Merge branch 'PaddlePaddle:develop' into develop

audio/tests/features/base.py

@@ -37,7 +37,7 @@ class FeatTest(unittest.TestCase):
         self.waveform, self.sr = load(os.path.abspath(os.path.basename(url)))
         self.waveform = self.waveform.astype(
             np.float32
-        )  # paddlespeech.s2t.transform.spectrogram only supports float32 
+        )  # paddlespeech.audio.transform.spectrogram only supports float32
         dim = len(self.waveform.shape)
 
         assert dim in [1, 2]

audio/tests/features/test_istft.py

@@ -18,8 +18,8 @@ import paddle
 from paddleaudio.functional.window import get_window
 
 from .base import FeatTest
-from paddlespeech.s2t.transform.spectrogram import IStft
+from paddlespeech.audio.transform.spectrogram import IStft
-from paddlespeech.s2t.transform.spectrogram import Stft
+from paddlespeech.audio.transform.spectrogram import Stft
 
 
 class TestIstft(FeatTest):

audio/tests/features/test_log_melspectrogram.py

@@ -18,7 +18,7 @@ import paddle
 import paddleaudio
 
 from .base import FeatTest
-from paddlespeech.s2t.transform.spectrogram import LogMelSpectrogram
+from paddlespeech.audio.transform.spectrogram import LogMelSpectrogram
 
 
 class TestLogMelSpectrogram(FeatTest):

audio/tests/features/test_spectrogram.py

@@ -18,7 +18,7 @@ import paddle
 import paddleaudio
 
 from .base import FeatTest
-from paddlespeech.s2t.transform.spectrogram import Spectrogram
+from paddlespeech.audio.transform.spectrogram import Spectrogram
 
 
 class TestSpectrogram(FeatTest):

audio/tests/features/test_stft.py

@@ -18,7 +18,7 @@ import paddle
 from paddleaudio.functional.window import get_window
 
 from .base import FeatTest
-from paddlespeech.s2t.transform.spectrogram import Stft
+from paddlespeech.audio.transform.spectrogram import Stft
 
 
 class TestStft(FeatTest):

demos/README.md

@@ -18,4 +18,4 @@ This directory contains many speech applications in multiple scenarios.
 * style_fs2 - multi style control for FastSpeech2 model  
 * text_to_speech - convert text into speech 
 * self supervised pretraining - speech feature extraction and speech recognition based on wav2vec2
-* Wishper - speech recognize and translate based on Whisper model
+* Whisper - speech recognize and translate based on Whisper model

docs/tutorial/asr/tutorial_transformer.ipynb

@@ -236,8 +236,8 @@
     "warnings.filterwarnings('ignore')\n",
     "\n",
     "from yacs.config import CfgNode\n",
-    "from paddlespeech.s2t.transform.spectrogram import LogMelSpectrogramKaldi\n",
+    "from paddlespeech.audio.transform.spectrogram import LogMelSpectrogramKaldi\n",
-    "from paddlespeech.s2t.transform.cmvn import GlobalCMVN\n",
+    "from paddlespeech.audio.transform.cmvn import GlobalCMVN\n",
     "from paddlespeech.s2t.frontend.featurizer.text_featurizer import TextFeaturizer\n",
     "from paddlespeech.s2t.models.u2 import U2Model\n",
     "\n",

paddlespeech/audio/transform/perturb.py

@@ -14,6 +14,7 @@
 # Modified from espnet(https://github.com/espnet/espnet)
 import io
 import os
+import sys
 
 import h5py
 import librosa
@@ -98,7 +99,7 @@ class SoundHDF5File():
     def __contains__(self, item):
         return item in self.file
 
-    def __len__(self, item):
+    def __len__(self):
         return len(self.file)
 
     def __enter__(self):

paddlespeech/cli/asr/infer.py

@@ -274,7 +274,7 @@ class ASRExecutor(BaseExecutor):
             # fbank
             audio = preprocessing(audio, **preprocess_args)
 
-            audio_len = paddle.to_tensor(audio.shape[0]).unsqueeze(axis=0)
+            audio_len = paddle.to_tensor([audio.shape[0]]).unsqueeze(axis=0)
             audio = paddle.to_tensor(audio, dtype='float32').unsqueeze(axis=0)
 
             self._inputs["audio"] = audio

paddlespeech/s2t/io/reader.py

@@ -14,9 +14,12 @@
 # Modified from espnet(https://github.com/espnet/espnet)
 from collections import OrderedDict
 
+import io
+import os
 import kaldiio
 import numpy as np
 import soundfile
+import h5py
 
 from .utility import feat_type
 from paddlespeech.audio.transform.transformation import Transformation
@@ -401,7 +404,7 @@ class SoundHDF5File():
     def __contains__(self, item):
         return item in self.file
 
-    def __len__(self, item):
+    def __len__(self):
         return len(self.file)
 
     def __enter__(self):

paddlespeech/server/engine/tts/online/onnx/tts_engine.py

@@ -154,7 +154,7 @@ class TTSServerExecutor(TTSExecutor):
         self.voc_sess = get_sess(self.voc_ckpt, voc_sess_conf)
         logger.debug("Create voc sess successfully.")
 
-        with open(self.phones_dict, "r") as f:
+        with open(self.phones_dict, "r", encoding='utf-8') as f:
             phn_id = [line.strip().split() for line in f.readlines()]
         self.vocab_size = len(phn_id)
         logger.debug(f"vocab_size: {self.vocab_size}")

paddlespeech/t2s/frontend/tone_sandhi.py

@@ -237,30 +237,25 @@ class ToneSandhi():
     # output seg: [['听一听', 'v']]
     def _merge_yi(self, seg: List[Tuple[str, str]]) -> List[Tuple[str, str]]:
         new_seg = []
+        skip_next = False
         # function 1
         for i, (word, pos) in enumerate(seg):
-            if i - 1 >= 0 and word == "一" and i + 1 < len(seg) and seg[i - 1][
+            if skip_next:
-                    0] == seg[i + 1][0] and seg[i - 1][1] == "v":
+                skip_next = False
-                if i - 1 < len(new_seg):
+                continue
-                    new_seg[i -
+            if i - 1 >= 0 and word == "一" and i + 1 < len(seg) and seg[i - 1][0] == seg[i + 1][0] and seg[i - 1][1] == "v":
-                            1][0] = new_seg[i - 1][0] + "一" + new_seg[i - 1][0]
+                new_seg[-1] = (new_seg[-1][0] + "一" + seg[i + 1][0], new_seg[-1][1])
-                else:
+                skip_next = True
-                    new_seg.append([word, pos])
-                    new_seg.append([seg[i + 1][0], pos])
             else:
-                if i - 2 >= 0 and seg[i - 1][0] == "一" and seg[i - 2][
+                new_seg.append((word, pos))
-                        0] == word and pos == "v":
-                    continue
-                else:
-                    new_seg.append([word, pos])
         seg = new_seg
         new_seg = []
         # function 2
         for i, (word, pos) in enumerate(seg):
             if new_seg and new_seg[-1][0] == "一":
-                new_seg[-1][0] = new_seg[-1][0] + word
+                new_seg[-1] = (new_seg[-1][0] + word, new_seg[-1][1])
             else:
-                new_seg.append([word, pos])
+                new_seg.append((word, pos))
         return new_seg
 
     # the first and the second words are all_tone_three

setup.py

@@ -53,6 +53,7 @@ base = [
     "pandas",
     "paddleaudio>=1.1.0",
     "paddlenlp>=2.4.8",
+    "paddlepaddle==2.5.1",
     "paddleslim>=2.3.4",
     "ppdiffusers>=0.9.0",
     "paddlespeech_feat",

tests/unit/cli/test_cli.sh

@@ -10,11 +10,12 @@ paddlespeech cls --input ./cat.wav --topk 10
 paddlespeech text --input 今天的天气真不错啊你下午有空吗我想约你一起去吃饭 --model ernie_linear_p3_wudao_fast
 
 # Speech SSL
+wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav
 paddlespeech ssl --task asr --lang en --input ./en.wav
 paddlespeech ssl --task vector --lang en --input ./en.wav
 
 # Speech_recognition
-wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/zh.wav https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav https://paddlespeech.bj.bcebos.com/PaddleAudio/ch_zh_mix.wav
+wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/zh.wav https://paddlespeech.bj.bcebos.com/PaddleAudio/ch_zh_mix.wav
 paddlespeech asr --input ./zh.wav
 paddlespeech asr --model conformer_aishell --input ./zh.wav
 paddlespeech asr --model conformer_online_aishell --input ./zh.wav
@@ -110,5 +111,7 @@ paddlespeech whisper --task transcribe --input ./zh.wav
 # whisper recognize text and translate to English
 paddlespeech whisper --task translate --input ./zh.wav
 
+# to change model English-Only model
+paddlespeech whisper --lang en --size base --task transcribe  --input ./en.wav
 
 echo -e "\033[32mTest success !!!\033[0m"

tests/unit/doc/test_cli.md

@@ -0,0 +1,29 @@
+# test CLI 测试文档
+
+ 该文档为 CLI 测试说明，该测试目前覆盖大部分 paddlespeech 中的 CLI 推理。该 CI 建立后用于快速验证修复是否正确。
+
+ # 测试流程
+ ## 1. 环境安装
+
+ CI 重建时在已有通过版本 paddlepaddle-gpu==2.5.1, paddlepseech==develop 下运行。
+
+ CI 重建后在 paddlepaddle-gpu==develop, paddlepseech==develop 下运行。
+ 
+ ### 其他相关依赖
+
+ gcc >= 4.8.5,
+ python >= 3.8
+
+ ## 2. 功能测试
+
+ 在 repo 的 tests/unit/cli 中运行：
+
+  ```shell
+
+  source path.sh
+  bash test_cli.sh
+
+  ```
+## 3. 预期结果
+
+ 输出 "Test success"，且运行过程中无报错或 Error 即为成功。

utils/apply-cmvn.py

@@ -6,7 +6,7 @@ import kaldiio
 import numpy
 from distutils.util import strtobool
 
-from paddlespeech.s2t.transform.cmvn import CMVN
+from paddlespeech.audio.transform.cmvn import CMVN
 from paddlespeech.s2t.utils.cli_readers import file_reader_helper
 from paddlespeech.s2t.utils.cli_utils import get_commandline_args
 from paddlespeech.s2t.utils.cli_utils import is_scipy_wav_style

utils/compute-cmvn-stats.py

@@ -5,7 +5,7 @@ import logging
 import kaldiio
 import numpy as np
 
-from paddlespeech.s2t.transform.transformation import Transformation
+from paddlespeech.audio.transform.transformation import Transformation
 from paddlespeech.s2t.utils.cli_readers import file_reader_helper
 from paddlespeech.s2t.utils.cli_utils import get_commandline_args
 from paddlespeech.s2t.utils.cli_utils import is_scipy_wav_style

utils/copy-feats.py

@@ -4,7 +4,7 @@ import logging
 
 from distutils.util import strtobool
 
-from paddlespeech.s2t.transform.transformation import Transformation
+from paddlespeech.audio.transform.transformation import Transformation
 from paddlespeech.s2t.utils.cli_readers import file_reader_helper
 from paddlespeech.s2t.utils.cli_utils import get_commandline_args
 from paddlespeech.s2t.utils.cli_utils import is_scipy_wav_style

utils/feat-to-shape.py

@@ -3,7 +3,7 @@ import argparse
 import logging
 import sys
 
-from paddlespeech.s2t.transform.transformation import Transformation
+from paddlespeech.audio.transform.transformation import Transformation
 from paddlespeech.s2t.utils.cli_readers import file_reader_helper
 from paddlespeech.s2t.utils.cli_utils import get_commandline_args
 from paddlespeech.s2t.utils.cli_utils import is_scipy_wav_style