LLM: using html to visualize the perf result for Arc (#9228)

* LLM: using html to visualize the perf result for Arc

* deploy the html file

* add python license

* reslove some comments

.github/workflows/llm_performance_tests.yml

@@ -127,5 +127,8 @@ jobs:
           cd python/llm/dev/benchmark/all-in-one
           export http_proxy=${HTTP_PROXY}
           export https_proxy=${HTTPS_PROXY}
-          taskset -c 0-$((THREAD_NUM - 1)) python run.py
+          python run.py
           curl -T ./*.csv ${LLM_FTP_URL}/llm/ggml-actions/perf/
+          cd ../../../test/benchmark
+          python csv_to_html.py -f ../../dev/benchmark/all-in-one
+          cp ./*.html /mnt/disk1/nightly_perf/

python/llm/dev/benchmark/all-in-one/run.py

@@ -59,9 +59,9 @@ def run_model(repo_id, test_api, in_out_pairs, local_model_hub=None, warm_up=1,
     for in_out_pair in in_out_pairs:
         if result:
             results.append([repo_id,
-                            np.mean(result[in_out_pair], axis=0)[0],
+                            round(np.mean(result[in_out_pair], axis=0)[0]*1000.0, 2),
-                            np.mean(result[in_out_pair], axis=0)[1],
+                            round(np.mean(result[in_out_pair], axis=0)[1]*1000.0, 2),
-                            np.mean(result[in_out_pair], axis=0)[2],
+                            round(np.mean(result[in_out_pair], axis=0)[2]*1000.0, 2),
                             in_out_pair,
                             f'{int(np.mean(result[in_out_pair], axis=0)[3])}' +
                             f'-{int(np.mean(result[in_out_pair], axis=0)[4])}',
@@ -545,7 +545,7 @@ if __name__ == '__main__':
     for api in conf.test_api:
         for model in conf.repo_id:
             run_model(model, api, conf['in_out_pairs'], conf['local_model_hub'], conf['warm_up'], conf['num_trials'], conf['num_beams'])
-        df = pd.DataFrame(results, columns=['model', '1st token avg latency (s)', '2+ avg latency (s/token)', 'encoder time (s)',
+        df = pd.DataFrame(results, columns=['model', '1st token avg latency (ms)', '2+ avg latency (ms/token)', 'encoder time (ms)',
                                             'input/output tokens', 'actual input/output tokens', 'num_beams'])
         df.to_csv(f'{current_dir}/{api}-results-{today}.csv')
         results = []

python/llm/test/benchmark/csv_to_html.py

@@ -0,0 +1,39 @@
+#
+# Copyright 2016 The BigDL Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+# Python program to convert CSV to HTML Table
+
+import os
+import sys
+import argparse
+import pandas as pd
+
+def main():
+    parser = argparse.ArgumentParser(description="convert .csv file to .html file")
+    parser.add_argument("-f", "--folder_path", type=str, dest="folder_path",
+                        help="The directory which stores the .csv file", default="../../dev/benchmark/all-in-one")
+    args = parser.parse_args()
+
+    csv_files = []
+    for file_name in os.listdir(args.folder_path):
+        file_path = os.path.join(args.folder_path, file_name)
+        if os.path.isfile(file_path) and file_name.endswith(".csv"):
+            csv_files.append(file_path)
+
+    a = pd.read_csv(csv_files[0], index_col=0).to_html(csv_files[0].split("/")[-1].split(".")[0]+".html")
+
+if __name__ == "__main__":
+    sys.exit(main())