In llm-harness-evaluation, add new models and change schedule to nightly (#10072)

* add new models and change schedule to nightly

* correct syntax error

* modify env set up and job

* change label and schedule time

* change schedule time

* change label

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

@@ -8,7 +8,7 @@ concurrency:
 # Controls when the action will run.
 on:
   schedule:
-    - cron: "00 13 * * 5" # GMT time, 13:00 GMT == 21:00 China
+    - cron: "00 16 * * *" # GMT time, 16:00 GMT == 00:00 China
   pull_request:
     branches: [main]
     paths:
@@ -39,6 +39,7 @@ on:
 jobs:
   llm-cpp-build:
     uses: ./.github/workflows/llm-binary-build.yml
+  # Set the testing matrix based on the event (schedule, PR, or manual dispatch)
   set-matrix:
     runs-on: ubuntu-latest
     outputs:
@@ -50,10 +51,11 @@ jobs:
       - name: set-nightly-env
         if: ${{github.event_name == 'schedule'}}
         env:
-          NIGHTLY_MATRIX_MODEL_NAME: '["stablelm-3b-4e1t","Mistral-7B-v0.1"]'
-          NIGHTLY_MATRIX_TASK: '["truthfulqa", "arc"]'
-          NIGHTLY_MATRIX_PRECISION: '["mixed_fp4", "fp8"]'
-          NIGHTLY_LABELS: '["self-hosted", "llm", "accuracy"]'
+          NIGHTLY_MATRIX_MODEL_NAME: '["Llama2-7b-guanaco-dolphin-500", "falcon-7b-instruct-with-patch", 
+                        "Mistral-7B-v0.1", "mpt-7b-chat", "Baichuan2-7B-Chat-LLaMAfied", "stablelm-3b"]'
+          NIGHTLY_MATRIX_TASK: '["arc", "truthfulqa", "winogrande"]'
+          NIGHTLY_MATRIX_PRECISION: '["sym_int4", "fp8"]'
+          NIGHTLY_LABELS: '["self-hosted", "llm", "accuracy-nightly"]'
         run: |
             echo "model_name=$NIGHTLY_MATRIX_MODEL_NAME" >> $GITHUB_ENV
             echo "precision=$NIGHTLY_MATRIX_PRECISION" >> $GITHUB_ENV
@@ -63,9 +65,9 @@ jobs:
       - name: set-pr-env
         if: ${{github.event_name == 'pull_request'}}
         env:
-          PR_MATRIX_MODEL_NAME: '["stablelm-3b-4e1t"]'
+          PR_MATRIX_MODEL_NAME: '["stablelm-3b-4e1t", "Mistral-7B-v0.1"]'
           PR_MATRIX_TASK: '["truthfulqa"]'
-          PR_MATRIX_PRECISION: '["mixed_fp4", "fp8"]'
+          PR_MATRIX_PRECISION: '["fp8"]'
           PR_LABELS: '["self-hosted", "llm", "temp-arc01"]'
         run: |
             echo "model_name=$PR_MATRIX_MODEL_NAME" >> $GITHUB_ENV
@@ -122,6 +124,7 @@ jobs:
       - name: Install dependencies
         shell: bash
         run: |
+          set -e
           python -m pip install --upgrade pip
           python -m pip install --upgrade setuptools==58.0.4
           python -m pip install --upgrade wheel
@@ -132,7 +135,7 @@ jobs:
       - name: Run LLM install (all) test
         uses: ./.github/actions/llm/setup-llm-env
         with:
-          extra-dependency: "xpu_2.0"
+          extra-dependency: "xpu_2.1"
 
       - name: Install harness
         working-directory: ${{ github.workspace }}/python/llm/dev/benchmark/harness/
@@ -154,7 +157,13 @@ jobs:
       - name: Upgrade packages
         shell: bash
         run: |
-          pip install --upgrade transformers==4.34.0 datasets==2.14.6
+          pip install --upgrade datasets==2.14.6 
+          if [ "${{ matrix.model_name }}" = "Mistral-7B-v0.1" ]; then
+          pip install --upgrade transformers==4.36
+          else
+          pip install --upgrade transformers==4.31
+          fi
+      
 
       - name: Run harness
         shell: bash
@@ -166,7 +175,7 @@ jobs:
           export HF_HOME=${HARNESS_HF_HOME}
           export HF_DATASETS=$HARNESS_HF_HOME/datasets
           export HF_DATASETS_CACHE=$HARNESS_HF_HOME/datasets
-          source $HOME/intel/oneapi/setvars.sh
+          source /opt/intel/oneapi/setvars.sh
           python run_llb.py \
           --model bigdl-llm \
           --pretrained ${MODEL_PATH} \