chore: import upstream snapshot with attribution

runtime/python/grpc/Readme.md

@@ -0,0 +1,27 @@
+# GRPC python Client for 2pass decoding
+The client can send streaming or full audio data to server as you wish, and get transcribed text once the server respond (depends on mode)
+
+In the demo client, audio_chunk_duration is set to 1000ms, and send_interval is set to 100ms
+
+### 1. Install the requirements
+```shell
+git clone https://github.com/alibaba/FunASR.git && cd FunASR/funasr/runtime/python/grpc
+pip install -r requirements.txt
+```
+
+### 2. Generate protobuf file
+```shell
+# paraformer_pb2.py and paraformer_pb2_grpc.py are already generated, 
+# regenerate it only when you make changes to ./proto/paraformer.proto file.
+python -m grpc_tools.protoc --proto_path=./proto -I ./proto --python_out=. --grpc_python_out=./ ./proto/paraformer.proto
+```
+
+### 3. Start grpc client
+```
+# Start client.
+python grpc_main_client.py --host 127.0.0.1 --port 10100 --wav_path /path/to/your_test_wav.wav
+```
+
+## Acknowledge
+1. This project is maintained by [FunASR community](https://github.com/modelscope/FunASR).
+2. We acknowledge burkliu (刘柏基, liubaiji@xverse.cn) for contributing the grpc service.

runtime/python/grpc/grpc_main_client.py

@@ -0,0 +1,83 @@
+"""
+  Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights
+  Reserved. MIT License  (https://opensource.org/licenses/MIT)
+  2023 by burkliu(刘柏基) liubaiji@xverse.cn
+"""
+
+import logging
+import argparse
+import soundfile as sf
+import time
+
+import grpc
+import paraformer_pb2_grpc
+from paraformer_pb2 import Request, WavFormat, DecodeMode
+
+
+class GrpcClient:
+    def __init__(self, wav_path, uri, mode):
+        self.wav, self.sampling_rate = sf.read(wav_path, dtype="int16")
+        self.wav_format = WavFormat.pcm
+        self.audio_chunk_duration = 1000  # ms
+        self.audio_chunk_size = int(self.sampling_rate * self.audio_chunk_duration / 1000)
+        self.send_interval = 100  # ms
+        self.mode = mode
+
+        # connect to grpc server
+        channel = grpc.insecure_channel(uri)
+        self.stub = paraformer_pb2_grpc.ASRStub(channel)
+
+        # start request
+        for respond in self.stub.Recognize(self.request_iterator()):
+            logging.info(
+                "[receive] mode {}, text {}, is final {}".format(
+                    DecodeMode.Name(respond.mode), respond.text, respond.is_final
+                )
+            )
+
+    def request_iterator(self, mode=DecodeMode.two_pass):
+        is_first_pack = True
+        is_final = False
+        for start in range(0, len(self.wav), self.audio_chunk_size):
+            request = Request()
+            audio_chunk = self.wav[start : start + self.audio_chunk_size]
+
+            if is_first_pack:
+                is_first_pack = False
+                request.sampling_rate = self.sampling_rate
+                request.mode = self.mode
+                request.wav_format = self.wav_format
+                if request.mode == DecodeMode.two_pass or request.mode == DecodeMode.online:
+                    request.chunk_size.extend([5, 10, 5])
+
+            if start + self.audio_chunk_size >= len(self.wav):
+                is_final = True
+            request.is_final = is_final
+            request.audio_data = audio_chunk.tobytes()
+            logging.info(
+                "[request] audio_data len {}, is final {}".format(
+                    len(request.audio_data), request.is_final
+                )
+            )  # int16 = 2bytes
+            time.sleep(self.send_interval / 1000)
+            yield request
+
+
+if __name__ == "__main__":
+    logging.basicConfig(filename="", format="%(asctime)s %(message)s", level=logging.INFO)
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--host", type=str, default="127.0.0.1", required=False, help="grpc server host ip"
+    )
+    parser.add_argument("--port", type=int, default=10100, required=False, help="grpc server port")
+    parser.add_argument("--wav_path", type=str, required=True, help="audio wav path")
+    args = parser.parse_args()
+
+    for mode in [DecodeMode.offline, DecodeMode.online, DecodeMode.two_pass]:
+        mode_name = DecodeMode.Name(mode)
+        logging.info("[request] start requesting with mode {}".format(mode_name))
+
+        st = time.time()
+        uri = "{}:{}".format(args.host, args.port)
+        client = GrpcClient(args.wav_path, uri, mode)
+        logging.info("mode {}, time pass: {}".format(mode_name, time.time() - st))

runtime/python/grpc/proto/Readme.md

@@ -0,0 +1,25 @@
+```	
+service ASR { //grpc service
+  rpc Recognize (stream Request) returns (stream Response) {} //Stub
+}
+
+message Request { //request data
+  bytes audio_data = 1; //audio data in bytes.
+  string user = 2; //user allowed.
+  string language = 3; //language, zh-CN for now.
+  bool speaking = 4; //flag for speaking. 
+  bool isEnd = 5; //flag for end. set isEnd to true when you stop asr:
+  //vad:is_speech then speaking=True & isEnd = False, audio data will be appended for the specfied user.
+  //vad:silence then speaking=False & isEnd = False, clear audio buffer and do asr inference.
+}
+
+message Response { //response data.
+  string sentence = 1; //json, includes flag for success and asr text .
+  string user = 2; //same to request user.
+  string language = 3; //same to request language.
+  string action = 4; //server status:
+  //terminate：asr stopped; 
+  //speaking：user is speaking, audio data is appended; 
+  //decoding: server is decoding; 
+  //finish: get asr text, most used.
+}

runtime/python/grpc/proto/paraformer.proto

@@ -0,0 +1,39 @@
+// Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights
+// Reserved. MIT License  (https://opensource.org/licenses/MIT)
+//
+// 2023 by burkliu(刘柏基) liubaiji@xverse.cn
+
+syntax = "proto3";
+
+option objc_class_prefix = "paraformer";
+
+package paraformer;
+
+service ASR {
+  rpc Recognize (stream Request) returns (stream Response) {}
+}
+
+enum WavFormat {
+  pcm = 0;
+}
+
+enum DecodeMode {
+  offline = 0;
+  online = 1;
+  two_pass = 2;
+}
+
+message Request {
+  DecodeMode mode = 1;
+  WavFormat wav_format = 2;
+  int32 sampling_rate = 3;
+  repeated int32 chunk_size = 4;
+  bool is_final = 5;
+  bytes audio_data = 6;
+}
+
+message Response {
+  DecodeMode mode = 1;
+  string text = 2;
+  bool is_final = 3;
+}

runtime/python/grpc/requirements.txt

@@ -0,0 +1,2 @@
+grpcio
+grpcio-tools