update nightly spr perf test (#10178)

* update nightly spr perf test

* update

* update runner lable

* update

* update

* update folder

* revert

.github/workflows/llm_performance_tests.yml

@@ -164,7 +164,7 @@ jobs:
       fail-fast: false
       matrix:
         python-version: ["3.9"]
-    runs-on: [self-hosted, llm, spr-perf]
+    runs-on: [self-hosted, llm, spr01-perf]
     env:
       OMP_NUM_THREADS: 16
       THREAD_NUM: 16
@@ -185,6 +185,8 @@ jobs:
           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 BigDL-LLM from Pypi" part for PR tests
       # - name: Download llm binary
@@ -213,13 +215,19 @@ jobs:
           cd python/llm/dev/benchmark/all-in-one
           export http_proxy=${HTTP_PROXY}
           export https_proxy=${HTTPS_PROXY}
+          source bigdl-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/nightly_perf_cpu/
+          cp ./*.csv /models/nightly_perf_cpu
           cd ../../../test/benchmark
           python -m pip install pandas==1.5.3
-          python csv_to_html.py -f /mnt/disk1/nightly_perf_cpu/
+          python csv_to_html.py -f /models/nightly_perf_cpu
+          cd /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

python/llm/test/benchmark/cpu-perf-test.yaml

@@ -1,6 +1,12 @@
 repo_id:
   - 'meta-llama/Llama-2-7b-chat-hf'
-local_model_hub: '/mnt/disk1/models'
+  - 'meta-llama/Llama-2-13b-chat-hf'
+  - 'THUDM/chatglm2-6b'
+  - 'THUDM/chatglm3-6b'
+  - 'baichuan-inc/Baichuan2-7B-Chat'
+  - 'baichuan-inc/Baichuan2-13B-Chat'
+  - 'Qwen/Qwen-14B-Chat'
+local_model_hub: '/models'
 warm_up: 1
 num_trials: 3
 num_beams: 1 # default to greedy search
@@ -9,6 +15,7 @@ batch_size: 1 # default to 1
 in_out_pairs:
   - '32-32'
   - '1024-128'
+  - '2048-256'
 test_api:
   - "transformer_int4"
   # - "native_int4"