update server cli, test=doc

demos/speech_server/conf/application.yaml

@@ -9,9 +9,17 @@ port: 8090
 ##################################################################
 #                     CONFIG FILE                                #
 ##################################################################
-# add engine type (Options: asr, tts) and config file here.
+# The engine_type of speech task needs to keep the same type as the config file of speech task.
+# E.g: The engine_type of asr is 'python', the engine_backend of asr is 'XX/asr.yaml'
+# E.g: The engine_type of asr is 'inference', the engine_backend of asr is 'XX/asr_pd.yaml'
+#
+# add engine type (Options: python, inference) 
+engine_type:
+    asr: 'inference'
+    tts: 'inference'
 
+# add engine backend type (Options: asr, tts) and config file here.
+# Adding a speech task to engine_backend means starting the service.
 engine_backend:
-    asr: 'conf/asr/asr.yaml'
-    tts: 'conf/tts/tts.yaml'
-
+    asr: 'conf/asr/asr_pd.yaml'
+    tts: 'conf/tts/tts_pd.yaml'

demos/speech_server/conf/asr/asr.yaml

@@ -1,7 +1,8 @@
 model: 'conformer_wenetspeech'
 lang: 'zh'
 sample_rate: 16000
-cfg_path: 
-ckpt_path: 
+cfg_path: # [optional]
+ckpt_path: # [optional]
 decode_method: 'attention_rescoring'
-force_yes: False
+force_yes: True
+device: 'gpu:3'  # set 'gpu:id' or 'cpu'

demos/speech_server/conf/asr/asr_pd.yaml

@@ -0,0 +1,25 @@
+# This is the parameter configuration file for ASR server.
+# These are the static models that support paddle inference.
+
+##################################################################
+#                  ACOUSTIC MODEL SETTING                        #
+# am choices=['deepspeech2offline_aishell'] TODO
+##################################################################
+model_type: 'deepspeech2offline_aishell'
+am_model: # the pdmodel file of am static model [optional]
+am_params:  # the pdiparams file of am static model [optional]
+lang: 'zh'
+sample_rate: 16000
+cfg_path: 
+decode_method: 
+force_yes: True
+
+am_predictor_conf:
+  device: 'gpu:3'  # set 'gpu:id' or 'cpu'
+  enable_mkldnn: True
+  switch_ir_optim: True
+
+
+##################################################################
+#                            OTHERS                              #
+##################################################################

demos/speech_server/conf/tts/tts.yaml

@@ -29,4 +29,4 @@ voc_stat:
 #                            OTHERS                              #
 ##################################################################
 lang: 'zh'
-device: 'gpu:2'
+device: 'gpu:3'  # set 'gpu:id' or 'cpu'

demos/speech_server/conf/tts/tts_pd.yaml

@@ -6,8 +6,8 @@
 # am choices=['speedyspeech_csmsc', 'fastspeech2_csmsc']
 ##################################################################
 am: 'fastspeech2_csmsc'   
-am_model: # the pdmodel file of am static model
-am_params: # the pdiparams file of am static model
+am_model: # the pdmodel file of your am static model (XX.pdmodel)
+am_params: # the pdiparams file of your am static model (XX.pdipparams)
 am_sample_rate: 24000
 phones_dict: 
 tones_dict: 
@@ -15,9 +15,9 @@ speaker_dict:
 spk_id: 0
 
 am_predictor_conf:
-  use_gpu: True
-  enable_mkldnn: True
-  switch_ir_optim: True
+  device: 'gpu:3'  # set 'gpu:id' or 'cpu'
+  enable_mkldnn: False
+  switch_ir_optim: False
 
 
 ##################################################################
@@ -25,17 +25,16 @@ am_predictor_conf:
 # voc choices=['pwgan_csmsc', 'mb_melgan_csmsc','hifigan_csmsc']
 ##################################################################
 voc: 'pwgan_csmsc'
-voc_model: # the pdmodel file of vocoder static model
-voc_params: # the pdiparams file of vocoder static model 
+voc_model: # the pdmodel file of your vocoder static model (XX.pdmodel)
+voc_params: # the pdiparams file of your vocoder static model (XX.pdipparams)
 voc_sample_rate: 24000
 
 voc_predictor_conf:
-  use_gpu: True
-  enable_mkldnn: True  
-  switch_ir_optim: True  
+  device: 'gpu:3'  # set 'gpu:id' or 'cpu'
+  enable_mkldnn: False  
+  switch_ir_optim: False  
 
 ##################################################################
 #                            OTHERS                              #
 ##################################################################
 lang: 'zh'
-device: paddle.get_device()

paddlespeech/server/bin/paddlespeech_server.py

@@ -20,7 +20,7 @@ from fastapi import FastAPI
 from ..executor import BaseExecutor
 from ..util import cli_server_register
 from ..util import stats_wrapper
-from paddlespeech.server.engine.engine_factory import EngineFactory
+from paddlespeech.server.engine.engine_pool import init_engine_pool
 from paddlespeech.server.restful.api import setup_router
 from paddlespeech.server.utils.config import get_config
 
@@ -51,8 +51,10 @@ class ServerExecutor(BaseExecutor):
 
     def init(self, config) -> bool:
         """system initialization
+
         Args:
             config (CfgNode): config object
+
         Returns:
             bool: 
         """
@@ -61,13 +63,8 @@ class ServerExecutor(BaseExecutor):
         api_router = setup_router(api_list)
         app.include_router(api_router)
 
-        # init engine
-        engine_pool = []
-        for engine in config.engine_backend:
-            engine_pool.append(EngineFactory.get_engine(engine_name=engine))
-            if not engine_pool[-1].init(
-                    config_file=config.engine_backend[engine]):
-                return False
+        if not init_engine_pool(config):
+            return False
 
         return True
 

paddlespeech/server/conf/application.yaml

@@ -9,12 +9,17 @@ port: 8090
 ##################################################################
 #                     CONFIG FILE                                #
 ##################################################################
+# The engine_type of speech task needs to keep the same type as the config file of speech task.
+# E.g: The engine_type of asr is 'python', the engine_backend of asr is 'XX/asr.yaml'
+# E.g: The engine_type of asr is 'inference', the engine_backend of asr is 'XX/asr_pd.yaml'
+#
 # add engine type (Options: python, inference) 
 engine_type:
-    asr: 'inference'
-    # tts: 'inference'
+    asr: 'python'
+    tts: 'python'
 
 # add engine backend type (Options: asr, tts) and config file here.
+# Adding a speech task to engine_backend means starting the service.
 engine_backend:
-    asr: 'conf/asr/asr_pd.yaml'
-    #tts: 'conf/tts/tts_pd.yaml'
+    asr: 'conf/asr/asr.yaml'
+    tts: 'conf/tts/tts.yaml'

paddlespeech/server/conf/asr/asr.yaml

@@ -5,3 +5,4 @@ cfg_path: # [optional]
 ckpt_path: # [optional]
 decode_method: 'attention_rescoring'
 force_yes: True
+device: 'gpu:3'  # set 'gpu:id' or 'cpu'

paddlespeech/server/conf/asr/asr_pd.yaml

@@ -15,7 +15,7 @@ decode_method:
 force_yes: True
 
 am_predictor_conf:
-  use_gpu: True
+  device: 'gpu:3'  # set 'gpu:id' or 'cpu'
   enable_mkldnn: True
   switch_ir_optim: True
 

paddlespeech/server/conf/tts/tts.yaml

@@ -29,4 +29,4 @@ voc_stat:
 #                            OTHERS                              #
 ##################################################################
 lang: 'zh'
-device: paddle.get_device()
+device: 'gpu:3'  # set 'gpu:id' or 'cpu'

paddlespeech/server/conf/tts/tts_pd.yaml

@@ -6,18 +6,18 @@
 # am choices=['speedyspeech_csmsc', 'fastspeech2_csmsc']
 ##################################################################
 am: 'fastspeech2_csmsc'   
-am_model: # the pdmodel file of am static model
-am_params: # the pdiparams file of am static model
-am_sample_rate: 24000
+am_model: # the pdmodel file of your am static model (XX.pdmodel)
+am_params: # the pdiparams file of your am static model (XX.pdipparams)
+am_sample_rate: 24000    # must match the model
 phones_dict: 
 tones_dict: 
 speaker_dict: 
 spk_id: 0
 
 am_predictor_conf:
-  use_gpu: True
-  enable_mkldnn: True
-  switch_ir_optim: True
+  device: 'gpu:3'  # set 'gpu:id' or 'cpu'
+  enable_mkldnn: False
+  switch_ir_optim: False
 
 
 ##################################################################
@@ -25,17 +25,16 @@ am_predictor_conf:
 # voc choices=['pwgan_csmsc', 'mb_melgan_csmsc','hifigan_csmsc']
 ##################################################################
 voc: 'pwgan_csmsc'
-voc_model: # the pdmodel file of vocoder static model
-voc_params: # the pdiparams file of vocoder static model 
-voc_sample_rate: 24000
+voc_model: # the pdmodel file of your vocoder static model (XX.pdmodel)
+voc_params: # the pdiparams file of your vocoder static model (XX.pdipparams)
+voc_sample_rate: 24000    #must match the model
 
 voc_predictor_conf:
-  use_gpu: True
-  enable_mkldnn: True  
-  switch_ir_optim: True  
+  device: 'gpu:3'   # set 'gpu:id' or 'cpu'
+  enable_mkldnn: False
+  switch_ir_optim: False  
 
 ##################################################################
 #                            OTHERS                              #
 ##################################################################
 lang: 'zh'
-device: paddle.get_device()

paddlespeech/server/engine/asr/python/asr_engine.py

@@ -12,21 +12,11 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import io
-import os
-from typing import List
-from typing import Optional
-from typing import Union
 
-import librosa
 import paddle
-import soundfile
 
 from paddlespeech.cli.asr.infer import ASRExecutor
 from paddlespeech.cli.log import logger
-from paddlespeech.s2t.frontend.featurizer.text_featurizer import TextFeaturizer
-from paddlespeech.s2t.transform.transformation import Transformation
-from paddlespeech.s2t.utils.dynamic_import import dynamic_import
-from paddlespeech.s2t.utils.utility import UpdateConfig
 from paddlespeech.server.engine.base_engine import BaseEngine
 from paddlespeech.server.utils.config import get_config
 
@@ -63,7 +53,7 @@ class ASREngine(BaseEngine):
         self.executor = ASRServerExecutor()
 
         self.config = get_config(config_file)
-        paddle.set_device(paddle.get_device())
+        paddle.set_device(self.config.device)
         self.executor._init_from_path(
             self.config.model, self.config.lang, self.config.sample_rate,
             self.config.cfg_path, self.config.decode_method,

paddlespeech/server/engine/tts/paddleinference/tts_engine.py

@@ -344,7 +344,6 @@ class TTSEngine(BaseEngine):
 
         try:
             self.config = get_config(config_file)
-
             self.executor._init_from_path(
                 am=self.config.am,
                 am_model=self.config.am_model,

paddlespeech/server/restful/tts_api.py

@@ -16,7 +16,7 @@ from typing import Union
 
 from fastapi import APIRouter
 
-from paddlespeech.server.engine.tts.paddleinference.tts_engine import TTSEngine
+from paddlespeech.server.engine.engine_pool import get_engine_pool
 from paddlespeech.server.restful.request import TTSRequest
 from paddlespeech.server.restful.response import ErrorResponse
 from paddlespeech.server.restful.response import TTSResponse
@@ -60,28 +60,41 @@ def tts(request_body: TTSRequest):
     Returns:
         json: [description]
     """
-    # json to dict 
-    item_dict = request_body.dict()
-    sentence = item_dict['text']
-    spk_id = item_dict['spk_id']
-    speed = item_dict['speed']
-    volume = item_dict['volume']
-    sample_rate = item_dict['sample_rate']
-    save_path = item_dict['save_path']
+    # get params
+    text = request_body.text
+    spk_id = request_body.spk_id
+    speed = request_body.speed
+    volume = request_body.volume
+    sample_rate = request_body.sample_rate
+    save_path = request_body.save_path
 
     # Check parameters
-    if speed <=0 or speed > 3 or volume <=0 or volume > 3 or \
-        sample_rate not in [0, 16000, 8000] or \
-        (save_path is not None and not save_path.endswith("pcm") and not save_path.endswith("wav")):
-        return failed_response(ErrorCode.SERVER_PARAM_ERR)
-
-    # single
-    tts_engine = TTSEngine()
+    if speed <= 0 or speed > 3:
+        return failed_response(
+            ErrorCode.SERVER_PARAM_ERR,
+            "invalid speed value, the value should be between 0 and 3.")
+    if volume <= 0 or volume > 3:
+        return failed_response(
+            ErrorCode.SERVER_PARAM_ERR,
+            "invalid volume value, the value should be between 0 and 3.")
+    if sample_rate not in [0, 16000, 8000]:
+        return failed_response(
+            ErrorCode.SERVER_PARAM_ERR,
+            "invalid sample_rate value, the choice of value is 0, 8000, 16000.")
+    if save_path is not None and not save_path.endswith(
+            "pcm") and not save_path.endswith("wav"):
+        return failed_response(
+            ErrorCode.SERVER_PARAM_ERR,
+            "invalid save_path, saved audio formats support pcm and wav")
 
     # run
     try:
+        # get single engine from engine pool
+        engine_pool = get_engine_pool()
+        tts_engine = engine_pool['tts']
+
         lang, target_sample_rate, wav_base64 = tts_engine.run(
-            sentence, spk_id, speed, volume, sample_rate, save_path)
+            text, spk_id, speed, volume, sample_rate, save_path)
 
         response = {
             "success": True,

paddlespeech/server/utils/paddle_predictor.py

@@ -41,8 +41,9 @@ def init_predictor(model_dir: Optional[os.PathLike]=None,
         config = Config(model_file, params_file)
 
     config.enable_memory_optim()
-    if predictor_conf["use_gpu"]:
-        config.enable_use_gpu(1000, 0)
+    if "gpu" in predictor_conf["device"]:
+        gpu_id = predictor_conf["device"].split(":")[-1]
+        config.enable_use_gpu(1000, int(gpu_id))
     if predictor_conf["enable_mkldnn"]:
         config.enable_mkldnn()
     if predictor_conf["switch_ir_optim"]: