s3 upload

pytorch · May 21, 2024 · 8c488c0 · 8c488c0
1 parent f9a629d
commit 8c488c0
Show file tree

Hide file tree

Showing 3 changed files with 27 additions and 35 deletions.
diff --git a/.github/workflows/benchmark_nightly_lmi.yml b/.github/workflows/benchmark_nightly_lmi.yml
@@ -56,33 +56,25 @@ jobs:
       - name: Clean up weights
         run: |
           rm -rf /home/ubuntu/Llama-2-7b-chat-hf
-      # - name: Save benchmark artifacts
-      #   uses: actions/upload-artifact@v2
-      #   with:
-      #     name: nightly ${{ matrix.hardware }} artifact
-      #     path: /tmp/ts_benchmark
-      # - name: Download benchmark artifacts for auto validation
-      #   uses: dawidd6/action-download-artifact@v2
-      #   with:
-      #     workflow: ${{ github.event.workflow_run.workflow_id }}
-      #     workflow_conclusion: success
-      #     if_no_artifact_found: ignore
-      #     path: /tmp/ts_artifacts
-      #     name: ${{ matrix.hardware }}_benchmark_validation
-      # - name: Validate Benchmark result
-      #   run: python benchmarks/validate_report.py --input-artifacts-dir /tmp/ts_artifacts/${{ matrix.hardware }}_benchmark_validation
-      # - name: Update benchmark artifacts for auto validation
-      #   run: python benchmarks/utils/update_artifacts.py --output /tmp/ts_artifacts/${{ matrix.hardware }}_benchmark_validation
-      # - name: Upload the updated benchmark artifacts for auto validation
-      #   uses: actions/upload-artifact@v2
-      #   with:
-      #     name: ${{ matrix.hardware }}_benchmark_validation
-      #     path: /tmp/ts_artifacts
-      # - name: Open issue on failure
-      #   if: ${{ failure() && github.event_name  == 'schedule' && matrix.hardware == 'cpu' }}
-      #   uses: dacbd/create-issue-action@v1
-      #   with:
-      #     token: ${{ secrets.GITHUB_TOKEN }}
-      #     title: Nightly ${{ matrix.hardware }} benchmark failed
-      #     body:  Commit ${{ github.sha }} daily scheduled [CI run](https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }}) failed, please check why
-      #     assignees: ''
+      - name: Save benchmark artifacts
+        uses: actions/upload-artifact@v2
+        with:
+          name: nightly ${{ matrix.hardware }} artifact
+          path: /tmp/ts_benchmark
+      - name: Download benchmark artifacts for auto validation
+        uses: dawidd6/action-download-artifact@v2
+        with:
+          workflow: ${{ github.event.workflow_run.workflow_id }}
+          workflow_conclusion: success
+          if_no_artifact_found: ignore
+          path: /tmp/ts_artifacts
+          name: ${{ matrix.hardware }}_benchmark_validation
+      - name: Validate Benchmark result
+        run: python benchmarks/validate_report.py --input-artifacts-dir /tmp/ts_artifacts/${{ matrix.hardware }}_benchmark_validation
+      - name: Update benchmark artifacts for auto validation
+        run: python benchmarks/utils/update_artifacts.py --output /tmp/ts_artifacts/${{ matrix.hardware }}_benchmark_validation
+      - name: Upload the updated benchmark artifacts for auto validation
+        uses: actions/upload-artifact@v2
+        with:
+          name: ${{ matrix.hardware }}_benchmark_validation
+          path: /tmp/ts_artifacts
diff --git a/benchmarks/benchmark_config_lmi.yaml b/benchmarks/benchmark_config_lmi.yaml
@@ -9,9 +9,9 @@ ts_version:
 # a list of model configure yaml files defined in benchmarks/models_config
 # or a list of model configure yaml files with full path
 models:
-  - "llama-2-7b.yaml"
+  #- "llama-2-7b.yaml"
   #- "llama-2-13b.yaml"
-  #- "llama-2-70b.yaml"
+  - "llama-2-70b.yaml"
 
 # benchmark on "cpu" or "gpu".
 # "cpu" is set if "hardware" is not specified

diff --git a/benchmarks/models_config/llama-2-70b.yaml b/benchmarks/models_config/llama-2-70b.yaml
@@ -1,15 +1,15 @@
 ---
 llama-2-70b:
-  eager-mode:
+  int4:
     benchmark_backend: "ab"
-    url: https://torchserve.s3.amazonaws.com/mar_files/llama-2/llama-2-70b-int4.mar
+    url: https://torchserve.s3.amazonaws.com/mar_files/llama-2/mar+files/llama-2-70b-int4-tp8.mar
     workers:
       - 1
     batch_delay: 100
     batch_size:
       - 1
     input: "./examples/large_models/gpt_fast/request.json"
-    requests: 100
+    requests: 1000
     concurrency: 1
     backend_profiling: False
     exec_env: "local"