fix Run Harness job

.github/workflows/llm-harness-evaluation.yml

@@ -167,8 +167,7 @@ jobs:
           fi
       
 
-      - name: Run harness nightly
+      - name: Run harness
-        if: ${{github.event_name == 'schedule'}}
         shell: bash
         working-directory: ${{ github.workspace }}/python/llm/dev/benchmark/harness
         env:
@@ -180,6 +179,11 @@ jobs:
           export HF_DATASETS_CACHE=$HARNESS_HF_HOME/datasets
           source /opt/intel/oneapi/setvars.sh
           
+          # set --limit if it's pr-triggered to accelerate pr action
+          if ${{github.event_name == 'pull_request'}}; then
+            export LIMIT="--limit 4"
+          fi
+
           python run_llb.py \
             --model bigdl-llm \
             --pretrained ${MODEL_PATH} \
@@ -187,28 +191,7 @@ jobs:
             --device ${{ matrix.device }} \
             --tasks ${{ matrix.task }} \
             --batch_size 1 --no_cache --output_path results \
-      
+            $LIMIT
-      - name: Run harness pr
-        if: ${{github.event_name == 'pull_request'}}
-        shell: bash
-        working-directory: ${{ github.workspace }}/python/llm/dev/benchmark/harness
-        env:
-          USE_XETLA: OFF
-          # SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS: 1
-        run: |
-          export HF_HOME=${HARNESS_HF_HOME}
-          export HF_DATASETS=$HARNESS_HF_HOME/datasets
-          export HF_DATASETS_CACHE=$HARNESS_HF_HOME/datasets
-          source /opt/intel/oneapi/setvars.sh
-          
-          python run_llb.py \
-            --model bigdl-llm \
-            --pretrained ${MODEL_PATH} \
-            --precision ${{ matrix.precision }} \
-            --device ${{ matrix.device }} \
-            --tasks ${{ matrix.task }} \
-            --batch_size 1 --no_cache --output_path results \
-            --limit 3 \
 
       - uses: actions/upload-artifact@v3
         with: