[LLM] Enable more long context in-out pairs for iGPU perf tests (#9765)

* Add test for 1024-128 and enable more tests for 512-64

* Fix date in results csv name to the time when the performance is triggered

* Small fix

* Small fix

* further fixes

.github/workflows/llm_performance_tests.yml

@@ -275,7 +275,7 @@ jobs:
 
           call conda deactivate
 
-      - name: Set directory envs
+      - name: Set directory envs & and fix generated csv date name
         shell: bash
         run: |
           if [ ${{ github.event_name }} == 'schedule' ]; then
@@ -286,6 +286,8 @@ jobs:
           cur_date=$(date +%Y-%m-%d)
           echo "LOG_FILE=${cur_date}_output.txt" >> "$GITHUB_ENV"
 
+          sed -i "s/date.today()/\"$cur_date\"/g" python/llm/dev/benchmark/all-in-one/run.py
+
       - name: Prepare igpu perf test (32-32)
         shell: bash
         run: |
@@ -489,7 +491,7 @@ jobs:
         shell: bash
         run: |
           sed -i 's/32-512/512-64/g' python/llm/dev/benchmark/all-in-one/run.py
-          sed -i 's/{today}_test2/{today}/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i 's/{today}_test2/{today}_test1/g' python/llm/dev/benchmark/all-in-one/run.py
           sed -i "s/path to your local model hub/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/512-64.yaml
 
       - name: Test on igpu (512-64)
@@ -511,12 +513,37 @@ jobs:
 
           call conda deactivate
 
-      - name: Generate html (512-64)
+      - name: Prepare igpu perf test for Mistral (512-64)
+        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/512-64_434.yaml
+
+      - name: Test on igpu for Mistral (512-64)
+        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\512-64_434.yaml config.yaml
+          python run.py >> %CSV_SAVE_PATH%\512-64\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+      - name: Concat csv and generate html (512-64)
         shell: cmd
         run: |
           call conda activate html-gen
 
           cd python\llm\dev\benchmark\all-in-one
+          python ..\..\..\test\benchmark\concat_csv.py
+          del /q *test*.csv
           move *.csv %CSV_SAVE_PATH%\512-64\
           cd ..\..\..\test\benchmark
           python csv_to_html.py -f %CSV_SAVE_PATH%\512-64\
@@ -525,6 +552,72 @@ jobs:
 
           call conda deactivate
 
+      # 1024-128
+      - name: Prepare igpu perf test (1024-128)
+        shell: bash
+        run: |
+          sed -i 's/512-64/1024-128/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/$MODEL_HUB_DIR/g" python/llm/test/benchmark/igpu-perf/1024-128.yaml
+
+      - name: Test on igpu (1024-128)
+        shell: cmd
+        run: |
+          call conda activate igpu-perf
+          pip install transformers==4.31.0
+
+          call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+          set SYCL_CACHE_PERSISTENT=1
+          set BIGDL_LLM_XMX_DISABLED=1
+          REM for llava
+          set TRANSFORMERS_OFFLINE=1
+
+          cd python\llm\dev\benchmark\all-in-one
+          move ..\..\..\test\benchmark\igpu-perf\1024-128.yaml config.yaml
+          python run.py >> %CSV_SAVE_PATH%\1024-128\log\%LOG_FILE% 2>&1
+          if %ERRORLEVEL% neq 0 (exit /b 1)
+
+          call conda deactivate
+
+      - name: Prepare igpu perf test for Mistral (1024-128)
+        shell: bash
+        run: |
+          sed -i 's/{today}_test1/{today}_test2/g' python/llm/dev/benchmark/all-in-one/run.py
+          sed -i "s/path to your local model hub/$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
+          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
+          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
+
       # for test on machine when encountering error
       # - name: Remove conda env
       #   if: ${{ always() }}

python/llm/test/benchmark/igpu-perf/1024-128.yaml

@@ -0,0 +1,24 @@
+repo_id:
+  - 'THUDM/chatglm2-6b'
+  - 'THUDM/chatglm3-6b'
+  - 'baichuan-inc/Baichuan2-7B-Chat'
+  - 'internlm/internlm-chat-7b-8k'
+  # - 'Qwen/Qwen-7B-Chat'
+  - 'BAAI/AquilaChat2-7B'
+  - '01-ai/Yi-6B'
+  - 'meta-llama/Llama-2-7b-chat-hf'
+  # - 'WisdomShell/CodeShell-7B-Chat'
+  - 'tiiuae/falcon-7b-instruct-with-patch'
+  - 'mosaicml/mpt-7b-chat'
+  - 'liuhaotian/llava-v1.5-7b'
+  - 'RWKV/rwkv-4-world-7b'
+local_model_hub: 'path to your local model hub'
+warm_up: 1
+num_trials: 3
+num_beams: 1 # default to greedy search
+low_bit: 'sym_int4' # default to use 'sym_int4' (i.e. symmetric int4)
+in_out_pairs:
+  - '1024-128'
+test_api:
+  - "transformer_int4_gpu_win" # on Intel GPU for Windows (catch GPU peak memory)
+cpu_embedding: True # whether put embedding to CPU (only avaiable now for gpu win related test_api)

python/llm/test/benchmark/igpu-perf/1024-128_434.yaml

@@ -0,0 +1,12 @@
+repo_id:
+  - 'mistralai/Mistral-7B-Instruct-v0.1'
+local_model_hub: 'path to your local model hub'
+warm_up: 1
+num_trials: 3
+num_beams: 1 # default to greedy search
+low_bit: 'sym_int4' # default to use 'sym_int4' (i.e. symmetric int4)
+in_out_pairs:
+  - '1024-128'
+test_api:
+  - "transformer_int4_gpu_win" # on Intel GPU for Windows (catch GPU peak memory)
+cpu_embedding: True # whether put embedding to CPU (only avaiable now for gpu win related test_api)

python/llm/test/benchmark/igpu-perf/512-64.yaml

@@ -1,13 +1,13 @@
 repo_id:
   - 'THUDM/chatglm2-6b'
   - 'THUDM/chatglm3-6b'
-  # - 'baichuan-inc/Baichuan2-7B-Chat'
+  - 'baichuan-inc/Baichuan2-7B-Chat'
   - 'internlm/internlm-chat-7b-8k'
-  # - 'Qwen/Qwen-7B-Chat'
+  - 'Qwen/Qwen-7B-Chat'
   - 'BAAI/AquilaChat2-7B'
   - '01-ai/Yi-6B'
   - 'meta-llama/Llama-2-7b-chat-hf'
-  # - 'WisdomShell/CodeShell-7B-Chat'
+  - 'WisdomShell/CodeShell-7B-Chat'
   - 'tiiuae/falcon-7b-instruct-with-patch'
   - 'mosaicml/mpt-7b-chat'
   - 'liuhaotian/llava-v1.5-7b'