Add windows serving support for faq finance (PaddlePaddle#3491)

* Add windows serving support for faq finance

* Remove unused comments

applications/question_answering/faq_finance/README.md

@@ -443,17 +443,10 @@ sh scripts/export_to_serving.sh
 ```
 
 启动 Pipeline Server:
-修改Tokenizer：
-
-```
-self.tokenizer = AutoTokenizer.from_pretrained('rocketqa-zh-base-query-encoder')
-
-```
-然后运行：
 
 ```
 cd deploy/python/
-python web_service.py
+python web_service.py --model_name_or_path rocketqa-zh-base-query-encoder
 ```
 
 启动客户端调用 Server, 使用 POST的方式：
@@ -479,6 +472,22 @@ list_data = [
 python rpc_client.py
 ```
 
+对于Windows用户，启动下面的Pipeline Server:
+
+```
+python web_service_windows.py --model_name_or_path rocketqa-zh-base-query-encoder
+```
+
+启动客户端调用 Server, 使用 POST的方式(Windows不支持RPC的调用方式)，首先修改http_client.py中需要预测的样本：
+
+```
+data = {"feed": ["买了社保，是不是就不用买商业保险了？"], "fetch": ["output_embedding"]}
+```
+然后运行：
+```
+python http_client.py
+```
+
 ### 4.5 问答系统整个流程
 
 问答系统使用了Client Server的模式，即抽取向量的模型部署在服务端，然后启动客户端（Client）端去访问。

applications/question_answering/faq_finance/deploy/python/http_client.py

@@ -0,0 +1,31 @@
+# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+#
+# 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.
+
+import json
+import base64
+import time
+import numpy as np
+import requests
+
+headers = {"Content-type": "application/json"}
+url = "http://10.21.226.175:8080/ernie/prediction"  # XXX取决于服务端YourService的初始化name参数
+
+data = {"feed": ["买了社保，是不是就不用买商业保险了？"], "fetch": ["output_embedding"]}
+data = json.dumps(data)
+print(data)
+r = requests.post(url=url, headers=headers, data=data)
+print(r.json())
+json_data = r.json()
+data = np.array(json_data['result']['output_embedding'])
+print(data.shape)

applications/question_answering/faq_finance/deploy/python/web_service.py

@@ -14,9 +14,16 @@
 
 import numpy as np
 import sys
+import argparse
 
 from paddle_serving_server.web_service import WebService, Op
 
+# yapf: disable
+parser = argparse.ArgumentParser()
+parser.add_argument('--model_name_or_path', default="rocketqa-zh-base-query-encoder", help="Select tokenizer name to for model")
+args = parser.parse_args()
+# yapf: enable
+
 
 def convert_example(example,
                     tokenizer,
@@ -37,8 +44,7 @@ class ErnieOp(Op):
 
     def init_op(self):
         from paddlenlp.transformers import AutoTokenizer
-        self.tokenizer = AutoTokenizer.from_pretrained(
-            'rocketqa-zh-base-query-encoder')
+        self.tokenizer = AutoTokenizer.from_pretrained(args.model_name_or_path)
 
     def preprocess(self, input_dicts, data_id, log_id):
         from paddlenlp.data import Stack, Tuple, Pad
@@ -54,7 +60,7 @@ def preprocess(self, input_dicts, data_id, log_id):
         batchify_fn = lambda samples, fn=Tuple(
             Pad(axis=0, pad_val=self.tokenizer.pad_token_id, dtype="int64"
                 ),  # input
-            Pad(axis=0, pad_val=self.tokenizer.pad_token_id, dtype="int64"
+            Pad(axis=0, pad_val=self.tokenizer.pad_token_type_id, dtype="int64"
                 ),  # segment
         ): fn(samples)
         input_ids, segment_ids = batchify_fn(examples)
@@ -77,6 +83,7 @@ def get_pipeline_response(self, read_op):
         return ernie_op
 
 
-ernie_service = ErnieService(name="ernie")
-ernie_service.prepare_pipeline_config("config_nlp.yml")
-ernie_service.run_service()
+if __name__ == "__main__":
+    ernie_service = ErnieService(name="ernie")
+    ernie_service.prepare_pipeline_config("config_nlp.yml")
+    ernie_service.run_service()

applications/question_answering/faq_finance/deploy/python/web_service_windows.py

@@ -0,0 +1,82 @@
+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# 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.
+
+import numpy as np
+import sys
+import argparse
+
+from paddle_serving_server.web_service import WebService, Op
+
+# yapf: disable
+parser = argparse.ArgumentParser()
+parser.add_argument('--model_name_or_path', default="rocketqa-zh-base-query-encoder", help="Select tokenizer name to for model")
+args = parser.parse_args()
+# yapf: enable
+
+
+def convert_example(example,
+                    tokenizer,
+                    max_seq_length=512,
+                    pad_to_max_seq_len=False):
+    result = []
+    for text in example:
+        encoded_inputs = tokenizer(text=text,
+                                   max_seq_len=max_seq_length,
+                                   pad_to_max_seq_len=pad_to_max_seq_len)
+        input_ids = encoded_inputs["input_ids"]
+        token_type_ids = encoded_inputs["token_type_ids"]
+        result += [input_ids, token_type_ids]
+    return result
+
+
+class ErnieService(WebService):
+
+    def init_service(self):
+        from paddlenlp.transformers import AutoTokenizer
+        self.tokenizer = AutoTokenizer.from_pretrained(args.model_name_or_path)
+
+    def preprocess(self, feed=[], fetch=[]):
+        from paddlenlp.data import Stack, Tuple, Pad
+        print("input dict", feed)
+        batch_size = len(feed)
+        is_batch = True
+        examples = []
+        for i in range(batch_size):
+            input_ids, segment_ids = convert_example([feed[i]], self.tokenizer)
+            examples.append((input_ids, segment_ids))
+        batchify_fn = lambda samples, fn=Tuple(
+            Pad(axis=0, pad_val=self.tokenizer.pad_token_id, dtype="int64"
+                ),  # input
+            Pad(axis=0, pad_val=self.tokenizer.pad_token_type_id, dtype="int64"
+                ),  # segment
+        ): fn(samples)
+        input_ids, segment_ids = batchify_fn(examples)
+        feed_dict = {}
+        feed_dict['input_ids'] = input_ids
+        feed_dict['token_type_ids'] = segment_ids
+        return feed_dict, fetch, is_batch
+
+    def postprocess(self, feed=[], fetch=[], fetch_map=None):
+        for key in fetch_map:
+            fetch_map[key] = fetch_map[key].tolist()
+        return fetch_map
+
+
+if __name__ == "__main__":
+    ernie_service = ErnieService(name="ernie")
+    ernie_service.load_model_config("../../serving_server")
+    ernie_service.prepare_server(workdir="workdir", port=8080)
+    ernie_service.init_service()
+    ernie_service.run_debugger_service()
+    ernie_service.run_web_service()