update cli, test=doc

paddlespeech/cli/asr/infer.py

@@ -29,9 +29,10 @@ from ..download import get_path_from_url
 from ..executor import BaseExecutor
 from ..log import logger
 from ..utils import cli_register
-from ..utils import download_and_decompress
 from ..utils import MODEL_HOME
 from ..utils import stats_wrapper
+from .pretrained_models import model_alias
+from .pretrained_models import pretrained_models
 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
@@ -39,94 +40,14 @@ from paddlespeech.s2t.utils.utility import UpdateConfig
 
 __all__ = ['ASRExecutor']
 
-pretrained_models = {
-    # The tags for pretrained_models should be "{model_name}[_{dataset}][-{lang}][-...]".
-    # e.g. "conformer_wenetspeech-zh-16k" and "panns_cnn6-32k".
-    # Command line and python api use "{model_name}[_{dataset}]" as --model, usage:
-    # "paddlespeech asr --model conformer_wenetspeech --lang zh --sr 16000 --input ./input.wav"
-    "conformer_wenetspeech-zh-16k": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/s2t/wenetspeech/asr1_conformer_wenetspeech_ckpt_0.1.1.model.tar.gz',
-        'md5':
-        '76cb19ed857e6623856b7cd7ebbfeda4',
-        'cfg_path':
-        'model.yaml',
-        'ckpt_path':
-        'exp/conformer/checkpoints/wenetspeech',
-    },
-    "transformer_librispeech-en-16k": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/s2t/librispeech/asr1/asr1_transformer_librispeech_ckpt_0.1.1.model.tar.gz',
-        'md5':
-        '2c667da24922aad391eacafe37bc1660',
-        'cfg_path':
-        'model.yaml',
-        'ckpt_path':
-        'exp/transformer/checkpoints/avg_10',
-    },
-    "deepspeech2offline_aishell-zh-16k": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/s2t/aishell/asr0/asr0_deepspeech2_aishell_ckpt_0.1.1.model.tar.gz',
-        'md5':
-        '932c3593d62fe5c741b59b31318aa314',
-        'cfg_path':
-        'model.yaml',
-        'ckpt_path':
-        'exp/deepspeech2/checkpoints/avg_1',
-        'lm_url':
-        'https://deepspeech.bj.bcebos.com/zh_lm/zh_giga.no_cna_cmn.prune01244.klm',
-        'lm_md5':
-        '29e02312deb2e59b3c8686c7966d4fe3'
-    },
-    "deepspeech2online_aishell-zh-16k": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/s2t/aishell/asr0/asr0_deepspeech2_online_aishell_ckpt_0.2.0.model.tar.gz',
-        'md5':
-        '23e16c69730a1cb5d735c98c83c21e16',
-        'cfg_path':
-        'model.yaml',
-        'ckpt_path':
-        'exp/deepspeech2_online/checkpoints/avg_1',
-        'lm_url':
-        'https://deepspeech.bj.bcebos.com/zh_lm/zh_giga.no_cna_cmn.prune01244.klm',
-        'lm_md5':
-        '29e02312deb2e59b3c8686c7966d4fe3'
-    },
-    "deepspeech2offline_librispeech-en-16k": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/s2t/librispeech/asr0/asr0_deepspeech2_librispeech_ckpt_0.1.1.model.tar.gz',
-        'md5':
-        'f5666c81ad015c8de03aac2bc92e5762',
-        'cfg_path':
-        'model.yaml',
-        'ckpt_path':
-        'exp/deepspeech2/checkpoints/avg_1',
-        'lm_url':
-        'https://deepspeech.bj.bcebos.com/en_lm/common_crawl_00.prune01111.trie.klm',
-        'lm_md5':
-        '099a601759d467cd0a8523ff939819c5'
-    },
-}
-
-model_alias = {
-    "deepspeech2offline":
-    "paddlespeech.s2t.models.ds2:DeepSpeech2Model",
-    "deepspeech2online":
-    "paddlespeech.s2t.models.ds2_online:DeepSpeech2ModelOnline",
-    "conformer":
-    "paddlespeech.s2t.models.u2:U2Model",
-    "transformer":
-    "paddlespeech.s2t.models.u2:U2Model",
-    "wenetspeech":
-    "paddlespeech.s2t.models.u2:U2Model",
-}
-
 
 @cli_register(
     name='paddlespeech.asr', description='Speech to text infer command.')
 class ASRExecutor(BaseExecutor):
     def __init__(self):
-        super(ASRExecutor, self).__init__()
+        super().__init__()
+        self.model_alias = model_alias
+        self.pretrained_models = pretrained_models
 
         self.parser = argparse.ArgumentParser(
             prog='paddlespeech.asr', add_help=True)
@@ -136,7 +57,9 @@ class ASRExecutor(BaseExecutor):
             '--model',
             type=str,
             default='conformer_wenetspeech',
-            choices=[tag[:tag.index('-')] for tag in pretrained_models.keys()],
+            choices=[
+                tag[:tag.index('-')] for tag in self.pretrained_models.keys()
+            ],
             help='Choose model type of asr task.')
         self.parser.add_argument(
             '--lang',
@@ -192,23 +115,6 @@ class ASRExecutor(BaseExecutor):
             action='store_true',
             help='Increase logger verbosity of current task.')
 
-    def _get_pretrained_path(self, tag: str) -> os.PathLike:
-        """
-        Download and returns pretrained resources path of current task.
-        """
-        support_models = list(pretrained_models.keys())
-        assert tag in pretrained_models, 'The model "{}" you want to use has not been supported, please choose other models.\nThe support models includes:\n\t\t{}\n'.format(
-            tag, '\n\t\t'.join(support_models))
-
-        res_path = os.path.join(MODEL_HOME, tag)
-        decompressed_path = download_and_decompress(pretrained_models[tag],
-                                                    res_path)
-        decompressed_path = os.path.abspath(decompressed_path)
-        logger.info(
-            'Use pretrained model stored in: {}'.format(decompressed_path))
-
-        return decompressed_path
-
     def _init_from_path(self,
                         model_type: str='wenetspeech',
                         lang: str='zh',
@@ -228,10 +134,11 @@ class ASRExecutor(BaseExecutor):
             tag = model_type + '-' + lang + '-' + sample_rate_str
             res_path = self._get_pretrained_path(tag)  # wenetspeech_zh
             self.res_path = res_path
-            self.cfg_path = os.path.join(res_path,
-                                         pretrained_models[tag]['cfg_path'])
+            self.cfg_path = os.path.join(
+                res_path, self.pretrained_models[tag]['cfg_path'])
             self.ckpt_path = os.path.join(
-                res_path, pretrained_models[tag]['ckpt_path'] + ".pdparams")
+                res_path,
+                self.pretrained_models[tag]['ckpt_path'] + ".pdparams")
             logger.info(res_path)
             logger.info(self.cfg_path)
             logger.info(self.ckpt_path)
@@ -255,8 +162,8 @@ class ASRExecutor(BaseExecutor):
                 self.collate_fn_test = SpeechCollator.from_config(self.config)
                 self.text_feature = TextFeaturizer(
                     unit_type=self.config.unit_type, vocab=self.vocab)
-                lm_url = pretrained_models[tag]['lm_url']
-                lm_md5 = pretrained_models[tag]['lm_md5']
+                lm_url = self.pretrained_models[tag]['lm_url']
+                lm_md5 = self.pretrained_models[tag]['lm_md5']
                 self.download_lm(
                     lm_url,
                     os.path.dirname(self.config.decode.lang_model_path), lm_md5)
@@ -274,7 +181,7 @@ class ASRExecutor(BaseExecutor):
                 raise Exception("wrong type")
         model_name = model_type[:model_type.rindex(
             '_')]  # model_type: {model_name}_{dataset}
-        model_class = dynamic_import(model_name, model_alias)
+        model_class = dynamic_import(model_name, self.model_alias)
         model_conf = self.config
         model = model_class.from_config(model_conf)
         self.model = model

paddlespeech/cli/asr/pretrained_models.py

+# 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.
+
+pretrained_models = {
+    # The tags for pretrained_models should be "{model_name}[_{dataset}][-{lang}][-...]".
+    # e.g. "conformer_wenetspeech-zh-16k" and "panns_cnn6-32k".
+    # Command line and python api use "{model_name}[_{dataset}]" as --model, usage:
+    # "paddlespeech asr --model conformer_wenetspeech --lang zh --sr 16000 --input ./input.wav"
+    "conformer_wenetspeech-zh-16k": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/s2t/wenetspeech/asr1_conformer_wenetspeech_ckpt_0.1.1.model.tar.gz',
+        'md5':
+        '76cb19ed857e6623856b7cd7ebbfeda4',
+        'cfg_path':
+        'model.yaml',
+        'ckpt_path':
+        'exp/conformer/checkpoints/wenetspeech',
+    },
+    "transformer_librispeech-en-16k": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/s2t/librispeech/asr1/asr1_transformer_librispeech_ckpt_0.1.1.model.tar.gz',
+        'md5':
+        '2c667da24922aad391eacafe37bc1660',
+        'cfg_path':
+        'model.yaml',
+        'ckpt_path':
+        'exp/transformer/checkpoints/avg_10',
+    },
+    "deepspeech2offline_aishell-zh-16k": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/s2t/aishell/asr0/asr0_deepspeech2_aishell_ckpt_0.1.1.model.tar.gz',
+        'md5':
+        '932c3593d62fe5c741b59b31318aa314',
+        'cfg_path':
+        'model.yaml',
+        'ckpt_path':
+        'exp/deepspeech2/checkpoints/avg_1',
+        'lm_url':
+        'https://deepspeech.bj.bcebos.com/zh_lm/zh_giga.no_cna_cmn.prune01244.klm',
+        'lm_md5':
+        '29e02312deb2e59b3c8686c7966d4fe3'
+    },
+    "deepspeech2online_aishell-zh-16k": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/s2t/aishell/asr0/asr0_deepspeech2_online_aishell_ckpt_0.2.0.model.tar.gz',
+        'md5':
+        '23e16c69730a1cb5d735c98c83c21e16',
+        'cfg_path':
+        'model.yaml',
+        'ckpt_path':
+        'exp/deepspeech2_online/checkpoints/avg_1',
+        'lm_url':
+        'https://deepspeech.bj.bcebos.com/zh_lm/zh_giga.no_cna_cmn.prune01244.klm',
+        'lm_md5':
+        '29e02312deb2e59b3c8686c7966d4fe3'
+    },
+    "deepspeech2offline_librispeech-en-16k": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/s2t/librispeech/asr0/asr0_deepspeech2_librispeech_ckpt_0.1.1.model.tar.gz',
+        'md5':
+        'f5666c81ad015c8de03aac2bc92e5762',
+        'cfg_path':
+        'model.yaml',
+        'ckpt_path':
+        'exp/deepspeech2/checkpoints/avg_1',
+        'lm_url':
+        'https://deepspeech.bj.bcebos.com/en_lm/common_crawl_00.prune01111.trie.klm',
+        'lm_md5':
+        '099a601759d467cd0a8523ff939819c5'
+    },
+}
+
+model_alias = {
+    "deepspeech2offline":
+    "paddlespeech.s2t.models.ds2:DeepSpeech2Model",
+    "deepspeech2online":
+    "paddlespeech.s2t.models.ds2_online:DeepSpeech2ModelOnline",
+    "conformer":
+    "paddlespeech.s2t.models.u2:U2Model",
+    "transformer":
+    "paddlespeech.s2t.models.u2:U2Model",
+    "wenetspeech":
+    "paddlespeech.s2t.models.u2:U2Model",
+}

paddlespeech/cli/cls/infer.py

@@ -25,55 +25,23 @@ import yaml
 from ..executor import BaseExecutor
 from ..log import logger
 from ..utils import cli_register
-from ..utils import download_and_decompress
-from ..utils import MODEL_HOME
 from ..utils import stats_wrapper
+from .pretrained_models import model_alias
+from .pretrained_models import pretrained_models
 from paddleaudio import load
 from paddleaudio.features import LogMelSpectrogram
 from paddlespeech.s2t.utils.dynamic_import import dynamic_import
 
 __all__ = ['CLSExecutor']
 
-pretrained_models = {
-    # The tags for pretrained_models should be "{model_name}[_{dataset}][-{lang}][-...]".
-    # e.g. "conformer_wenetspeech-zh-16k", "transformer_aishell-zh-16k" and "panns_cnn6-32k".
-    # Command line and python api use "{model_name}[_{dataset}]" as --model, usage:
-    # "paddlespeech asr --model conformer_wenetspeech --lang zh --sr 16000 --input ./input.wav"
-    "panns_cnn6-32k": {
-        'url': 'https://paddlespeech.bj.bcebos.com/cls/panns_cnn6.tar.gz',
-        'md5': '4cf09194a95df024fd12f84712cf0f9c',
-        'cfg_path': 'panns.yaml',
-        'ckpt_path': 'cnn6.pdparams',
-        'label_file': 'audioset_labels.txt',
-    },
-    "panns_cnn10-32k": {
-        'url': 'https://paddlespeech.bj.bcebos.com/cls/panns_cnn10.tar.gz',
-        'md5': 'cb8427b22176cc2116367d14847f5413',
-        'cfg_path': 'panns.yaml',
-        'ckpt_path': 'cnn10.pdparams',
-        'label_file': 'audioset_labels.txt',
-    },
-    "panns_cnn14-32k": {
-        'url': 'https://paddlespeech.bj.bcebos.com/cls/panns_cnn14.tar.gz',
-        'md5': 'e3b9b5614a1595001161d0ab95edee97',
-        'cfg_path': 'panns.yaml',
-        'ckpt_path': 'cnn14.pdparams',
-        'label_file': 'audioset_labels.txt',
-    },
-}
-
-model_alias = {
-    "panns_cnn6": "paddlespeech.cls.models.panns:CNN6",
-    "panns_cnn10": "paddlespeech.cls.models.panns:CNN10",
-    "panns_cnn14": "paddlespeech.cls.models.panns:CNN14",
-}
-
 
 @cli_register(
     name='paddlespeech.cls', description='Audio classification infer command.')
 class CLSExecutor(BaseExecutor):
     def __init__(self):
-        super(CLSExecutor, self).__init__()
+        super().__init__()
+        self.model_alias = model_alias
+        self.pretrained_models = pretrained_models
 
         self.parser = argparse.ArgumentParser(
             prog='paddlespeech.cls', add_help=True)
@@ -83,7 +51,9 @@ class CLSExecutor(BaseExecutor):
             '--model',
             type=str,
             default='panns_cnn14',
-            choices=[tag[:tag.index('-')] for tag in pretrained_models.keys()],
+            choices=[
+                tag[:tag.index('-')] for tag in self.pretrained_models.keys()
+            ],
             help='Choose model type of cls task.')
         self.parser.add_argument(
             '--config',
@@ -121,23 +91,6 @@ class CLSExecutor(BaseExecutor):
             action='store_true',
             help='Increase logger verbosity of current task.')
 
-    def _get_pretrained_path(self, tag: str) -> os.PathLike:
-        """
-            Download and returns pretrained resources path of current task.
-        """
-        support_models = list(pretrained_models.keys())
-        assert tag in pretrained_models, 'The model "{}" you want to use has not been supported, please choose other models.\nThe support models includes:\n\t\t{}\n'.format(
-            tag, '\n\t\t'.join(support_models))
-
-        res_path = os.path.join(MODEL_HOME, tag)
-        decompressed_path = download_and_decompress(pretrained_models[tag],
-                                                    res_path)
-        decompressed_path = os.path.abspath(decompressed_path)
-        logger.info(
-            'Use pretrained model stored in: {}'.format(decompressed_path))
-
-        return decompressed_path
-
     def _init_from_path(self,
                         model_type: str='panns_cnn14',
                         cfg_path: Optional[os.PathLike]=None,
@@ -153,12 +106,12 @@ class CLSExecutor(BaseExecutor):
         if label_file is None or ckpt_path is None:
             tag = model_type + '-' + '32k'  # panns_cnn14-32k
             self.res_path = self._get_pretrained_path(tag)
-            self.cfg_path = os.path.join(self.res_path,
-                                         pretrained_models[tag]['cfg_path'])
-            self.label_file = os.path.join(self.res_path,
-                                           pretrained_models[tag]['label_file'])
-            self.ckpt_path = os.path.join(self.res_path,
-                                          pretrained_models[tag]['ckpt_path'])
+            self.cfg_path = os.path.join(
+                self.res_path, self.pretrained_models[tag]['cfg_path'])
+            self.label_file = os.path.join(
+                self.res_path, self.pretrained_models[tag]['label_file'])
+            self.ckpt_path = os.path.join(
+                self.res_path, self.pretrained_models[tag]['ckpt_path'])
         else:
             self.cfg_path = os.path.abspath(cfg_path)
             self.label_file = os.path.abspath(label_file)
@@ -175,7 +128,7 @@ class CLSExecutor(BaseExecutor):
                 self._label_list.append(line.strip())
 
         # model
-        model_class = dynamic_import(model_type, model_alias)
+        model_class = dynamic_import(model_type, self.model_alias)
         model_dict = paddle.load(self.ckpt_path)
         self.model = model_class(extract_embedding=False)
         self.model.set_state_dict(model_dict)

paddlespeech/cli/cls/pretrained_models.py

+# 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.
+
+pretrained_models = {
+    # The tags for pretrained_models should be "{model_name}[_{dataset}][-{lang}][-...]".
+    # e.g. "conformer_wenetspeech-zh-16k", "transformer_aishell-zh-16k" and "panns_cnn6-32k".
+    # Command line and python api use "{model_name}[_{dataset}]" as --model, usage:
+    # "paddlespeech asr --model conformer_wenetspeech --lang zh --sr 16000 --input ./input.wav"
+    "panns_cnn6-32k": {
+        'url': 'https://paddlespeech.bj.bcebos.com/cls/panns_cnn6.tar.gz',
+        'md5': '4cf09194a95df024fd12f84712cf0f9c',
+        'cfg_path': 'panns.yaml',
+        'ckpt_path': 'cnn6.pdparams',
+        'label_file': 'audioset_labels.txt',
+    },
+    "panns_cnn10-32k": {
+        'url': 'https://paddlespeech.bj.bcebos.com/cls/panns_cnn10.tar.gz',
+        'md5': 'cb8427b22176cc2116367d14847f5413',
+        'cfg_path': 'panns.yaml',
+        'ckpt_path': 'cnn10.pdparams',
+        'label_file': 'audioset_labels.txt',
+    },
+    "panns_cnn14-32k": {
+        'url': 'https://paddlespeech.bj.bcebos.com/cls/panns_cnn14.tar.gz',
+        'md5': 'e3b9b5614a1595001161d0ab95edee97',
+        'cfg_path': 'panns.yaml',
+        'ckpt_path': 'cnn14.pdparams',
+        'label_file': 'audioset_labels.txt',
+    },
+}
+
+model_alias = {
+    "panns_cnn6": "paddlespeech.cls.models.panns:CNN6",
+    "panns_cnn10": "paddlespeech.cls.models.panns:CNN10",
+    "panns_cnn14": "paddlespeech.cls.models.panns:CNN14",
+}

paddlespeech/cli/executor.py

@@ -25,6 +25,8 @@ from typing import Union
 import paddle
 
 from .log import logger
+from .utils import download_and_decompress
+from .utils import MODEL_HOME
 
 
 class BaseExecutor(ABC):
@@ -35,19 +37,8 @@ class BaseExecutor(ABC):
     def __init__(self):
         self._inputs = OrderedDict()
         self._outputs = OrderedDict()
-
-    @abstractmethod
-    def _get_pretrained_path(self, tag: str) -> os.PathLike:
-        """
-        Download and returns pretrained resources path of current task.
-
-        Args:
-            tag (str): A tag of pretrained model.
-
-        Returns:
-            os.PathLike: The path on which resources of pretrained model locate. 
-        """
-        pass
+        self.pretrained_models = OrderedDict()
+        self.model_alias = OrderedDict()
 
     @abstractmethod
     def _init_from_path(self, *args, **kwargs):
@@ -227,3 +218,20 @@ class BaseExecutor(ABC):
         ]
         for l in loggers:
             l.disabled = True
+
+    def _get_pretrained_path(self, tag: str) -> os.PathLike:
+        """
+        Download and returns pretrained resources path of current task.
+        """
+        support_models = list(self.pretrained_models.keys())
+        assert tag in self.pretrained_models, 'The model "{}" you want to use has not been supported, please choose other models.\nThe support models includes:\n\t\t{}\n'.format(
+            tag, '\n\t\t'.join(support_models))
+
+        res_path = os.path.join(MODEL_HOME, tag)
+        decompressed_path = download_and_decompress(self.pretrained_models[tag],
+                                                    res_path)
+        decompressed_path = os.path.abspath(decompressed_path)
+        logger.info(
+            'Use pretrained model stored in: {}'.format(decompressed_path))
+
+        return decompressed_path

paddlespeech/cli/st/infer.py

@@ -32,40 +32,24 @@ from ..utils import cli_register
 from ..utils import download_and_decompress
 from ..utils import MODEL_HOME
 from ..utils import stats_wrapper
+from .pretrained_models import kaldi_bins
+from .pretrained_models import model_alias
+from .pretrained_models import pretrained_models
 from paddlespeech.s2t.frontend.featurizer.text_featurizer import TextFeaturizer
 from paddlespeech.s2t.utils.dynamic_import import dynamic_import
 from paddlespeech.s2t.utils.utility import UpdateConfig
 
 __all__ = ["STExecutor"]
 
-pretrained_models = {
-    "fat_st_ted-en-zh": {
-        "url":
-        "https://paddlespeech.bj.bcebos.com/s2t/ted_en_zh/st1/st1_transformer_mtl_noam_ted-en-zh_ckpt_0.1.1.model.tar.gz",
-        "md5":
-        "d62063f35a16d91210a71081bd2dd557",
-        "cfg_path":
-        "model.yaml",
-        "ckpt_path":
-        "exp/transformer_mtl_noam/checkpoints/fat_st_ted-en-zh.pdparams",
-    }
-}
-
-model_alias = {"fat_st": "paddlespeech.s2t.models.u2_st:U2STModel"}
-
-kaldi_bins = {
-    "url":
-    "https://paddlespeech.bj.bcebos.com/s2t/ted_en_zh/st1/kaldi_bins.tar.gz",
-    "md5":
-    "c0682303b3f3393dbf6ed4c4e35a53eb",
-}
-
 
 @cli_register(
     name="paddlespeech.st", description="Speech translation infer command.")
 class STExecutor(BaseExecutor):
     def __init__(self):
-        super(STExecutor, self).__init__()
+        super().__init__()
+        self.model_alias = model_alias
+        self.pretrained_models = pretrained_models
+        self.kaldi_bins = kaldi_bins
 
         self.parser = argparse.ArgumentParser(
             prog="paddlespeech.st", add_help=True)
@@ -75,7 +59,9 @@ class STExecutor(BaseExecutor):
             "--model",
             type=str,
             default="fat_st_ted",
-            choices=[tag[:tag.index('-')] for tag in pretrained_models.keys()],
+            choices=[
+                tag[:tag.index('-')] for tag in self.pretrained_models.keys()
+            ],
             help="Choose model type of st task.")
         self.parser.add_argument(
             "--src_lang",
@@ -119,28 +105,11 @@ class STExecutor(BaseExecutor):
             action='store_true',
             help='Increase logger verbosity of current task.')
 
-    def _get_pretrained_path(self, tag: str) -> os.PathLike:
-        """
-            Download and returns pretrained resources path of current task.
-        """
-        support_models = list(pretrained_models.keys())
-        assert tag in pretrained_models, 'The model "{}" you want to use has not been supported, please choose other models.\nThe support models includes:\n\t\t{}\n'.format(
-            tag, '\n\t\t'.join(support_models))
-
-        res_path = os.path.join(MODEL_HOME, tag)
-        decompressed_path = download_and_decompress(pretrained_models[tag],
-                                                    res_path)
-        decompressed_path = os.path.abspath(decompressed_path)
-        logger.info(
-            "Use pretrained model stored in: {}".format(decompressed_path))
-
-        return decompressed_path
-
     def _set_kaldi_bins(self) -> os.PathLike:
         """
             Download and returns kaldi_bins resources path of current task.
         """
-        decompressed_path = download_and_decompress(kaldi_bins, MODEL_HOME)
+        decompressed_path = download_and_decompress(self.kaldi_bins, MODEL_HOME)
         decompressed_path = os.path.abspath(decompressed_path)
         logger.info("Kaldi_bins stored in: {}".format(decompressed_path))
         if "LD_LIBRARY_PATH" in os.environ:
@@ -197,7 +166,7 @@ class STExecutor(BaseExecutor):
         model_conf = self.config
         model_name = model_type[:model_type.rindex(
             '_')]  # model_type: {model_name}_{dataset}
-        model_class = dynamic_import(model_name, model_alias)
+        model_class = dynamic_import(model_name, self.model_alias)
         self.model = model_class.from_config(model_conf)
         self.model.eval()
 

paddlespeech/cli/st/pretrained_models.py

+# 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.
+
+pretrained_models = {
+    "fat_st_ted-en-zh": {
+        "url":
+        "https://paddlespeech.bj.bcebos.com/s2t/ted_en_zh/st1/st1_transformer_mtl_noam_ted-en-zh_ckpt_0.1.1.model.tar.gz",
+        "md5":
+        "d62063f35a16d91210a71081bd2dd557",
+        "cfg_path":
+        "model.yaml",
+        "ckpt_path":
+        "exp/transformer_mtl_noam/checkpoints/fat_st_ted-en-zh.pdparams",
+    }
+}
+
+model_alias = {"fat_st": "paddlespeech.s2t.models.u2_st:U2STModel"}
+
+kaldi_bins = {
+    "url":
+    "https://paddlespeech.bj.bcebos.com/s2t/ted_en_zh/st1/kaldi_bins.tar.gz",
+    "md5":
+    "c0682303b3f3393dbf6ed4c4e35a53eb",
+}

paddlespeech/cli/stats/infer.py

@@ -16,7 +16,6 @@ from typing import List
 
 from prettytable import PrettyTable
 
-from ..log import logger
 from ..utils import cli_register
 from ..utils import stats_wrapper
 
@@ -27,7 +26,8 @@ model_name_format = {
     'cls': 'Model-Sample Rate',
     'st': 'Model-Source language-Target language',
     'text': 'Model-Task-Language',
-    'tts': 'Model-Language'
+    'tts': 'Model-Language',
+    'vector': 'Model-Sample Rate'
 }
 
 
@@ -36,18 +36,18 @@ model_name_format = {
     description='Get speech tasks support models list.')
 class StatsExecutor():
     def __init__(self):
-        super(StatsExecutor, self).__init__()
+        super().__init__()
 
         self.parser = argparse.ArgumentParser(
             prog='paddlespeech.stats', add_help=True)
+        self.task_choices = ['asr', 'cls', 'st', 'text', 'tts', 'vector']
         self.parser.add_argument(
             '--task',
             type=str,
             default='asr',
-            choices=['asr', 'cls', 'st', 'text', 'tts'],
+            choices=self.task_choices,
             help='Choose speech task.',
             required=True)
-        self.task_choices = ['asr', 'cls', 'st', 'text', 'tts']
 
     def show_support_models(self, pretrained_models: dict):
         fields = model_name_format[self.task].split("-")
@@ -61,73 +61,15 @@ class StatsExecutor():
             Command line entry.
         """
         parser_args = self.parser.parse_args(argv)
-        self.task = parser_args.task
-        if self.task not in self.task_choices:
-            logger.error(
-                "Please input correct speech task, choices = ['asr', 'cls', 'st', 'text', 'tts']"
-            )
+        has_exceptions = False
+        try:
+            self(parser_args.task)
+        except Exception as e:
+            has_exceptions = True
+        if has_exceptions:
             return False
-
-        elif self.task == 'asr':
-            try:
-                from ..asr.infer import pretrained_models
-                logger.info(
-                    "Here is the list of ASR pretrained models released by PaddleSpeech that can be used by command line and python API"
-                )
-                self.show_support_models(pretrained_models)
-                return True
-            except BaseException:
-                logger.error("Failed to get the list of ASR pretrained models.")
-                return False
-
-        elif self.task == 'cls':
-            try:
-                from ..cls.infer import pretrained_models
-                logger.info(
-                    "Here is the list of CLS pretrained models released by PaddleSpeech that can be used by command line and python API"
-                )
-                self.show_support_models(pretrained_models)
-                return True
-            except BaseException:
-                logger.error("Failed to get the list of CLS pretrained models.")
-                return False
-
-        elif self.task == 'st':
-            try:
-                from ..st.infer import pretrained_models
-                logger.info(
-                    "Here is the list of ST pretrained models released by PaddleSpeech that can be used by command line and python API"
-                )
-                self.show_support_models(pretrained_models)
-                return True
-            except BaseException:
-                logger.error("Failed to get the list of ST pretrained models.")
-                return False
-
-        elif self.task == 'text':
-            try:
-                from ..text.infer import pretrained_models
-                logger.info(
-                    "Here is the list of TEXT pretrained models released by PaddleSpeech that can be used by command line and python API"
-                )
-                self.show_support_models(pretrained_models)
-                return True
-            except BaseException:
-                logger.error(
-                    "Failed to get the list of TEXT pretrained models.")
-                return False
-
-        elif self.task == 'tts':
-            try:
-                from ..tts.infer import pretrained_models
-                logger.info(
-                    "Here is the list of TTS pretrained models released by PaddleSpeech that can be used by command line and python API"
-                )
-                self.show_support_models(pretrained_models)
-                return True
-            except BaseException:
-                logger.error("Failed to get the list of TTS pretrained models.")
-                return False
+        else:
+            return True
 
     @stats_wrapper
     def __call__(
@@ -138,13 +80,12 @@ class StatsExecutor():
         """
         self.task = task
         if self.task not in self.task_choices:
-            print(
-                "Please input correct speech task, choices = ['asr', 'cls', 'st', 'text', 'tts']"
-            )
+            print("Please input correct speech task, choices = " + str(
+                self.task_choices))
 
         elif self.task == 'asr':
             try:
-                from ..asr.infer import pretrained_models
+                from ..asr.pretrained_models import pretrained_models
                 print(
                     "Here is the list of ASR pretrained models released by PaddleSpeech that can be used by command line and python API"
                 )
@@ -154,7 +95,7 @@ class StatsExecutor():
 
         elif self.task == 'cls':
             try:
-                from ..cls.infer import pretrained_models
+                from ..cls.pretrained_models import pretrained_models
                 print(
                     "Here is the list of CLS pretrained models released by PaddleSpeech that can be used by command line and python API"
                 )
@@ -164,7 +105,7 @@ class StatsExecutor():
 
         elif self.task == 'st':
             try:
-                from ..st.infer import pretrained_models
+                from ..st.pretrained_models import pretrained_models
                 print(
                     "Here is the list of ST pretrained models released by PaddleSpeech that can be used by command line and python API"
                 )
@@ -174,7 +115,7 @@ class StatsExecutor():
 
         elif self.task == 'text':
             try:
-                from ..text.infer import pretrained_models
+                from ..text.pretrained_models import pretrained_models
                 print(
                     "Here is the list of TEXT pretrained models released by PaddleSpeech that can be used by command line and python API"
                 )
@@ -184,10 +125,22 @@ class StatsExecutor():
 
         elif self.task == 'tts':
             try:
-                from ..tts.infer import pretrained_models
+                from ..tts.pretrained_models import pretrained_models
                 print(
                     "Here is the list of TTS pretrained models released by PaddleSpeech that can be used by command line and python API"
                 )
                 self.show_support_models(pretrained_models)
             except BaseException:
                 print("Failed to get the list of TTS pretrained models.")
+
+        elif self.task == 'vector':
+            try:
+                from ..vector.pretrained_models import pretrained_models
+                print(
+                    "Here is the list of Speaker Recognition pretrained models released by PaddleSpeech that can be used by command line and python API"
+                )
+                self.show_support_models(pretrained_models)
+            except BaseException:
+                print(
+                    "Failed to get the list of Speaker Recognition pretrained models."
+                )

paddlespeech/cli/text/infer.py

@@ -25,58 +25,21 @@ from ...s2t.utils.dynamic_import import dynamic_import
 from ..executor import BaseExecutor
 from ..log import logger
 from ..utils import cli_register
-from ..utils import download_and_decompress
-from ..utils import MODEL_HOME
 from ..utils import stats_wrapper
+from .pretrained_models import model_alias
+from .pretrained_models import pretrained_models
+from .pretrained_models import tokenizer_alias
 
 __all__ = ['TextExecutor']
 
-pretrained_models = {
-    # The tags for pretrained_models should be "{model_name}[_{dataset}][-{lang}][-...]".
-    # e.g. "conformer_wenetspeech-zh-16k", "transformer_aishell-zh-16k" and "panns_cnn6-32k".
-    # Command line and python api use "{model_name}[_{dataset}]" as --model, usage:
-    # "paddlespeech asr --model conformer_wenetspeech --lang zh --sr 16000 --input ./input.wav"
-    "ernie_linear_p7_wudao-punc-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/text/ernie_linear_p7_wudao-punc-zh.tar.gz',
-        'md5':
-        '12283e2ddde1797c5d1e57036b512746',
-        'cfg_path':
-        'ckpt/model_config.json',
-        'ckpt_path':
-        'ckpt/model_state.pdparams',
-        'vocab_file':
-        'punc_vocab.txt',
-    },
-    "ernie_linear_p3_wudao-punc-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/text/ernie_linear_p3_wudao-punc-zh.tar.gz',
-        'md5':
-        '448eb2fdf85b6a997e7e652e80c51dd2',
-        'cfg_path':
-        'ckpt/model_config.json',
-        'ckpt_path':
-        'ckpt/model_state.pdparams',
-        'vocab_file':
-        'punc_vocab.txt',
-    },
-}
-
-model_alias = {
-    "ernie_linear_p7": "paddlespeech.text.models:ErnieLinear",
-    "ernie_linear_p3": "paddlespeech.text.models:ErnieLinear",
-}
-
-tokenizer_alias = {
-    "ernie_linear_p7": "paddlenlp.transformers:ErnieTokenizer",
-    "ernie_linear_p3": "paddlenlp.transformers:ErnieTokenizer",
-}
-
 
 @cli_register(name='paddlespeech.text', description='Text infer command.')
 class TextExecutor(BaseExecutor):
     def __init__(self):
-        super(TextExecutor, self).__init__()
+        super().__init__()
+        self.model_alias = model_alias
+        self.pretrained_models = pretrained_models
+        self.tokenizer_alias = tokenizer_alias
 
         self.parser = argparse.ArgumentParser(
             prog='paddlespeech.text', add_help=True)
@@ -92,7 +55,9 @@ class TextExecutor(BaseExecutor):
             '--model',
             type=str,
             default='ernie_linear_p7_wudao',
-            choices=[tag[:tag.index('-')] for tag in pretrained_models.keys()],
+            choices=[
+                tag[:tag.index('-')] for tag in self.pretrained_models.keys()
+            ],
             help='Choose model type of text task.')
         self.parser.add_argument(
             '--lang',
@@ -131,23 +96,6 @@ class TextExecutor(BaseExecutor):
             action='store_true',
             help='Increase logger verbosity of current task.')
 
-    def _get_pretrained_path(self, tag: str) -> os.PathLike:
-        """
-            Download and returns pretrained resources path of current task.
-        """
-        support_models = list(pretrained_models.keys())
-        assert tag in pretrained_models, 'The model "{}" you want to use has not been supported, please choose other models.\nThe support models includes:\n\t\t{}\n'.format(
-            tag, '\n\t\t'.join(support_models))
-
-        res_path = os.path.join(MODEL_HOME, tag)
-        decompressed_path = download_and_decompress(pretrained_models[tag],
-                                                    res_path)
-        decompressed_path = os.path.abspath(decompressed_path)
-        logger.info(
-            'Use pretrained model stored in: {}'.format(decompressed_path))
-
-        return decompressed_path
-
     def _init_from_path(self,
                         task: str='punc',
                         model_type: str='ernie_linear_p7_wudao',
@@ -167,12 +115,12 @@ class TextExecutor(BaseExecutor):
         if cfg_path is None or ckpt_path is None or vocab_file is None:
             tag = '-'.join([model_type, task, lang])
             self.res_path = self._get_pretrained_path(tag)
-            self.cfg_path = os.path.join(self.res_path,
-                                         pretrained_models[tag]['cfg_path'])
-            self.ckpt_path = os.path.join(self.res_path,
-                                          pretrained_models[tag]['ckpt_path'])
-            self.vocab_file = os.path.join(self.res_path,
-                                           pretrained_models[tag]['vocab_file'])
+            self.cfg_path = os.path.join(
+                self.res_path, self.pretrained_models[tag]['cfg_path'])
+            self.ckpt_path = os.path.join(
+                self.res_path, self.pretrained_models[tag]['ckpt_path'])
+            self.vocab_file = os.path.join(
+                self.res_path, self.pretrained_models[tag]['vocab_file'])
         else:
             self.cfg_path = os.path.abspath(cfg_path)
             self.ckpt_path = os.path.abspath(ckpt_path)
@@ -187,8 +135,8 @@ class TextExecutor(BaseExecutor):
                     self._punc_list.append(line.strip())
 
             # model
-            model_class = dynamic_import(model_name, model_alias)
-            tokenizer_class = dynamic_import(model_name, tokenizer_alias)
+            model_class = dynamic_import(model_name, self.model_alias)
+            tokenizer_class = dynamic_import(model_name, self.tokenizer_alias)
             self.model = model_class(
                 cfg_path=self.cfg_path, ckpt_path=self.ckpt_path)
             self.tokenizer = tokenizer_class.from_pretrained('ernie-1.0')

paddlespeech/cli/text/pretrained_models.py

+# 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.
+
+pretrained_models = {
+    # The tags for pretrained_models should be "{model_name}[_{dataset}][-{lang}][-...]".
+    # e.g. "conformer_wenetspeech-zh-16k", "transformer_aishell-zh-16k" and "panns_cnn6-32k".
+    # Command line and python api use "{model_name}[_{dataset}]" as --model, usage:
+    # "paddlespeech asr --model conformer_wenetspeech --lang zh --sr 16000 --input ./input.wav"
+    "ernie_linear_p7_wudao-punc-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/text/ernie_linear_p7_wudao-punc-zh.tar.gz',
+        'md5':
+        '12283e2ddde1797c5d1e57036b512746',
+        'cfg_path':
+        'ckpt/model_config.json',
+        'ckpt_path':
+        'ckpt/model_state.pdparams',
+        'vocab_file':
+        'punc_vocab.txt',
+    },
+    "ernie_linear_p3_wudao-punc-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/text/ernie_linear_p3_wudao-punc-zh.tar.gz',
+        'md5':
+        '448eb2fdf85b6a997e7e652e80c51dd2',
+        'cfg_path':
+        'ckpt/model_config.json',
+        'ckpt_path':
+        'ckpt/model_state.pdparams',
+        'vocab_file':
+        'punc_vocab.txt',
+    },
+}
+
+model_alias = {
+    "ernie_linear_p7": "paddlespeech.text.models:ErnieLinear",
+    "ernie_linear_p3": "paddlespeech.text.models:ErnieLinear",
+}
+
+tokenizer_alias = {
+    "ernie_linear_p7": "paddlenlp.transformers:ErnieTokenizer",
+    "ernie_linear_p3": "paddlenlp.transformers:ErnieTokenizer",
+}

paddlespeech/cli/tts/infer.py

@@ -29,9 +29,9 @@ from yacs.config import CfgNode
 from ..executor import BaseExecutor
 from ..log import logger
 from ..utils import cli_register
-from ..utils import download_and_decompress
-from ..utils import MODEL_HOME
 from ..utils import stats_wrapper
+from .pretrained_models import model_alias
+from .pretrained_models import pretrained_models
 from paddlespeech.s2t.utils.dynamic_import import dynamic_import
 from paddlespeech.t2s.frontend import English
 from paddlespeech.t2s.frontend.zh_frontend import Frontend
@@ -39,299 +39,14 @@ from paddlespeech.t2s.modules.normalizer import ZScore
 
 __all__ = ['TTSExecutor']
 
-pretrained_models = {
-    # speedyspeech
-    "speedyspeech_csmsc-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_ckpt_0.2.0.zip',
-        'md5':
-        '6f6fa967b408454b6662c8c00c0027cb',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_30600.pdz',
-        'speech_stats':
-        'feats_stats.npy',
-        'phones_dict':
-        'phone_id_map.txt',
-        'tones_dict':
-        'tone_id_map.txt',
-    },
-
-    # fastspeech2
-    "fastspeech2_csmsc-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_baker_ckpt_0.4.zip',
-        'md5':
-        '637d28a5e53aa60275612ba4393d5f22',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_76000.pdz',
-        'speech_stats':
-        'speech_stats.npy',
-        'phones_dict':
-        'phone_id_map.txt',
-    },
-    "fastspeech2_ljspeech-en": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_ljspeech_ckpt_0.5.zip',
-        'md5':
-        'ffed800c93deaf16ca9b3af89bfcd747',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_100000.pdz',
-        'speech_stats':
-        'speech_stats.npy',
-        'phones_dict':
-        'phone_id_map.txt',
-    },
-    "fastspeech2_aishell3-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_aishell3_ckpt_0.4.zip',
-        'md5':
-        'f4dd4a5f49a4552b77981f544ab3392e',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_96400.pdz',
-        'speech_stats':
-        'speech_stats.npy',
-        'phones_dict':
-        'phone_id_map.txt',
-        'speaker_dict':
-        'speaker_id_map.txt',
-    },
-    "fastspeech2_vctk-en": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_vctk_ckpt_0.5.zip',
-        'md5':
-        '743e5024ca1e17a88c5c271db9779ba4',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_66200.pdz',
-        'speech_stats':
-        'speech_stats.npy',
-        'phones_dict':
-        'phone_id_map.txt',
-        'speaker_dict':
-        'speaker_id_map.txt',
-    },
-    # tacotron2
-    "tacotron2_csmsc-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/tacotron2/tacotron2_csmsc_ckpt_0.2.0.zip',
-        'md5':
-        '0df4b6f0bcbe0d73c5ed6df8867ab91a',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_30600.pdz',
-        'speech_stats':
-        'speech_stats.npy',
-        'phones_dict':
-        'phone_id_map.txt',
-    },
-    "tacotron2_ljspeech-en": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/tacotron2/tacotron2_ljspeech_ckpt_0.2.0.zip',
-        'md5':
-        '6a5eddd81ae0e81d16959b97481135f3',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_60300.pdz',
-        'speech_stats':
-        'speech_stats.npy',
-        'phones_dict':
-        'phone_id_map.txt',
-    },
-
-    # pwgan
-    "pwgan_csmsc-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_baker_ckpt_0.4.zip',
-        'md5':
-        '2e481633325b5bdf0a3823c714d2c117',
-        'config':
-        'pwg_default.yaml',
-        'ckpt':
-        'pwg_snapshot_iter_400000.pdz',
-        'speech_stats':
-        'pwg_stats.npy',
-    },
-    "pwgan_ljspeech-en": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_ljspeech_ckpt_0.5.zip',
-        'md5':
-        '53610ba9708fd3008ccaf8e99dacbaf0',
-        'config':
-        'pwg_default.yaml',
-        'ckpt':
-        'pwg_snapshot_iter_400000.pdz',
-        'speech_stats':
-        'pwg_stats.npy',
-    },
-    "pwgan_aishell3-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_aishell3_ckpt_0.5.zip',
-        'md5':
-        'd7598fa41ad362d62f85ffc0f07e3d84',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_1000000.pdz',
-        'speech_stats':
-        'feats_stats.npy',
-    },
-    "pwgan_vctk-en": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_vctk_ckpt_0.1.1.zip',
-        'md5':
-        'b3da1defcde3e578be71eb284cb89f2c',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_1500000.pdz',
-        'speech_stats':
-        'feats_stats.npy',
-    },
-    # mb_melgan
-    "mb_melgan_csmsc-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_ckpt_0.1.1.zip',
-        'md5':
-        'ee5f0604e20091f0d495b6ec4618b90d',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_1000000.pdz',
-        'speech_stats':
-        'feats_stats.npy',
-    },
-    # style_melgan
-    "style_melgan_csmsc-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/style_melgan/style_melgan_csmsc_ckpt_0.1.1.zip',
-        'md5':
-        '5de2d5348f396de0c966926b8c462755',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_1500000.pdz',
-        'speech_stats':
-        'feats_stats.npy',
-    },
-    # hifigan
-    "hifigan_csmsc-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_ckpt_0.1.1.zip',
-        'md5':
-        'dd40a3d88dfcf64513fba2f0f961ada6',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_2500000.pdz',
-        'speech_stats':
-        'feats_stats.npy',
-    },
-    "hifigan_ljspeech-en": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_ckpt_0.2.0.zip',
-        'md5':
-        '70e9131695decbca06a65fe51ed38a72',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_2500000.pdz',
-        'speech_stats':
-        'feats_stats.npy',
-    },
-    "hifigan_aishell3-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_ckpt_0.2.0.zip',
-        'md5':
-        '3bb49bc75032ed12f79c00c8cc79a09a',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_2500000.pdz',
-        'speech_stats':
-        'feats_stats.npy',
-    },
-    "hifigan_vctk-en": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_ckpt_0.2.0.zip',
-        'md5':
-        '7da8f88359bca2457e705d924cf27bd4',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_2500000.pdz',
-        'speech_stats':
-        'feats_stats.npy',
-    },
-
-    # wavernn
-    "wavernn_csmsc-zh": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/wavernn/wavernn_csmsc_ckpt_0.2.0.zip',
-        'md5':
-        'ee37b752f09bcba8f2af3b777ca38e13',
-        'config':
-        'default.yaml',
-        'ckpt':
-        'snapshot_iter_400000.pdz',
-        'speech_stats':
-        'feats_stats.npy',
-    }
-}
-
-model_alias = {
-    # acoustic model
-    "speedyspeech":
-    "paddlespeech.t2s.models.speedyspeech:SpeedySpeech",
-    "speedyspeech_inference":
-    "paddlespeech.t2s.models.speedyspeech:SpeedySpeechInference",
-    "fastspeech2":
-    "paddlespeech.t2s.models.fastspeech2:FastSpeech2",
-    "fastspeech2_inference":
-    "paddlespeech.t2s.models.fastspeech2:FastSpeech2Inference",
-    "tacotron2":
-    "paddlespeech.t2s.models.tacotron2:Tacotron2",
-    "tacotron2_inference":
-    "paddlespeech.t2s.models.tacotron2:Tacotron2Inference",
-    # voc
-    "pwgan":
-    "paddlespeech.t2s.models.parallel_wavegan:PWGGenerator",
-    "pwgan_inference":
-    "paddlespeech.t2s.models.parallel_wavegan:PWGInference",
-    "mb_melgan":
-    "paddlespeech.t2s.models.melgan:MelGANGenerator",
-    "mb_melgan_inference":
-    "paddlespeech.t2s.models.melgan:MelGANInference",
-    "style_melgan":
-    "paddlespeech.t2s.models.melgan:StyleMelGANGenerator",
-    "style_melgan_inference":
-    "paddlespeech.t2s.models.melgan:StyleMelGANInference",
-    "hifigan":
-    "paddlespeech.t2s.models.hifigan:HiFiGANGenerator",
-    "hifigan_inference":
-    "paddlespeech.t2s.models.hifigan:HiFiGANInference",
-    "wavernn":
-    "paddlespeech.t2s.models.wavernn:WaveRNN",
-    "wavernn_inference":
-    "paddlespeech.t2s.models.wavernn:WaveRNNInference",
-}
-
 
 @cli_register(
     name='paddlespeech.tts', description='Text to Speech infer command.')
 class TTSExecutor(BaseExecutor):
     def __init__(self):
         super().__init__()
+        self.model_alias = model_alias
+        self.pretrained_models = pretrained_models
 
         self.parser = argparse.ArgumentParser(
             prog='paddlespeech.tts', add_help=True)
@@ -449,22 +164,6 @@ class TTSExecutor(BaseExecutor):
             action='store_true',
             help='Increase logger verbosity of current task.')
 
-    def _get_pretrained_path(self, tag: str) -> os.PathLike:
-        """
-        Download and returns pretrained resources path of current task.
-        """
-        support_models = list(pretrained_models.keys())
-        assert tag in pretrained_models, 'The model "{}" you want to use has not been supported, please choose other models.\nThe support models includes:\n\t\t{}\n'.format(
-            tag, '\n\t\t'.join(support_models))
-
-        res_path = os.path.join(MODEL_HOME, tag)
-        decompressed_path = download_and_decompress(pretrained_models[tag],
-                                                    res_path)
-        decompressed_path = os.path.abspath(decompressed_path)
-        logger.info(
-            'Use pretrained model stored in: {}'.format(decompressed_path))
-        return decompressed_path
-
     def _init_from_path(
             self,
             am: str='fastspeech2_csmsc',
@@ -490,16 +189,15 @@ class TTSExecutor(BaseExecutor):
         if am_ckpt is None or am_config is None or am_stat is None or phones_dict is None:
             am_res_path = self._get_pretrained_path(am_tag)
             self.am_res_path = am_res_path
-            self.am_config = os.path.join(am_res_path,
-                                          pretrained_models[am_tag]['config'])
+            self.am_config = os.path.join(
+                am_res_path, self.pretrained_models[am_tag]['config'])
             self.am_ckpt = os.path.join(am_res_path,
-                                        pretrained_models[am_tag]['ckpt'])
+                                        self.pretrained_models[am_tag]['ckpt'])
             self.am_stat = os.path.join(
-                am_res_path, pretrained_models[am_tag]['speech_stats'])
+                am_res_path, self.pretrained_models[am_tag]['speech_stats'])
             # must have phones_dict in acoustic
             self.phones_dict = os.path.join(
-                am_res_path, pretrained_models[am_tag]['phones_dict'])
-            print("self.phones_dict:", self.phones_dict)
+                am_res_path, self.pretrained_models[am_tag]['phones_dict'])
             logger.info(am_res_path)
             logger.info(self.am_config)
             logger.info(self.am_ckpt)
@@ -509,21 +207,20 @@ class TTSExecutor(BaseExecutor):
             self.am_stat = os.path.abspath(am_stat)
             self.phones_dict = os.path.abspath(phones_dict)
             self.am_res_path = os.path.dirname(os.path.abspath(self.am_config))
-        print("self.phones_dict:", self.phones_dict)
 
         # for speedyspeech
         self.tones_dict = None
-        if 'tones_dict' in pretrained_models[am_tag]:
+        if 'tones_dict' in self.pretrained_models[am_tag]:
             self.tones_dict = os.path.join(
-                am_res_path, pretrained_models[am_tag]['tones_dict'])
+                am_res_path, self.pretrained_models[am_tag]['tones_dict'])
             if tones_dict:
                 self.tones_dict = tones_dict
 
         # for multi speaker fastspeech2
         self.speaker_dict = None
-        if 'speaker_dict' in pretrained_models[am_tag]:
+        if 'speaker_dict' in self.pretrained_models[am_tag]:
             self.speaker_dict = os.path.join(
-                am_res_path, pretrained_models[am_tag]['speaker_dict'])
+                am_res_path, self.pretrained_models[am_tag]['speaker_dict'])
             if speaker_dict:
                 self.speaker_dict = speaker_dict
 
@@ -532,12 +229,12 @@ class TTSExecutor(BaseExecutor):
         if voc_ckpt is None or voc_config is None or voc_stat is None:
             voc_res_path = self._get_pretrained_path(voc_tag)
             self.voc_res_path = voc_res_path
-            self.voc_config = os.path.join(voc_res_path,
-                                           pretrained_models[voc_tag]['config'])
-            self.voc_ckpt = os.path.join(voc_res_path,
-                                         pretrained_models[voc_tag]['ckpt'])
+            self.voc_config = os.path.join(
+                voc_res_path, self.pretrained_models[voc_tag]['config'])
+            self.voc_ckpt = os.path.join(
+                voc_res_path, self.pretrained_models[voc_tag]['ckpt'])
             self.voc_stat = os.path.join(
-                voc_res_path, pretrained_models[voc_tag]['speech_stats'])
+                voc_res_path, self.pretrained_models[voc_tag]['speech_stats'])
             logger.info(voc_res_path)
             logger.info(self.voc_config)
             logger.info(self.voc_ckpt)
@@ -588,8 +285,9 @@ class TTSExecutor(BaseExecutor):
         # model: {model_name}_{dataset}
         am_name = am[:am.rindex('_')]
 
-        am_class = dynamic_import(am_name, model_alias)
-        am_inference_class = dynamic_import(am_name + '_inference', model_alias)
+        am_class = dynamic_import(am_name, self.model_alias)
+        am_inference_class = dynamic_import(am_name + '_inference',
+                                            self.model_alias)
 
         if am_name == 'fastspeech2':
             am = am_class(
@@ -618,9 +316,9 @@ class TTSExecutor(BaseExecutor):
         # vocoder
         # model: {model_name}_{dataset}
         voc_name = voc[:voc.rindex('_')]
-        voc_class = dynamic_import(voc_name, model_alias)
+        voc_class = dynamic_import(voc_name, self.model_alias)
         voc_inference_class = dynamic_import(voc_name + '_inference',
-                                             model_alias)
+                                             self.model_alias)
         if voc_name != 'wavernn':
             voc = voc_class(**self.voc_config["generator_params"])
             voc.set_state_dict(paddle.load(self.voc_ckpt)["generator_params"])
@@ -735,7 +433,6 @@ class TTSExecutor(BaseExecutor):
         am_ckpt = args.am_ckpt
         am_stat = args.am_stat
         phones_dict = args.phones_dict
-        print("phones_dict:", phones_dict)
         tones_dict = args.tones_dict
         speaker_dict = args.speaker_dict
         voc = args.voc

paddlespeech/cli/tts/pretrained_models.py

+# 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.
+
+pretrained_models = {
+    # speedyspeech
+    "speedyspeech_csmsc-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_ckpt_0.2.0.zip',
+        'md5':
+        '6f6fa967b408454b6662c8c00c0027cb',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_30600.pdz',
+        'speech_stats':
+        'feats_stats.npy',
+        'phones_dict':
+        'phone_id_map.txt',
+        'tones_dict':
+        'tone_id_map.txt',
+    },
+
+    # fastspeech2
+    "fastspeech2_csmsc-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_baker_ckpt_0.4.zip',
+        'md5':
+        '637d28a5e53aa60275612ba4393d5f22',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_76000.pdz',
+        'speech_stats':
+        'speech_stats.npy',
+        'phones_dict':
+        'phone_id_map.txt',
+    },
+    "fastspeech2_ljspeech-en": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_ljspeech_ckpt_0.5.zip',
+        'md5':
+        'ffed800c93deaf16ca9b3af89bfcd747',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_100000.pdz',
+        'speech_stats':
+        'speech_stats.npy',
+        'phones_dict':
+        'phone_id_map.txt',
+    },
+    "fastspeech2_aishell3-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_aishell3_ckpt_0.4.zip',
+        'md5':
+        'f4dd4a5f49a4552b77981f544ab3392e',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_96400.pdz',
+        'speech_stats':
+        'speech_stats.npy',
+        'phones_dict':
+        'phone_id_map.txt',
+        'speaker_dict':
+        'speaker_id_map.txt',
+    },
+    "fastspeech2_vctk-en": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_vctk_ckpt_0.5.zip',
+        'md5':
+        '743e5024ca1e17a88c5c271db9779ba4',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_66200.pdz',
+        'speech_stats':
+        'speech_stats.npy',
+        'phones_dict':
+        'phone_id_map.txt',
+        'speaker_dict':
+        'speaker_id_map.txt',
+    },
+    # tacotron2
+    "tacotron2_csmsc-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/tacotron2/tacotron2_csmsc_ckpt_0.2.0.zip',
+        'md5':
+        '0df4b6f0bcbe0d73c5ed6df8867ab91a',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_30600.pdz',
+        'speech_stats':
+        'speech_stats.npy',
+        'phones_dict':
+        'phone_id_map.txt',
+    },
+    "tacotron2_ljspeech-en": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/tacotron2/tacotron2_ljspeech_ckpt_0.2.0.zip',
+        'md5':
+        '6a5eddd81ae0e81d16959b97481135f3',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_60300.pdz',
+        'speech_stats':
+        'speech_stats.npy',
+        'phones_dict':
+        'phone_id_map.txt',
+    },
+
+    # pwgan
+    "pwgan_csmsc-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_baker_ckpt_0.4.zip',
+        'md5':
+        '2e481633325b5bdf0a3823c714d2c117',
+        'config':
+        'pwg_default.yaml',
+        'ckpt':
+        'pwg_snapshot_iter_400000.pdz',
+        'speech_stats':
+        'pwg_stats.npy',
+    },
+    "pwgan_ljspeech-en": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_ljspeech_ckpt_0.5.zip',
+        'md5':
+        '53610ba9708fd3008ccaf8e99dacbaf0',
+        'config':
+        'pwg_default.yaml',
+        'ckpt':
+        'pwg_snapshot_iter_400000.pdz',
+        'speech_stats':
+        'pwg_stats.npy',
+    },
+    "pwgan_aishell3-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_aishell3_ckpt_0.5.zip',
+        'md5':
+        'd7598fa41ad362d62f85ffc0f07e3d84',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_1000000.pdz',
+        'speech_stats':
+        'feats_stats.npy',
+    },
+    "pwgan_vctk-en": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_vctk_ckpt_0.1.1.zip',
+        'md5':
+        'b3da1defcde3e578be71eb284cb89f2c',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_1500000.pdz',
+        'speech_stats':
+        'feats_stats.npy',
+    },
+    # mb_melgan
+    "mb_melgan_csmsc-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_ckpt_0.1.1.zip',
+        'md5':
+        'ee5f0604e20091f0d495b6ec4618b90d',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_1000000.pdz',
+        'speech_stats':
+        'feats_stats.npy',
+    },
+    # style_melgan
+    "style_melgan_csmsc-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/style_melgan/style_melgan_csmsc_ckpt_0.1.1.zip',
+        'md5':
+        '5de2d5348f396de0c966926b8c462755',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_1500000.pdz',
+        'speech_stats':
+        'feats_stats.npy',
+    },
+    # hifigan
+    "hifigan_csmsc-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_ckpt_0.1.1.zip',
+        'md5':
+        'dd40a3d88dfcf64513fba2f0f961ada6',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_2500000.pdz',
+        'speech_stats':
+        'feats_stats.npy',
+    },
+    "hifigan_ljspeech-en": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_ckpt_0.2.0.zip',
+        'md5':
+        '70e9131695decbca06a65fe51ed38a72',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_2500000.pdz',
+        'speech_stats':
+        'feats_stats.npy',
+    },
+    "hifigan_aishell3-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_ckpt_0.2.0.zip',
+        'md5':
+        '3bb49bc75032ed12f79c00c8cc79a09a',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_2500000.pdz',
+        'speech_stats':
+        'feats_stats.npy',
+    },
+    "hifigan_vctk-en": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_ckpt_0.2.0.zip',
+        'md5':
+        '7da8f88359bca2457e705d924cf27bd4',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_2500000.pdz',
+        'speech_stats':
+        'feats_stats.npy',
+    },
+
+    # wavernn
+    "wavernn_csmsc-zh": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/Parakeet/released_models/wavernn/wavernn_csmsc_ckpt_0.2.0.zip',
+        'md5':
+        'ee37b752f09bcba8f2af3b777ca38e13',
+        'config':
+        'default.yaml',
+        'ckpt':
+        'snapshot_iter_400000.pdz',
+        'speech_stats':
+        'feats_stats.npy',
+    }
+}
+
+model_alias = {
+    # acoustic model
+    "speedyspeech":
+    "paddlespeech.t2s.models.speedyspeech:SpeedySpeech",
+    "speedyspeech_inference":
+    "paddlespeech.t2s.models.speedyspeech:SpeedySpeechInference",
+    "fastspeech2":
+    "paddlespeech.t2s.models.fastspeech2:FastSpeech2",
+    "fastspeech2_inference":
+    "paddlespeech.t2s.models.fastspeech2:FastSpeech2Inference",
+    "tacotron2":
+    "paddlespeech.t2s.models.tacotron2:Tacotron2",
+    "tacotron2_inference":
+    "paddlespeech.t2s.models.tacotron2:Tacotron2Inference",
+    # voc
+    "pwgan":
+    "paddlespeech.t2s.models.parallel_wavegan:PWGGenerator",
+    "pwgan_inference":
+    "paddlespeech.t2s.models.parallel_wavegan:PWGInference",
+    "mb_melgan":
+    "paddlespeech.t2s.models.melgan:MelGANGenerator",
+    "mb_melgan_inference":
+    "paddlespeech.t2s.models.melgan:MelGANInference",
+    "style_melgan":
+    "paddlespeech.t2s.models.melgan:StyleMelGANGenerator",
+    "style_melgan_inference":
+    "paddlespeech.t2s.models.melgan:StyleMelGANInference",
+    "hifigan":
+    "paddlespeech.t2s.models.hifigan:HiFiGANGenerator",
+    "hifigan_inference":
+    "paddlespeech.t2s.models.hifigan:HiFiGANInference",
+    "wavernn":
+    "paddlespeech.t2s.models.wavernn:WaveRNN",
+    "wavernn_inference":
+    "paddlespeech.t2s.models.wavernn:WaveRNNInference",
+}

paddlespeech/cli/vector/infer.py

@@ -27,45 +27,24 @@ from yacs.config import CfgNode
 from ..executor import BaseExecutor
 from ..log import logger
 from ..utils import cli_register
-from ..utils import download_and_decompress
-from ..utils import MODEL_HOME
 from ..utils import stats_wrapper
+from .pretrained_models import model_alias
+from .pretrained_models import pretrained_models
 from paddleaudio.backends import load as load_audio
 from paddleaudio.compliance.librosa import melspectrogram
 from paddlespeech.s2t.utils.dynamic_import import dynamic_import
 from paddlespeech.vector.io.batch import feature_normalize
 from paddlespeech.vector.modules.sid_model import SpeakerIdetification
 
-pretrained_models = {
-    # The tags for pretrained_models should be "{model_name}[-{dataset}][-{sr}][-...]".
-    # e.g. "ecapatdnn_voxceleb12-16k".
-    # Command line and python api use "{model_name}[-{dataset}]" as --model, usage:
-    # "paddlespeech vector --task spk --model ecapatdnn_voxceleb12-16k --sr 16000 --input ./input.wav"
-    "ecapatdnn_voxceleb12-16k": {
-        'url':
-        'https://paddlespeech.bj.bcebos.com/vector/voxceleb/sv0_ecapa_tdnn_voxceleb12_ckpt_0_2_0.tar.gz',
-        'md5':
-        'cc33023c54ab346cd318408f43fcaf95',
-        'cfg_path':
-        'conf/model.yaml',  # the yaml config path
-        'ckpt_path':
-        'model/model',  # the format is ${dir}/{model_name}, 
-        # so the first 'model' is dir, the second 'model' is the name
-        # this means we have a model stored as model/model.pdparams
-    },
-}
-
-model_alias = {
-    "ecapatdnn": "paddlespeech.vector.models.ecapa_tdnn:EcapaTdnn",
-}
-
 
 @cli_register(
     name="paddlespeech.vector",
     description="Speech to vector embedding infer command.")
 class VectorExecutor(BaseExecutor):
     def __init__(self):
-        super(VectorExecutor, self).__init__()
+        super().__init__()
+        self.model_alias = model_alias
+        self.pretrained_models = pretrained_models
 
         self.parser = argparse.ArgumentParser(
             prog="paddlespeech.vector", add_help=True)
@@ -128,8 +107,8 @@ class VectorExecutor(BaseExecutor):
 
         Returns:
             bool: 
-                 False: some audio occurs error
-                 True: all audio process success
+                False: some audio occurs error
+                True: all audio process success
         """
         # stage 0: parse the args and get the required args
         parser_args = self.parser.parse_args(argv)
@@ -289,32 +268,6 @@ class VectorExecutor(BaseExecutor):
 
         return res
 
-    def _get_pretrained_path(self, tag: str) -> os.PathLike:
-        """get the neural network path from the pretrained model list
-           we stored all the pretained mode in the variable `pretrained_models`
-
-        Args:
-            tag (str): model tag in the pretrained model list
-
-        Returns:
-            os.PathLike: the downloaded pretrained model path in the disk
-        """
-        support_models = list(pretrained_models.keys())
-        assert tag in pretrained_models, \
-            'The model "{}" you want to use has not been supported,'\
-            'please choose other models.\n' \
-            'The support models includes\n\t\t{}'.format(tag, "\n\t\t".join(support_models))
-
-        res_path = os.path.join(MODEL_HOME, tag)
-        decompressed_path = download_and_decompress(pretrained_models[tag],
-                                                    res_path)
-
-        decompressed_path = os.path.abspath(decompressed_path)
-        logger.info(
-            'Use pretrained model stored in: {}'.format(decompressed_path))
-
-        return decompressed_path
-
     def _init_from_path(self,
                         model_type: str='ecapatdnn_voxceleb12',
                         sample_rate: int=16000,
@@ -350,10 +303,11 @@ class VectorExecutor(BaseExecutor):
             res_path = self._get_pretrained_path(tag)
             self.res_path = res_path
 
-            self.cfg_path = os.path.join(res_path,
-                                         pretrained_models[tag]['cfg_path'])
+            self.cfg_path = os.path.join(
+                res_path, self.pretrained_models[tag]['cfg_path'])
             self.ckpt_path = os.path.join(
-                res_path, pretrained_models[tag]['ckpt_path'] + '.pdparams')
+                res_path,
+                self.pretrained_models[tag]['ckpt_path'] + '.pdparams')
         else:
             # get the model from disk
             self.cfg_path = os.path.abspath(cfg_path)
@@ -373,7 +327,7 @@ class VectorExecutor(BaseExecutor):
         logger.info("start to dynamic import the model class")
         model_name = model_type[:model_type.rindex('_')]
         logger.info(f"model name {model_name}")
-        model_class = dynamic_import(model_name, model_alias)
+        model_class = dynamic_import(model_name, self.model_alias)
         model_conf = self.config.model
         backbone = model_class(**model_conf)
         model = SpeakerIdetification(

paddlespeech/cli/vector/pretrained_models.py

+# 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.
+
+pretrained_models = {
+    # The tags for pretrained_models should be "{model_name}[-{dataset}][-{sr}][-...]".
+    # e.g. "ecapatdnn_voxceleb12-16k".
+    # Command line and python api use "{model_name}[-{dataset}]" as --model, usage:
+    # "paddlespeech vector --task spk --model ecapatdnn_voxceleb12-16k --sr 16000 --input ./input.wav"
+    "ecapatdnn_voxceleb12-16k": {
+        'url':
+        'https://paddlespeech.bj.bcebos.com/vector/voxceleb/sv0_ecapa_tdnn_voxceleb12_ckpt_0_2_0.tar.gz',
+        'md5':
+        'cc33023c54ab346cd318408f43fcaf95',
+        'cfg_path':
+        'conf/model.yaml',  # the yaml config path
+        'ckpt_path':
+        'model/model',  # the format is ${dir}/{model_name}, 
+        # so the first 'model' is dir, the second 'model' is the name
+        # this means we have a model stored as model/model.pdparams
+    },
+}
+
+model_alias = {
+    "ecapatdnn": "paddlespeech.vector.models.ecapa_tdnn:EcapaTdnn",
+}

tests/unit/cli/test_cli.sh

 #!/bin/bash
 set -e
+
 # Audio classification
 wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/cat.wav https://paddlespeech.bj.bcebos.com/PaddleAudio/dog.wav
 paddlespeech cls --input ./cat.wav --topk 10
@@ -28,26 +29,16 @@ paddlespeech tts --am tacotron2_csmsc --input "你好，欢迎使用百度飞桨
 paddlespeech tts --am tacotron2_csmsc --voc wavernn_csmsc --input "你好，欢迎使用百度飞桨深度学习框架！"
 paddlespeech tts --am tacotron2_ljspeech --voc pwgan_ljspeech --lang en --input "Life was like a box of chocolates, you never know what you're gonna get."
 
-
 # Speech Translation (only support linux)
 paddlespeech st --input ./en.wav
 
-
-# batch process
-echo -e "1 欢迎光临。\n2 谢谢惠顾。" | paddlespeech tts
-
-# shell pipeline
-paddlespeech asr --input ./zh.wav | paddlespeech text --task punc
-
-# stats
-paddlespeech stats --task asr
-paddlespeech stats --task tts
-paddlespeech stats --task cls
-
 # Speaker Verification 
 wget -c https://paddlespeech.bj.bcebos.com/vector/audio/85236145389.wav
 paddlespeech vector --task spk --input 85236145389.wav
 
+# batch process
+echo -e "1 欢迎光临。\n2 谢谢惠顾。" | paddlespeech tts
+
 echo -e "demo1 85236145389.wav \n demo2 85236145389.wav" > vec.job
 paddlespeech vector --task spk --input vec.job
 
@@ -55,4 +46,13 @@ echo -e "demo3 85236145389.wav \n demo4 85236145389.wav" | paddlespeech vector -
 rm 85236145389.wav 
 rm vec.job
 
+# shell pipeline
+paddlespeech asr --input ./zh.wav | paddlespeech text --task punc
 
+# stats
+paddlespeech stats --task asr
+paddlespeech stats --task tts
+paddlespeech stats --task cls
+paddlespeech stats --task text
+paddlespeech stats --task vector
+paddlespeech stats --task st