diff --git a/.github/actions/llm/example-test-436/action.yml b/.github/actions/llm/example-test-436/action.yml
new file mode 100644
index 00000000000..2871b87f73c
--- /dev/null
+++ b/.github/actions/llm/example-test-436/action.yml
@@ -0,0 +1,14 @@
+name: 'IPEX-LLM example tests'
+description: 'IPEX-LLM example tests'
+
+runs:
+  using: "composite"
+  steps:
+    - name: Test LLAMA2
+      shell: bash
+      env:
+        INT4_CKPT_DIR: ./llm/ggml-actions/stable
+        LLM_DIR: ./llm
+      
+      run: |
+        bash python/llm/dev/test/run-example-tests-436.sh
\ No newline at end of file
diff --git a/.github/workflows/llm-nightly-test-436.yml b/.github/workflows/llm-nightly-test-436.yml
new file mode 100644
index 00000000000..cf26f93b5c1
--- /dev/null
+++ b/.github/workflows/llm-nightly-test-436.yml
@@ -0,0 +1,113 @@
+name: LLM Nightly Tests for 4.36
+
+# Cancel previous runs in the PR when you push new commits
+concurrency:
+  group: ${{ github.workflow }}-llm-nightly-test-${{ github.event.pull_request.number || github.run_id }}
+  cancel-in-progress: true
+
+permissions:
+  contents: read
+
+# Controls when the action will run.
+on:
+  schedule:
+    - cron: "00 13 * * *" # GMT time, 13:00 GMT == 21:00 China
+  pull_request:
+    branches: [main]
+    paths:
+#      - ".github/workflows/llm-nightly-test.yml"
+#      - ".github/actions/llm/setup-llm-env/action.yml"
+#      - ".github/actions/llm/remove-llm-env/action.yml"
+#      - ".github/actions/llm/convert-test/action.yml"
+  # Allows you to run this workflow manually from the Actions tab
+  workflow_dispatch:
+
+# A workflow run is made up of one or more jobs that can run sequentially or in parallel
+jobs:
+  llm-cpp-build:
+    uses: ./.github/workflows/llm-binary-build.yml
+  llm-nightly-convert-test:
+    needs: llm-cpp-build
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - os: windows
+            instruction: AVX-VNNI-UT
+            python-version: "3.11"
+          - os: ubuntu-20.04-lts
+            instruction: avx512
+            python-version: "3.11"
+    runs-on: [self-hosted, llm, "${{matrix.instruction}}", "${{matrix.os}}"]
+    env:
+      ANALYTICS_ZOO_ROOT: ${{ github.workspace }}
+    steps:
+      - name: Set model directories
+        shell: bash
+        run: |
+          echo "ORIGIN_DIR=$(pwd)/../llm/origin-models" >> "$GITHUB_ENV"
+          echo "INT4_CKPT_DIR=$(pwd)/../llm/nightly-converted-models" >> "$GITHUB_ENV"
+      - name: Create model directories
+        shell: bash
+        run: |
+          if [ ! -d $ORIGIN_DIR ]; then
+            mkdir -p $ORIGIN_DIR
+          fi
+          if [ ! -d $INT4_CKPT_DIR ]; then
+            mkdir -p $INT4_CKPT_DIR
+          fi
+      - name: Set environment variables
+        shell: bash
+        run: |
+          echo "LLAMA_ORIGIN_PATH=${ORIGIN_DIR}/llama-7b-hf" >> "$GITHUB_ENV"
+          echo "GPTNEOX_ORIGIN_PATH=${ORIGIN_DIR}/gptneox-7b-redpajama-bf16" >> "$GITHUB_ENV"
+          echo "BLOOM_ORIGIN_PATH=${ORIGIN_DIR}/bloomz-7b1" >> "$GITHUB_ENV"
+          echo "STARCODER_ORIGIN_PATH=${ORIGIN_DIR}/gpt_bigcode-santacoder" >> "$GITHUB_ENV"
+
+          echo "LLAMA_INT4_CKPT_PATH=${INT4_CKPT_DIR}/bigdl_llm_llama_q4_0.bin" >> "$GITHUB_ENV"
+          echo "GPTNEOX_INT4_CKPT_PATH=${INT4_CKPT_DIR}/bigdl_llm_gptneox_q4_0.bin" >> "$GITHUB_ENV"
+          echo "BLOOM_INT4_CKPT_PATH=${INT4_CKPT_DIR}/bigdl_llm_bloom_q4_0.bin" >> "$GITHUB_ENV"
+          echo "STARCODER_INT4_CKPT_PATH=${INT4_CKPT_DIR}/bigdl_llm_starcoder_q4_0.bin" >> "$GITHUB_ENV"
+      - uses: actions/checkout@f43a0e5ff2bd294095638e18286ca9a3d1956744 # actions/checkout@v3
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v4
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Install dependencies
+        shell: bash
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install --upgrade setuptools==58.0.4
+          python -m pip install --upgrade wheel
+
+      - name: Download llm binary
+        uses: ./.github/actions/llm/download-llm-binary
+
+      - name: Install IPEX-LLM
+        uses: ./.github/actions/llm/setup-llm-env
+
+      - name: Download original models & convert
+        uses: ./.github/actions/llm/convert-test
+
+      - name: Upload ckpt to ftp
+        shell: bash
+        if: runner.os == 'Linux' && github.event_name == 'schedule'
+        run: |
+          curl -T $LLAMA_INT4_CKPT_PATH ${LLM_FTP_URL}/llm/ggml-actions/nightly/bigdl_llm_llama_7b_q4_0.bin
+          curl -T $GPTNEOX_INT4_CKPT_PATH ${LLM_FTP_URL}/llm/ggml-actions/nightly/bigdl_llm_redpajama_7b_q4_0.bin
+          curl -T $BLOOM_INT4_CKPT_PATH ${LLM_FTP_URL}/llm/ggml-actions/nightly/bigdl_llm_bloom_7b_q4_0.bin
+          curl -T $STARCODER_INT4_CKPT_PATH ${LLM_FTP_URL}/llm/ggml-actions/nightly/bigdl_llm_santacoder_1b_q4_0.bin
+      - name: Delete ckpt
+        shell: bash
+        run: |
+          rm -rf $LLAMA_INT4_CKPT_PATH
+          rm -rf $GPTNEOX_INT4_CKPT_PATH
+          rm -rf $BLOOM_INT4_CKPT_PATH
+          rm -rf $STARCODER_INT4_CKPT_PATH
+
+  llm-unit-tests:
+    needs: llm-cpp-build
+    uses: ./.github/workflows/llm_unit_tests_4.36.yml
+  llm-example-test:
+    needs: llm-cpp-build
+    uses: ./.github/workflows/llm_example_tests_436.yml
diff --git a/.github/workflows/llm_example_tests_436.yml b/.github/workflows/llm_example_tests_436.yml
new file mode 100644
index 00000000000..662a9f2c5cb
--- /dev/null
+++ b/.github/workflows/llm_example_tests_436.yml
@@ -0,0 +1,81 @@
+name: LLM Example Test for 4.36
+
+# Cancel previous runs in the PR when you push new commits
+concurrency:
+  group: ${{ github.workflow }}-llm-example-tests-${{ github.event.pull_request.number || github.run_id }}
+  cancel-in-progress: true
+
+permissions:
+  contents: read
+
+# Controls when the action will run. 
+on:
+  # schedule:
+  #   - cron: '00 13 * * *' # GMT time, 13:00 GMT == 21:00 China
+  pull_request:
+    branches: [ main ]
+    paths:
+      - '.github/workflows/llm_example_tests.yml'
+      - '.github/workflows/llm-binary-build.yml'
+      - '.github/actions/llm/example-test/action.yml'
+      - '.github/actions/llm/setup-llm-env/action.yml'
+      - '.github/actions/llm/remove-llm-env/action.yml'
+      - '.github/actions/llm/download-llm-binary/action.yml'
+      - 'python/llm/dev/test/run-example-tests.sh'
+      - 'python/llm/example/**'
+  workflow_dispatch:
+  workflow_call:
+
+env:
+  INT4_CKPT_DIR: ./llm/ggml-actions/stable
+  LLM_DIR: ./llm
+
+# A workflow run is made up of one or more jobs that can run sequentially or in parallel
+jobs:
+  llm-cpp-build:
+    uses: ./.github/workflows/llm-binary-build.yml
+  llm-example-test:
+    needs: llm-cpp-build
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.11"]
+        instruction: ["AVX512"]
+    runs-on: [ self-hosted, llm,"${{matrix.instruction}}", ubuntu-20.04-lts ]
+    env:
+      THREAD_NUM: 24
+    steps:
+      - uses: actions/checkout@ee0669bd1cc54295c223e0bb666b733df41de1c5 # actions/checkout@v2
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v2
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install --upgrade setuptools==58.0.4
+          python -m pip install --upgrade wheel
+
+#      - name: Download llm binary
+#        uses: ./.github/actions/llm/download-llm-binary
+#
+#      - name: Run LLM install (all) test
+#        uses: ./.github/actions/llm/setup-llm-env
+#        env:
+#          ANALYTICS_ZOO_ROOT: ${{ github.workspace }}
+
+      - name: Install IPEX-LLM from Pypi
+        shell: bash
+        run: |
+          pip install --pre --upgrade ipex-llm[all]
+          pip install transformers==4.36.2
+
+      - name: Run LLM example test
+        uses: ./.github/actions/llm/example-test
+        env:
+          ANALYTICS_ZOO_ROOT: ${{ github.workspace }}
+
+      # - name: Clean up test environment
+      #   uses: ./.github/actions/llm/remove-llm-env
+      #   env:
+      #     ANALYTICS_ZOO_ROOT: ${{ github.workspace }}
diff --git a/.github/workflows/llm_performance_tests_436.yml b/.github/workflows/llm_performance_tests_436.yml
new file mode 100644
index 00000000000..ec4294c4578
--- /dev/null
+++ b/.github/workflows/llm_performance_tests_436.yml
@@ -0,0 +1,885 @@
+name: LLM Performance Test for 4.36
+
+# Cancel previous runs in the PR when you push new commits
+concurrency:
+  group: ${{ github.workflow }}-llm-performance-tests-${{ github.event.pull_request.number || github.run_id }}
+  cancel-in-progress: true
+
+permissions:
+  contents: read
+
+# Controls when the action will run.
+on:
+  schedule:
+    - cron: "30 16 * * *" # GMT time, 16:30 GMT == 00:30 China
+  # please uncomment it for PR tests
+   pull_request:
+     branches: [main]
+  #   paths:
+  #     - ".github/workflows/llm_performance_tests.yml"
+  #     - "python/llm/test/benchmark/**"
+  #     - "python/llm/dev/benchmark/all-in-one/**"
+  workflow_dispatch:
+  workflow_call:
+
+# A workflow run is made up of one or more jobs that can run sequentially or in parallel
+jobs:
+  # llm-cpp-build: # please uncomment it for PR tests
+  #   uses: ./.github/workflows/llm-binary-build.yml
+
+  llm-performance-test-on-arc:
+    if: ${{ github.event.schedule || github.event.inputs.artifact == 'llm-performance-test-on-arc' || github.event.inputs.artifact == 'all' }} # please comment it for PR tests
+    # needs: llm-cpp-build # please uncomment it for PR tests
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.11"]
+    runs-on: [self-hosted, llm, perf]
+    env:
+      OMP_NUM_THREADS: 16
+      THREAD_NUM: 16
+      ANALYTICS_ZOO_ROOT: ${{ github.workspace }}
+      CSV_SAVE_PATH: ${{ github.event.schedule && '/mnt/disk1/nightly_perf_gpu/' || '/mnt/disk1/pr_perf_gpu/' }}
+
+    steps:
+      - uses: actions/checkout@f43a0e5ff2bd294095638e18286ca9a3d1956744 # actions/checkout@v3
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v4
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Install dependencies
+        shell: bash
+        # pip install transformers_stream_generator for model internlm-chat-7b-8k
+        # pip install tiktoken for model Qwen-7B-Chat-10-12
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install --upgrade wheel
+          python -m pip install --upgrade omegaconf
+          python -m pip install --upgrade pandas
+          python -m pip install --upgrade einops
+          python -m pip install --upgrade transformers_stream_generator
+          python -m pip install --upgrade tiktoken
+
+      # please uncomment it and comment the "Install IPEX-LLM from Pypi" part for PR tests
+      # - name: Download llm binary
+      #   uses: ./.github/actions/llm/download-llm-binary
+
+      # - name: Run LLM install (all) test
+      #   uses: ./.github/actions/llm/setup-llm-env
+      #   with:
+      #     extra-dependency: "xpu_2.1"
+
+      - name: Install IPEX-LLM from Pypi
+        shell: bash
+        run: |
+          pip install --pre --upgrade ipex-llm[xpu] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/cn/
+          test_version_date=`date -d 'yesterday' '+%Y%m%d'`
+          if ! pip show ipex-llm | grep $test_version_date; then
+            echo "Did not install ipex-llm with excepted version $test_version_date"
+            exit 1
+          fi
+          pip install transformers==4.36.2
+
+      - name: Test installed xpu version
+        shell: bash
+        run: |
+          source /opt/intel/oneapi/setvars.sh
+          bash python/llm/test/run-llm-install-tests.sh
+
+      - name: Test on xpu(transformers==4.36.2)
+        shell: bash
+        run: |
+          date_for_test_version=$(date -d yesterday +%Y-%m-%d)
+          sed -i "s/date.today()/\"$date_for_test_version\"/g" python/llm/dev/benchmark/all-in-one/run.py
+
+          source /opt/intel/oneapi/setvars.sh
+          export USE_XETLA=OFF
+          export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
+          cp python/llm/test/benchmark/arc-perf-test.yaml python/llm/dev/benchmark/all-in-one/config.yaml
+          cd python/llm/dev/benchmark/all-in-one
+          # hide time info
+          sed -i 's/str(end - st)/"xxxxxx"/g' run.py
+          # change csv name
+          sed -i 's/{today}/{today}_test1/g' run.py
+          python run.py
+
+#      - name: Test on xpu(transformers==4.34.0)
+#        shell: bash
+#        run: |
+#          source /opt/intel/oneapi/setvars.sh
+#          export USE_XETLA=OFF
+#          export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
+#          # upgrade transformers for model Mistral-7B-v0.1
+#          python -m pip install transformers==4.34.0
+#          cp python/llm/test/benchmark/arc-perf-transformers-434.yaml python/llm/dev/benchmark/all-in-one/config.yaml
+#          cd python/llm/dev/benchmark/all-in-one
+#          # change csv name
+#          sed -i 's/test1/test2/g' run.py
+#          python run.py
+
+      - name: Test on xpu(transformers==4.37.0)
+        shell: bash
+        run: |
+          source /opt/intel/oneapi/setvars.sh
+          export USE_XETLA=OFF
+          export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
+          # upgrade transformers for model Qwen/Qwen1.5-7B-Chat
+          python -m pip install transformers==4.37.0
+          cp python/llm/test/benchmark/arc-perf-transformers-437.yaml python/llm/dev/benchmark/all-in-one/config.yaml
+          cd python/llm/dev/benchmark/all-in-one
+          # change csv name
+          sed -i 's/test1/test2/g' run.py
+          python run.py
+
+      - name: Concat csv and generate html
+        shell: bash
+        run: |
+          cd python/llm/dev/benchmark/all-in-one
+          python ../../../test/benchmark/concat_csv.py
+          for file in *.csv; do
+              if [[ $file != *test* ]]; then
+                  cp "$file" $CSV_SAVE_PATH
+              fi
+          done
+          python -m pip install pandas==1.5.3
+          cd ../../../test/benchmark
+          python csv_to_html.py -f $CSV_SAVE_PATH
+
+      - name: Check and upload results to ftp
+        shell: bash
+        run: |
+          cd python/llm/dev/benchmark/all-in-one
+          python ../../../test/benchmark/check_results.py -c test1 -y ../../../test/benchmark/arc-perf-test.yaml
+          find . -name "*test*.csv" -delete
+          if [ ${{ github.event.schedule}} ]; then
+            curl -T ./*.csv ${LLM_FTP_URL}/llm/nightly_perf/gpu/
+          fi
+          
+  llm-performance-test-on-spr:
+    if: ${{ github.event.schedule || github.event.inputs.artifact == 'llm-performance-test-on-spr' || github.event.inputs.artifact == 'all' }} # please comment it for PR tests
+    # needs: llm-cpp-build # please uncomment it for PR tests
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.11"]
+    runs-on: [self-hosted, llm, spr-perf]
+    env:
+      OMP_NUM_THREADS: 16
+      THREAD_NUM: 16
+      ANALYTICS_ZOO_ROOT: ${{ github.workspace }}
+    steps:
+      - uses: actions/checkout@f43a0e5ff2bd294095638e18286ca9a3d1956744 # actions/checkout@v3
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v4
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Install dependencies
+        shell: bash
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install --upgrade wheel
+          python -m pip install --upgrade omegaconf
+          python -m pip install --upgrade pandas
+          python -m pip install --upgrade einops
+          python -m pip install --upgrade tiktoken
+          python -m pip install --upgrade transformers_stream_generator
+
+      # please uncomment it and comment the "Install IPEX-LLM from Pypi" part for PR tests
+      # - name: Download llm binary
+      #   uses: ./.github/actions/llm/download-llm-binary
+
+      # - name: Run LLM install (all) test
+      #   uses: ./.github/actions/llm/setup-llm-env
+
+      - name: Install IPEX-LLM from Pypi
+        shell: bash
+        run: |
+          pip install --pre --upgrade ipex-llm[all] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/cn/
+          test_version_date=`date -d 'yesterday' '+%Y%m%d'`
+          if ! pip show ipex-llm | grep $test_version_date; then
+            echo "Did not install ipex-llm with excepted version $test_version_date"
+            exit 1
+          fi
+          pip install tranformers==4.36.2
+
+      - name: Test on cpu
+        shell: bash
+        run: |
+          date_for_test_version=$(date -d yesterday +%Y-%m-%d)
+          sed -i "s/date.today()/\"$date_for_test_version\"/g" python/llm/dev/benchmark/all-in-one/run.py
+
+          mv python/llm/test/benchmark/cpu-perf-test.yaml python/llm/dev/benchmark/all-in-one/config.yaml
+          cd python/llm/dev/benchmark/all-in-one
+          export http_proxy=${HTTP_PROXY}
+          export https_proxy=${HTTPS_PROXY}
+          source ipex-llm-init -t
+          export OMP_NUM_THREADS=48
+          # hide time info
+          sed -i 's/str(end - st)/"xxxxxx"/g' run.py
+          python run.py
+          cp ./*.csv /mnt/disk1/models/nightly_perf_cpu
+          cd ../../../test/benchmark
+          python -m pip install pandas==1.5.3
+          python csv_to_html.py -f /mnt/disk1/models/nightly_perf_cpu
+          cd /mnt/disk1/models/nightly_perf_cpu
+          for f in *.html; do
+            curl -T "$f" ${LLM_FTP_URL}/llm/nightly_perf/nightly_perf_cpu/
+          done
+
+  llm-performance-test-on-core:
+    if: ${{ github.event.schedule || github.event.inputs.artifact == 'llm-performance-test-on-core' || github.event.inputs.artifact == 'all' }} # please comment it for PR tests
+    # needs: llm-cpp-build # please uncomment it for PR tests
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - os: windows
+            platform: dp
+            python-version: "3.11"
+          # - os: windows
+          #   platform: lp
+          #   python-version: "3.11"
+    runs-on: [self-hosted, "${{ matrix.os }}", llm, perf-core, "${{ matrix.platform }}"]
+    env:
+      ANALYTICS_ZOO_ROOT: ${{ github.workspace }}
+      CSV_SAVE_PATH: ${{ github.event.schedule && 'D:/action-runners/nightly_perf_core_' || 'D:/action-runners/pr_perf_core_' }}${{ matrix.platform }}/
+    steps:
+      - uses: actions/checkout@f43a0e5ff2bd294095638e18286ca9a3d1956744 # actions/checkout@v3
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v4
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Install dependencies
+        shell: bash
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install --upgrade wheel
+          python -m pip install --upgrade omegaconf pandas
+          python -m pip install --upgrade tiktoken einops transformers_stream_generator
+    
+      # please uncomment it and comment the "Install IPEX-LLM from Pypi" part for PR tests
+      # - name: Download llm binary
+      #   uses: ./.github/actions/llm/download-llm-binary
+
+      # - name: Run LLM install (all) test
+      #   uses: ./.github/actions/llm/setup-llm-env
+
+      - name: Install IPEX-LLM from Pypi
+        shell: bash
+        run: |
+          pip install --pre --upgrade ipex-llm[all] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/cn/
+          test_version_date=`date -d 'yesterday' '+%Y%m%d'`
+          if ! pip show ipex-llm | grep $test_version_date; then
+            echo "Did not install ipex-llm with excepted version $test_version_date"
+            exit 1
+          fi
+          pip install transformers==4.36.2
+
+      - name: Test on core ${{ matrix.platform }}
+        shell: bash
+        run: |
+          date_for_test_version=$(date -d yesterday +%Y-%m-%d)
+          sed -i "s/date.today()/\"$date_for_test_version\"/g" python/llm/dev/benchmark/all-in-one/run.py
+
+          mv python/llm/test/benchmark/core-perf-test.yaml python/llm/dev/benchmark/all-in-one/config.yaml
+          cd python/llm/dev/benchmark/all-in-one
+          export http_proxy=${HTTP_PROXY}
+          export https_proxy=${HTTPS_PROXY}
+          # hide time info
+          sed -i 's/str(end - st)/"xxxxxx"/g' run.py
+          python run.py
+          cp ./*.csv $CSV_SAVE_PATH
+          cd ../../../test/benchmark
+          python -m pip install pandas==1.5.3
+          python csv_to_html.py -f $CSV_SAVE_PATH
+          cd ../../dev/benchmark/all-in-one/
+          if [ ${{ github.event.schedule}} ]; then
+            curl -T ./*.csv ${LLM_FTP_URL}/llm/nightly_perf/core_${{ matrix.platform }}/
+          fi
+
+  llm-performance-test-on-igpu:
+    if: ${{ github.event.schedule || github.event.inputs.artifact == 'llm-performance-test-on-igpu' || github.event.inputs.artifact == 'all' }} # please comment it for PR tests
+    # needs: llm-cpp-build # please uncomment it for PR tests
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - os: windows
+            python-version: "3.11"
+    runs-on: [self-hosted, "${{ matrix.os }}", llm, perf-igpu]
+    env:
+      ANALYTICS_ZOO_ROOT: ${{ github.workspace }}
+    steps:
+      - uses: actions/checkout@f43a0e5ff2bd294095638e18286ca9a3d1956744 # actions/checkout@v3
+
+      # TODO: Put the ipex-llm related install process for win gpu into a action function
+
+      # Please uncomment it and commment the install from pypi for PR tests
+      # - name: Download llm binary
+      #   uses: ./.github/actions/llm/download-llm-binary
+
+      # - name: Prepare for install ipex-llm from source
+      #   shell: bash
+      #   run: |
+      #     sed -i 's/"bigdl-core-xe-21==" + VERSION + "/"bigdl-core-xe-21/g' python/llm/setup.py
+      #     sed -i 's/"bigdl-core-xe-21==" + VERSION/"bigdl-core-xe-21"/g' python/llm/setup.py
+
+      # - name: Install ipex-llm and other related packages (install from source)
+      #   shell: cmd
+      #   run: |
+      #     call conda create -n igpu-perf python=${{ matrix.python-version }} libuv -y
+      #     call conda activate igpu-perf
+
+      #     pip install --upgrade pip
+      #     pip install --upgrade wheel
+      #     pip install --upgrade omegaconf pandas
+      #     pip install --upgrade tiktoken einops transformers_stream_generator
+
+      #     cd python\llm
+      #     python setup.py clean --all bdist_wheel --win
+      #     if not exist dist\ipex_llm*.whl (exit /b 1)
+      #     for %%i in (dist\ipex_llm*.whl) do set whl_name=%%i
+
+      #     pip install --pre --upgrade %whl_name%[xpu] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/cn/
+      #     if %ERRORLEVEL% neq 0 (exit /b 1)
+      #     pip list
+
+      #     call conda deactivate
+
+      - name: Determine desired ipex-llm version
+        shell: bash
+        run: |
+          test_version_date=`date -d 'yesterday' '+%Y%m%d'`
+          echo "TEST_VERSION_DATE=${test_version_date}" >> "$GITHUB_ENV"
+
+      - name: Install ipex-llm and other related packages (install from pypi)
+        shell: cmd
+        run: |
+          call conda create -n igpu-perf python=${{ matrix.python-version }} libuv -y
+          call conda activate igpu-perf
+
+          pip install --upgrade pip
+          pip install --upgrade wheel
+          pip install --upgrade omegaconf pandas
+          pip install --upgrade tiktoken einops transformers_stream_generator
+
+          pip install --pre --upgrade ipex-llm[xpu] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/cn/
+#          pip show ipex-llm | findstr %TEST_VERSION_DATE%
+#          if %ERRORLEVEL% neq 0 (
+#            echo "Did not install ipex-llm with excepted version %TEST_VERSION_DATE%"
+#            exit /b 1
+#          )
+          pip install transformers==4.36.2
+          pip list
+
+          call conda deactivate
+
+      - name: Create env for html generation
+        shell: cmd
+        run: |
+          call conda create -n html-gen python=3.11 -y
+          call conda activate html-gen
+
+          pip install pandas==1.5.3
+          pip install Jinja2
+
+          call conda deactivate
+
+      - name: Set directory envs & and fix generated csv date name
+        shell: bash
+        run: |
+          if [ ${{ github.event_name }} == 'schedule' ]; then
+            echo "CSV_SAVE_PATH=${CSV_NIGHTLY_PATH}" >> "$GITHUB_ENV"
+          else
+            echo "CSV_SAVE_PATH=${CSV_PR_PATH}" >> "$GITHUB_ENV"
+          fi
+          date_for_test_version=$(date -d yesterday +%Y-%m-%d)
+          echo "LOG_FILE=${date_for_test_version}_output.txt" >> "$GITHUB_ENV"
+
+          sed -i "s/date.today()/\"$date_for_test_version\"/g" python/llm/dev/benchmark/all-in-one/run.py
+
+      - name: Prepare igpu perf test (32-32)
+        shell: bash
+        run: |
+          # hide time info
+          # sed -i 's/str(end - st)/"xxxxxx"/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i 's/{api}-results-{today}.csv/32-32-{api}-results-{today}_test1.csv/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/32-32.yaml
+
+      - name: Test on igpu (32-32)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+          REM for llava
+          set TRANSFORMERS_OFFLINE=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\32-32.yaml config.yaml
+          set PYTHONIOENCODING=utf-8
+          python run.py >> %CSV_SAVE_PATH%\32-32\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+#      - name: Prepare igpu perf test for Mistral (32-32)
+#        shell: bash
+#        run: |
+#          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+#          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/32-32_434.yaml
+
+#      - name: Test on igpu for Mistral (32-32)
+#        shell: cmd
+#        run: |
+#          call conda activate igpu-perf
+#          pip install transformers==4.34.0
+#
+#          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+#          set SYCL_CACHE_PERSISTENT=1
+#          set BIGDL_LLM_XMX_DISABLED=1
+#
+#          cd python\llm\dev\benchmark\all-in-one
+#          move ..\..\..\test\benchmark\igpu-perf\32-32_434.yaml config.yaml
+#          set PYTHONIOENCODING=utf-8
+#          python run.py >> %CSV_SAVE_PATH%\32-32\log\%LOG_FILE% 2>&1
+#          if %ERRORLEVEL% neq 0 (exit /b 1)
+#
+#          call conda deactivate
+
+      - name: Prepare igpu perf test for Qwen1.5 (32-32)
+        shell: bash
+        run: |
+          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/32-32_437.yaml
+
+      - name: Test on igpu for Qwen1.5 (32-32)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          pip install transformers==4.37.0
+
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\32-32_437.yaml config.yaml
+          set PYTHONIOENCODING=utf-8
+          python run.py >> %CSV_SAVE_PATH%\32-32\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+      - name: Concat csv and generate html (32-32)
+        shell: cmd
+        run: |
+          call conda activate html-gen
+
+          cd python\llm\dev\benchmark\all-in-one
+          python ..\..\..\test\benchmark\concat_csv.py
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+          del /q *test*.csv
+          move *.csv %CSV_SAVE_PATH%\32-32\
+          cd ..\..\..\test\benchmark
+          python csv_to_html.py -f %CSV_SAVE_PATH%\32-32\
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+          move %CSV_SAVE_PATH%\32-32\*.html %CSV_SAVE_PATH%
+
+          call conda deactivate
+
+      # TODO: create a action function here for different input
+      # 1024-128
+      - name: Prepare igpu perf test (1024-128)
+        shell: bash
+        run: |
+          sed -i 's/32-32/1024-128/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i 's/{today}_test2/{today}_test1/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/1024-128.yaml
+
+      - name: Test on igpu (1024-128)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          pip install transformers==4.36.2
+
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+          REM for llava
+          set TRANSFORMERS_OFFLINE=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\1024-128.yaml config.yaml
+          set PYTHONIOENCODING=utf-8
+          python run.py >> %CSV_SAVE_PATH%\1024-128\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+#      - name: Prepare igpu perf test for Mistral (1024-128)
+#        shell: bash
+#        run: |
+#          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+#          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/1024-128_434.yaml
+#
+#      - name: Test on igpu for Mistral (1024-128)
+#        shell: cmd
+#        run: |
+#          call conda activate igpu-perf
+#          pip install transformers==4.34.0
+#
+#          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+#          set SYCL_CACHE_PERSISTENT=1
+#          set BIGDL_LLM_XMX_DISABLED=1
+#
+#          cd python\llm\dev\benchmark\all-in-one
+#          move ..\..\..\test\benchmark\igpu-perf\1024-128_434.yaml config.yaml
+#          set PYTHONIOENCODING=utf-8
+#          python run.py >> %CSV_SAVE_PATH%\1024-128\log\%LOG_FILE% 2>&1
+#          if %ERRORLEVEL% neq 0 (exit /b 1)
+#
+#          call conda deactivate
+
+      - name: Prepare igpu perf test for Qwen 1.5 (1024-128)
+        shell: bash
+        run: |
+          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/1024-128_437.yaml
+
+      - name: Test on igpu for Qwen 1.5 (1024-128)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          pip install transformers==4.37.0
+
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\1024-128_437.yaml config.yaml
+          set PYTHONIOENCODING=utf-8
+          python run.py >> %CSV_SAVE_PATH%\1024-128\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+      - name: Concat csv and generate html (1024-128)
+        shell: cmd
+        run: |
+          call conda activate html-gen
+
+          cd python\llm\dev\benchmark\all-in-one
+          python ..\..\..\test\benchmark\concat_csv.py
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+          del /q *test*.csv
+          move *.csv %CSV_SAVE_PATH%\1024-128\
+          cd ..\..\..\test\benchmark
+          python csv_to_html.py -f %CSV_SAVE_PATH%\1024-128\
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+          move %CSV_SAVE_PATH%\1024-128\*.html %CSV_SAVE_PATH%
+
+          call conda deactivate
+
+      # 2048-256
+      - name: Prepare igpu perf test (2048-256)
+        shell: bash
+        run: |
+          sed -i 's/1024-128/2048-256/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i 's/{today}_test2/{today}_test1/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/2048-256.yaml
+
+      - name: Test on igpu (2048-256)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          pip install transformers==4.36.2
+
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+          REM for llava
+          set TRANSFORMERS_OFFLINE=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\2048-256.yaml config.yaml
+          set PYTHONIOENCODING=utf-8
+          python run.py >> %CSV_SAVE_PATH%\2048-256\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+#      - name: Prepare igpu perf test for Mistral (2048-256)
+#        shell: bash
+#        run: |
+#          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+#          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/2048-256_434.yaml
+
+#      - name: Test on igpu for Mistral (2048-256)
+#        shell: cmd
+#        run: |
+#          call conda activate igpu-perf
+#          pip install transformers==4.34.0
+#
+#          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+#          set SYCL_CACHE_PERSISTENT=1
+#          set BIGDL_LLM_XMX_DISABLED=1
+#
+#          cd python\llm\dev\benchmark\all-in-one
+#          move ..\..\..\test\benchmark\igpu-perf\2048-256_434.yaml config.yaml
+#          set PYTHONIOENCODING=utf-8
+#          python run.py >> %CSV_SAVE_PATH%\2048-256\log\%LOG_FILE% 2>&1
+#          if %ERRORLEVEL% neq 0 (exit /b 1)
+#
+#          call conda deactivate
+
+      - name: Prepare igpu perf test for Qwen 1.5 (2048-256)
+        shell: bash
+        run: |
+          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/2048-256_437.yaml
+
+      - name: Test on igpu for Qwen 1.5 (2048-256)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          pip install transformers==4.37.0
+
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\2048-256_437.yaml config.yaml
+          set PYTHONIOENCODING=utf-8
+          python run.py >> %CSV_SAVE_PATH%\2048-256\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+      - name: Concat csv and generate html (2048-256)
+        shell: cmd
+        run: |
+          call conda activate html-gen
+
+          cd python\llm\dev\benchmark\all-in-one
+          python ..\..\..\test\benchmark\concat_csv.py
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+          del /q *test*.csv
+          move *.csv %CSV_SAVE_PATH%\2048-256\
+          cd ..\..\..\test\benchmark
+          python csv_to_html.py -f %CSV_SAVE_PATH%\2048-256\
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+          move %CSV_SAVE_PATH%\2048-256\*.html %CSV_SAVE_PATH%
+
+          call conda deactivate
+
+      # load_low_bit 1024-128 
+      - name: Prepare igpu perf test (load_low_bit 1024-128)
+        shell: bash
+        run: |
+          sed -i 's/2048-256/1024-128/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i 's/{today}_test2/{today}_test1/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/1024-128_loadlowbit.yaml
+
+      - name: Test on igpu (load_low_bit 1024-128)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          pip install transformers==4.36.2
+
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+          REM for llava
+          set TRANSFORMERS_OFFLINE=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\1024-128_loadlowbit.yaml config.yaml
+          set PYTHONIOENCODING=utf-8
+          python run.py >> %CSV_SAVE_PATH%\1024-128_loadlowbit\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+#      - name: Prepare igpu perf test for Mistral (load_low_bit 1024-128)
+#        shell: bash
+#        run: |
+#          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+#          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/1024-128_loadlowbit_434.yaml
+#
+#      - name: Test on igpu for Mistral (load_low_bit 1024-128)
+#        shell: cmd
+#        run: |
+#          call conda activate igpu-perf
+#          pip install transformers==4.34.0
+#
+#          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+#          set SYCL_CACHE_PERSISTENT=1
+#          set BIGDL_LLM_XMX_DISABLED=1
+#
+#          cd python\llm\dev\benchmark\all-in-one
+#          move ..\..\..\test\benchmark\igpu-perf\1024-128_loadlowbit_434.yaml config.yaml
+#          set PYTHONIOENCODING=utf-8
+#          python run.py >> %CSV_SAVE_PATH%\1024-128_loadlowbit\log\%LOG_FILE% 2>&1
+#          if %ERRORLEVEL% neq 0 (exit /b 1)
+#
+#          call conda deactivate
+
+      - name: Prepare igpu perf test for Qwen 1.5 (load_low_bit 1024-128)
+        shell: bash
+        run: |
+          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/1024-128_loadlowbit_437.yaml
+
+      - name: Test on igpu for Qwen 1.5 (load_low_bit 1024-128)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          pip install transformers==4.37.0
+
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\1024-128_loadlowbit_437.yaml config.yaml
+          set PYTHONIOENCODING=utf-8
+          python run.py >> %CSV_SAVE_PATH%\1024-128_loadlowbit\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+      - name: Concat csv and generate html (load_low_bit 1024-128)
+        shell: cmd
+        run: |
+          call conda activate html-gen
+
+          cd python\llm\dev\benchmark\all-in-one
+          python ..\..\..\test\benchmark\concat_csv.py
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+          del /q *test*.csv
+          move *.csv %CSV_SAVE_PATH%\1024-128_loadlowbit\
+          cd ..\..\..\test\benchmark
+          python csv_to_html.py -f %CSV_SAVE_PATH%\1024-128_loadlowbit\
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+          move %CSV_SAVE_PATH%\1024-128_loadlowbit\*.html %CSV_SAVE_PATH%
+
+          call conda deactivate
+
+      - name: Prepare igpu perf test (int4+fp16 1024-128)
+        shell: bash
+        run: |
+          sed -i 's/{today}_test2/{today}_test1/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/1024-128_int4_fp16.yaml
+
+      - name: Test on igpu (int4+fp16 1024-128)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          pip install transformers==4.36.2
+
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+          REM for llava
+          set TRANSFORMERS_OFFLINE=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\1024-128_int4_fp16.yaml config.yaml
+          set PYTHONIOENCODING=utf-8
+          python run.py >> %CSV_SAVE_PATH%\1024-128_int4_fp16\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+#      - name: Prepare igpu perf test for Mistral (int4+fp16 1024-128)
+#        shell: bash
+#        run: |
+#          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+#          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/1024-128_int4_fp16_434.yaml
+#
+#      - name: Test on igpu for Mistral (int4+fp16 1024-128)
+#        shell: cmd
+#        run: |
+#          call conda activate igpu-perf
+#          pip install transformers==4.34.0
+#
+#          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+#          set SYCL_CACHE_PERSISTENT=1
+#          set BIGDL_LLM_XMX_DISABLED=1
+#
+#          cd python\llm\dev\benchmark\all-in-one
+#          move ..\..\..\test\benchmark\igpu-perf\1024-128_int4_fp16_434.yaml config.yaml
+#          set PYTHONIOENCODING=utf-8
+#          python run.py >> %CSV_SAVE_PATH%\1024-128_int4_fp16\log\%LOG_FILE% 2>&1
+#          if %ERRORLEVEL% neq 0 (exit /b 1)
+#
+#          call conda deactivate
+
+      - name: Prepare igpu perf test for Qwen 1.5 (int4+fp16 1024-128)
+        shell: bash
+        run: |
+          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/1024-128_int4_fp16_437.yaml
+
+      - name: Test on igpu for Qwen 1.5 (int4+fp16 1024-128)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          pip install transformers==4.37.0
+
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\1024-128_int4_fp16_437.yaml config.yaml
+          set PYTHONIOENCODING=utf-8
+          python run.py >> %CSV_SAVE_PATH%\1024-128_int4_fp16\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+      - name: Concat csv and generate html (int4+fp16 1024-128)
+        shell: cmd
+        run: |
+          call conda activate html-gen
+
+          cd python\llm\dev\benchmark\all-in-one
+          python ..\..\..\test\benchmark\concat_csv.py
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+          del /q *test*.csv
+          move *.csv %CSV_SAVE_PATH%\1024-128_int4_fp16\
+          cd ..\..\..\test\benchmark
+          python csv_to_html.py -f %CSV_SAVE_PATH%\1024-128_int4_fp16\
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+          move %CSV_SAVE_PATH%\1024-128_int4_fp16\*.html %CSV_SAVE_PATH%
+
+          call conda deactivate
+
+      - name: Upload results to ftp
+        if: ${{ always() }}
+        shell: cmd
+        run: |
+          cd %CSV_SAVE_PATH%
+          IF "${{ github.event_name }}"=="schedule" (
+            for %%f in (*.html) do (
+                curl -T "%%f" %FTP_IGPU_NIGHTLY_PERF_PATH%
+            )
+          )
+
+      # for test on machine when encountering error
+      # - name: Remove conda env
+      #   if: ${{ always() }}
+      #   shell: cmd
+      #   run: |
+      #     call conda env remove -n igpu-perf -y
diff --git a/.github/workflows/llm_unit_tests_4.36.yml b/.github/workflows/llm_unit_tests_4.36.yml
index a0a483e7243..345dd34124c 100644
--- a/.github/workflows/llm_unit_tests_4.36.yml
+++ b/.github/workflows/llm_unit_tests_4.36.yml
@@ -23,18 +23,18 @@ on:
       - ".github/actions/llm/cli-test-linux/action.yml"
       - ".github/actions/llm/cli-test-windows/action.yml"
       - ".github/actions/llm/download-llm-binary/action.yml"
-  pull_request:
-    branches: [main]
-    paths:
-      - "python/llm/**"
-      - ".github/workflows/llm_unit_tests.yml"
-      - ".github/workflows/llm_unit_tests_4.36.yml"
-      - ".github/workflows/llm-binary-build.yml"
-      - ".github/actions/llm/setup-llm-env/action.yml"
-      - ".github/actions/llm/remove-llm-env/action.yml"
-      - ".github/actions/llm/cli-test-linux/action.yml"
-      - ".github/actions/llm/cli-test-windows/action.yml"
-      - ".github/actions/llm/download-llm-binary/action.yml"
+#  pull_request:
+#    branches: [main]
+#    paths:
+#      - "python/llm/**"
+#      - ".github/workflows/llm_unit_tests.yml"
+#      - ".github/workflows/llm_unit_tests_4.36.yml"
+#      - ".github/workflows/llm-binary-build.yml"
+#      - ".github/actions/llm/setup-llm-env/action.yml"
+#      - ".github/actions/llm/remove-llm-env/action.yml"
+#      - ".github/actions/llm/cli-test-linux/action.yml"
+#      - ".github/actions/llm/cli-test-windows/action.yml"
+#      - ".github/actions/llm/download-llm-binary/action.yml"
   workflow_dispatch:
   workflow_call:
 
diff --git a/python/llm/dev/test/run-example-tests-436.sh b/python/llm/dev/test/run-example-tests-436.sh
new file mode 100644
index 00000000000..c3e92b0f032
--- /dev/null
+++ b/python/llm/dev/test/run-example-tests-436.sh
@@ -0,0 +1,85 @@
+if [ -z "$THREAD_NUM" ]; then
+  THREAD_NUM=2
+fi
+export OMP_NUM_THREADS=$THREAD_NUM
+
+######## LLAMA2
+# transformers
+export ORIGINAL_LLAMA2_PATH=./llm/Llama-2-7b-chat-hf/
+if [ ! -d $ORIGINAL_LLAMA2_PATH ]; then
+    echo "Directory $ORIGINAL_LLAMA2_PATH not found. Downloading from FTP server..."
+    wget -r -nH --no-verbose --cut-dirs=1 $LLM_FTP_URL/${ORIGINAL_LLAMA2_PATH:2} -P $LLM_DIR
+fi
+
+echo ">>> Testing LLAMA2 transformers API"
+std=$(taskset -c 0-$((THREAD_NUM - 1)) python python/llm/example/CPU/HF-Transformers-AutoModels/Model/llama2/generate.py --repo-id-or-model-path $ORIGINAL_LLAMA2_PATH)
+echo "the output of the example is: " 
+echo $std
+if [[ ! $std == *"AI is a term"* ]]; then
+    echo "The expected output is not met."
+    return 1
+fi
+# transformers low-bit
+echo ">>> Testing LLAMA2 transformers API sym_int4"
+std=$(taskset -c 0-$((THREAD_NUM - 1)) python python/llm/example/CPU/HF-Transformers-AutoModels/More-Data-Types/transformers_low_bit_pipeline.py --repo-id-or-model-path $ORIGINAL_LLAMA2_PATH)
+echo "the output of the example is: " 
+echo $std
+if [[ ! $std == *"But her parents were always telling her to stay close to home"* ]]; then
+    echo "The expected output is not met."
+    return 1
+fi
+
+echo ">>> Testing LLAMA2 transformers API sym_int5"
+std=$(taskset -c 0-$((THREAD_NUM - 1)) python python/llm/example/CPU/HF-Transformers-AutoModels/More-Data-Types/transformers_low_bit_pipeline.py --repo-id-or-model-path $ORIGINAL_LLAMA2_PATH --low-bit sym_int5)
+echo "the output of the example is: " 
+echo $std
+if [[ ! $std == *"She wanted to go to places and meet new people"* ]]; then
+    echo "The expected output is not met."
+    return 1
+fi
+echo ">>> Testing LLAMA2 transformers API sym_int8"
+std=$(taskset -c 0-$((THREAD_NUM - 1)) python python/llm/example/CPU/HF-Transformers-AutoModels/More-Data-Types/transformers_low_bit_pipeline.py --repo-id-or-model-path $ORIGINAL_LLAMA2_PATH --low-bit sym_int8)
+echo "the output of the example is: " 
+echo $std
+if [[ ! $std == *"She wanted to go to places and meet new people"* ]]; then
+    echo "The expected output is not met."
+    return 1
+fi
+echo ">>> Testing LLAMA2 transformers API asym_int4"
+std=$(taskset -c 0-$((THREAD_NUM - 1)) python python/llm/example/CPU/HF-Transformers-AutoModels/More-Data-Types/transformers_low_bit_pipeline.py --repo-id-or-model-path $ORIGINAL_LLAMA2_PATH --low-bit asym_int4)
+echo "the output of the example is: " 
+echo $std
+if [[ ! $std == *"She wanted to go to places and meet new people"* ]]; then
+    echo "The expected output is not met."
+    return 1
+fi
+
+echo ">>> Testing LLAMA2 transformers API asym_int5"
+std=$(taskset -c 0-$((THREAD_NUM - 1)) python python/llm/example/CPU/HF-Transformers-AutoModels/More-Data-Types/transformers_low_bit_pipeline.py --repo-id-or-model-path $ORIGINAL_LLAMA2_PATH --low-bit asym_int5)
+echo "the output of the example is: " 
+echo $std
+if [[ ! $std == *"She wanted to go to places and meet new people"* ]]; then
+    echo "The expected output is not met."
+    return 1
+fi
+
+
+########## ChatGLM2
+export ORIGINAL_CHATGLM2_PATH=./llm/chatglm2-6b/
+if [ ! -d $ORIGINAL_CHATGLM2_PATH ]; then
+    echo "Directory $ORIGINAL_CHATGLM2_PATH not found. Downloading from FTP server..."
+    wget -r -nH --no-verbose --cut-dirs=1 $LLM_FTP_URL/updated_for_4.36/${ORIGINAL_CHATGLM2_PATH:6} -P $LLM_DIR
+fi
+
+echo ">>> Testing ChatGLM2 transformers API"
+std=$(taskset -c 0-$((THREAD_NUM - 1)) python python/llm/example/CPU/HF-Transformers-AutoModels/Model/chatglm2/generate.py --repo-id-or-model-path $ORIGINAL_CHATGLM2_PATH)
+echo "the output of the example is: " 
+echo $std
+if [[ ! $std == *"AI指的是人工智能"* ]]; then
+    echo "The expected output is not met."
+    return 1
+fi
+
+
+
+
diff --git a/python/llm/test/benchmark/arc-perf-test.yaml b/python/llm/test/benchmark/arc-perf-test.yaml
index 47f74b20e7e..782aa5e2619 100644
--- a/python/llm/test/benchmark/arc-perf-test.yaml
+++ b/python/llm/test/benchmark/arc-perf-test.yaml
@@ -17,6 +17,7 @@ repo_id:
   - 'baichuan-inc/Baichuan2-13B-Chat-4bit'
   - 'bigscience/bloomz-7b1'
   - 'fnlp/moss-moon-003-sft-4bit'
+  - 'mistralai/Mistral-7B-v0.1'
 local_model_hub: '/mnt/disk1/models'
 warm_up: 1
 num_trials: 3
diff --git a/python/llm/test/benchmark/igpu-perf/32-32.yaml b/python/llm/test/benchmark/igpu-perf/32-32.yaml
index 35f6736645d..e446e641ac3 100644
--- a/python/llm/test/benchmark/igpu-perf/32-32.yaml
+++ b/python/llm/test/benchmark/igpu-perf/32-32.yaml
@@ -16,6 +16,7 @@ repo_id:
   - 'RWKV/rwkv-4-world-7b'
   - 'RWKV/rwkv-5-world-7b'
   - 'IEITYuan/Yuan2-2B-hf'
+  - 'mistralai/Mistral-7B-Instruct-v0.1'
 local_model_hub: 'path to your local model hub'
 warm_up: 3
 num_trials: 5