diff --git a/README.md b/README.md index 19ec61cb0..00367d787 100644 --- a/README.md +++ b/README.md @@ -7,7 +7,7 @@ - + @@ -179,6 +179,7 @@ Via the easy-to-use, efficient, flexible and scalable implementation, our vision ### Recent Update - 👑 2023.05.31: Add [WavLM ASR-en](https://github.com/PaddlePaddle/PaddleSpeech/blob/develop/examples/librispeech/asr5), WavLM fine-tuning for ASR on LibriSpeech. +- 🎉 2023.05.18: Add [Squeezeformer](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/aishell/asr1), Squeezeformer training for ASR on Aishell. - 👑 2023.05.04: Add [HuBERT ASR-en](https://github.com/PaddlePaddle/PaddleSpeech/blob/develop/examples/librispeech/asr4), HuBERT fine-tuning for ASR on LibriSpeech. - ⚡ 2023.04.28: Fix [0-d tensor](https://github.com/PaddlePaddle/PaddleSpeech/pull/3214), with the upgrade of paddlepaddle==2.5, the problem of modifying 0-d tensor has been solved. - 👑 2023.04.25: Add [AMP for U2 conformer](https://github.com/PaddlePaddle/PaddleSpeech/pull/3167). diff --git a/README_cn.md b/README_cn.md index 7aef30871..d70940dd2 100644 --- a/README_cn.md +++ b/README_cn.md @@ -184,6 +184,7 @@ ### 近期更新 - 👑 2023.05.31: 新增 [WavLM ASR-en](https://github.com/PaddlePaddle/PaddleSpeech/blob/develop/examples/librispeech/asr5), 基于WavLM的英语识别微调，使用LibriSpeech数据集 +- 🎉 2023.05.18: 新增 [Squeezeformer](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/aishell/asr1), 使用Squeezeformer进行训练，使用Aishell数据集 - 👑 2023.05.04: 新增 [HuBERT ASR-en](https://github.com/PaddlePaddle/PaddleSpeech/blob/develop/examples/librispeech/asr4), 基于HuBERT的英语识别微调，使用LibriSpeech数据集 - ⚡ 2023.04.28: 修正 [0-d tensor](https://github.com/PaddlePaddle/PaddleSpeech/pull/3214), 配合PaddlePaddle2.5升级修改了0-d tensor的问题。 - 👑 2023.04.25: 新增 [U2 conformer 的 AMP 训练](https://github.com/PaddlePaddle/PaddleSpeech/pull/3167). diff --git a/audio/README.md b/audio/README.md index d42d41229..a8c47efe8 100644 --- a/audio/README.md +++ b/audio/README.md @@ -14,7 +14,7 @@ Linux test build whl environment: * gcc/g++ - 8.2.0 * cmake - 3.18.0 (need install) -MAC：test build whl envrioment： +MAC：test build whl environment： * os * gcc/g++ 12.2.0 * cpu Intel Xeon E5 x86_64 diff --git a/audio/tests/features/base.py b/audio/tests/features/base.py index d183b72ad..3bb1d1dde 100644 --- a/audio/tests/features/base.py +++ b/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] diff --git a/audio/tests/features/test_istft.py b/audio/tests/features/test_istft.py index 9cf8cdd65..ea1ee5cb6 100644 --- a/audio/tests/features/test_istft.py +++ b/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.s2t.transform.spectrogram import Stft +from paddlespeech.audio.transform.spectrogram import IStft +from paddlespeech.audio.transform.spectrogram import Stft class TestIstft(FeatTest): diff --git a/audio/tests/features/test_log_melspectrogram.py b/audio/tests/features/test_log_melspectrogram.py index 7d5680387..b2765d3be 100644 --- a/audio/tests/features/test_log_melspectrogram.py +++ b/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): diff --git a/audio/tests/features/test_spectrogram.py b/audio/tests/features/test_spectrogram.py index 5fe5afee1..6f4609632 100644 --- a/audio/tests/features/test_spectrogram.py +++ b/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): diff --git a/audio/tests/features/test_stft.py b/audio/tests/features/test_stft.py index 58792ffe2..9511a2926 100644 --- a/audio/tests/features/test_stft.py +++ b/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): diff --git a/dataset/librispeech/librispeech.py b/dataset/librispeech/librispeech.py index 44567b0cf..2f5f9016c 100644 --- a/dataset/librispeech/librispeech.py +++ b/dataset/librispeech/librispeech.py @@ -30,6 +30,7 @@ import soundfile from paddlespeech.dataset.download import download from paddlespeech.dataset.download import unpack +from paddlespeech.utils.argparse import strtobool URL_ROOT = "http://openslr.elda.org/resources/12" #URL_ROOT = "https://openslr.magicdatatech.com/resources/12" @@ -63,7 +64,7 @@ parser.add_argument( parser.add_argument( "--full_download", default="True", - type=distutils.util.strtobool, + type=strtobool, help="Download all datasets for Librispeech." " If False, only download a minimal requirement (test-clean, dev-clean" " train-clean-100). (default: %(default)s)") diff --git a/demos/README.md b/demos/README.md index a41967864..6f9cd2e41 100644 --- a/demos/README.md +++ b/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 diff --git a/demos/audio_content_search/README.md b/demos/audio_content_search/README.md index 4428bf389..f04ac447e 100644 --- a/demos/audio_content_search/README.md +++ b/demos/audio_content_search/README.md @@ -19,7 +19,7 @@ You can choose one way from meduim and hard to install paddlespeech. The dependency refers to the requirements.txt, and install the dependency as follows: ``` -pip install -r requriement.txt +pip install -r requirements.txt ``` ### 2. Prepare Input File @@ -30,11 +30,20 @@ Here are sample files for this demo that can be downloaded: wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/zh.wav ``` -### 3. Usage +### 3. run paddlespeech_server +Before using the client, it is necessary to start paddlespeech_servers. + +Here are sample server configuration： +```bash +bash demos/audio_content_search/run.sh +``` +The logs of the two services will be recorded in 'acs.log' and 'streaming_asr.log' in this configuration. + +### 4. Usage - Command Line(Recommended) ```bash # Chinese - paddlespeech_client acs --server_ip 127.0.0.1 --port 8090 --input ./zh.wav + paddlespeech_client acs --server_ip 127.0.0.1 --port 8490 --input ./zh.wav ``` Usage: diff --git a/demos/audio_content_search/README_cn.md b/demos/audio_content_search/README_cn.md index 6f51c4cf2..16c1a3dd7 100644 --- a/demos/audio_content_search/README_cn.md +++ b/demos/audio_content_search/README_cn.md @@ -19,7 +19,7 @@ 依赖参见 requirements.txt, 安装依赖 ``` -pip install -r requriement.txt +pip install -r requirements.txt ``` ### 2. 准备输入 @@ -29,16 +29,26 @@ pip install -r requriement.txt ```bash wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/zh.wav ``` -### 3. 使用方法 + +### 3. 启动 server +使用 client 之前需要先启动 paddlespeech_server。 + +可以使用默认 server 配置： +```bash +bash demos/audio_content_search/run.sh +``` +该配置下两个服务的日志会被记录在 `acs.log` 和 `streaming_asr.log` 中。 + +### 4. 使用方法 - 命令行 (推荐使用) ```bash # 中文 - paddlespeech_client acs --server_ip 127.0.0.1 --port 8090 --input ./zh.wav + paddlespeech_client acs --server_ip 127.0.0.1 --port 8490 --input ./zh.wav ``` 使用方法： ```bash - paddlespeech acs --help + paddlespeech asr --help ``` 参数： - `input`(必须输入)：用于识别的音频文件。 diff --git a/demos/audio_content_search/conf/ws_conformer_application.yaml b/demos/audio_content_search/conf/ws_conformer_application.yaml index 97201382f..ad34ec9fd 100644 --- a/demos/audio_content_search/conf/ws_conformer_application.yaml +++ b/demos/audio_content_search/conf/ws_conformer_application.yaml @@ -26,8 +26,10 @@ asr_online: sample_rate: 16000 cfg_path: decode_method: 'attention_rescoring' + num_decoding_left_chunks: -1 # number of left chunks for decoding. Defaults to -1. force_yes: True device: 'cpu' # cpu or gpu:id + continuous_decoding: False # disable continue decoding when endpoint detected am_predictor_conf: device: # set 'gpu:id' or 'cpu' switch_ir_optim: True @@ -40,4 +42,4 @@ asr_online: window_ms: 25 # ms shift_ms: 10 # ms sample_rate: 16000 - sample_width: 2 + sample_width: 2 \ No newline at end of file diff --git a/demos/audio_content_search/conf/ws_conformer_wenetspeech_application.yaml b/demos/audio_content_search/conf/ws_conformer_wenetspeech_application.yaml index c23680bd5..ef1ce8d5c 100644 --- a/demos/audio_content_search/conf/ws_conformer_wenetspeech_application.yaml +++ b/demos/audio_content_search/conf/ws_conformer_wenetspeech_application.yaml @@ -31,6 +31,7 @@ asr_online: force_yes: True device: 'cpu' # cpu or gpu:id decode_method: "attention_rescoring" + num_decoding_left_chunks: -1 # number of left chunks for decoding. Defaults to -1. am_predictor_conf: device: # set 'gpu:id' or 'cpu' switch_ir_optim: True diff --git a/demos/audio_searching/requirements.txt b/demos/audio_searching/requirements.txt index 9d0f6419b..3c0f05afc 100644 --- a/demos/audio_searching/requirements.txt +++ b/demos/audio_searching/requirements.txt @@ -1,5 +1,5 @@ -diskcache==5.2.1 -dtaidistance==2.3.1 +diskcache +dtaidistane fastapi librosa==0.8.0 numpy==1.22.0 @@ -10,4 +10,4 @@ python-multipart soundfile==0.10.3.post1 starlette typing -uvicorn \ No newline at end of file +uvicorn diff --git a/demos/speech_ssl/README_cn.md b/demos/speech_ssl/README_cn.md index a18c778a7..5b209419a 100644 --- a/demos/speech_ssl/README_cn.md +++ b/demos/speech_ssl/README_cn.md @@ -56,7 +56,7 @@ wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav # 识别文本 text = ssl_executor( - model='wav2vec2, + model='wav2vec2', task='asr', lang='en', sample_rate=16000, diff --git a/demos/streaming_asr_server/README.md b/demos/streaming_asr_server/README.md index 31256d151..136863b96 100644 --- a/demos/streaming_asr_server/README.md +++ b/demos/streaming_asr_server/README.md @@ -429,7 +429,7 @@ bash server.sh If `127.0.0.1` is not accessible, you need to use the actual service IP address. ```bash - paddlespeech_client asr_online --server_ip 127.0.0.1 --port 8290 --punc.server_ip 127.0.0.1 --punc.port 8190 --input ./zh.wav + paddlespeech_client asr_online --server_ip 127.0.0.1 --port 8090 --punc.server_ip 127.0.0.1 --punc.port 8190 --input ./zh.wav ``` Output: ```text @@ -507,7 +507,7 @@ bash server.sh If `127.0.0.1` is not accessible, you need to use the actual service IP address. ```bash - python3 websocket_client.py --server_ip 127.0.0.1 --port 8290 --punc.server_ip 127.0.0.1 --punc.port 8190 --wavfile ./zh.wav + python3 local/websocket_client.py --server_ip 127.0.0.1 --port 8090 --punc.server_ip 127.0.0.1 --punc.port 8190 --wavfile ./zh.wav ``` Output: ```text diff --git a/demos/streaming_asr_server/README_cn.md b/demos/streaming_asr_server/README_cn.md index bbddd6932..f5f477ea1 100644 --- a/demos/streaming_asr_server/README_cn.md +++ b/demos/streaming_asr_server/README_cn.md @@ -428,7 +428,7 @@ bash server.sh 若 `127.0.0.1` 不能访问，则需要使用实际服务 IP 地址 ```bash - paddlespeech_client asr_online --server_ip 127.0.0.1 --port 8290 --punc.server_ip 127.0.0.1 --punc.port 8190 --input ./zh.wav + paddlespeech_client asr_online --server_ip 127.0.0.1 --port 8090 --punc.server_ip 127.0.0.1 --punc.port 8190 --input ./zh.wav ``` 输出: ```text @@ -506,7 +506,7 @@ bash server.sh 若 `127.0.0.1` 不能访问，则需要使用实际服务 IP 地址 ```bash - python3 websocket_client.py --server_ip 127.0.0.1 --port 8290 --punc.server_ip 127.0.0.1 --punc.port 8190 --wavfile ./zh.wav + python3 local/websocket_client.py --server_ip 127.0.0.1 --port 8090 --punc.server_ip 127.0.0.1 --punc.port 8190 --wavfile ./zh.wav ``` 输出: ```text diff --git a/demos/style_fs2/run.sh b/demos/style_fs2/run.sh index 6f6d60680..45fc0c104 100755 --- a/demos/style_fs2/run.sh +++ b/demos/style_fs2/run.sh @@ -32,7 +32,7 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then --pwg-config=download/pwg_baker_ckpt_0.4/pwg_default.yaml \ --pwg-checkpoint=download/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz \ --pwg-stat=download/pwg_baker_ckpt_0.4/pwg_stats.npy \ - --text=${BIN_DIR}/../sentences.txt \ + --text=./sentences.txt \ --output-dir=output \ --phones-dict=download/fastspeech2_nosil_baker_ckpt_0.4/phone_id_map.txt fi diff --git a/docs/tutorial/asr/tutorial_transformer.ipynb b/docs/tutorial/asr/tutorial_transformer.ipynb index dc3030061..77aed4bf8 100644 --- a/docs/tutorial/asr/tutorial_transformer.ipynb +++ b/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.s2t.transform.cmvn import GlobalCMVN\n", + "from paddlespeech.audio.transform.spectrogram import LogMelSpectrogramKaldi\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", diff --git a/examples/aishell/asr0/README.md b/examples/aishell/asr0/README.md index 131de36e3..a9469c460 100644 --- a/examples/aishell/asr0/README.md +++ b/examples/aishell/asr0/README.md @@ -103,12 +103,19 @@ If you want to train the model, you can use the script below to execute stage 0 ```bash bash run.sh --stage 0 --stop_stage 1 ``` -or you can run these scripts in the command line (only use CPU). +Or you can run these scripts in the command line (only use CPU). ```bash source path.sh bash ./local/data.sh -CUDA_VISIBLE_DEVICES= ./local/train.sh conf/deepspeech2.yaml deepspeech2 +CUDA_VISIBLE_DEVICES= ./local/train.sh conf/deepspeech2.yaml deepspeech2 ``` +If you want to use GPU, you can run these scripts in the command line (suppose you have only 1 GPU). +```bash +source path.sh +bash ./local/data.sh +CUDA_VISIBLE_DEVICES=0 ./local/train.sh conf/deepspeech2.yaml deepspeech2 +``` + ## Stage 2: Top-k Models Averaging After training the model, we need to get the final model for testing and inference. In every epoch, the model checkpoint is saved, so we can choose the best model from them based on the validation loss or we can sort them and average the parameters of the top-k models to get the final model. We can use stage 2 to do this, and the code is shown below: ```bash @@ -148,7 +155,7 @@ source path.sh bash ./local/data.sh CUDA_VISIBLE_DEVICES= ./local/train.sh conf/deepspeech2.yaml deepspeech2 avg.sh best exp/deepspeech2/checkpoints 1 -CUDA_VISIBLE_DEVICES= ./local/test.sh conf/deepspeech2.yaml exp/deepspeech2/checkpoints/avg_1 +CUDA_VISIBLE_DEVICES= ./local/test.sh conf/deepspeech2.yaml conf/tuning/decode.yaml exp/deepspeech2/checkpoints/avg_10 ``` ## Pretrained Model You can get the pretrained models from [this](../../../docs/source/released_model.md). @@ -157,14 +164,14 @@ using the `tar` scripts to unpack the model and then you can use the script to t For example: ``` -wget https://paddlespeech.bj.bcebos.com/s2t/aishell/asr0/asr0_deepspeech2_aishell_ckpt_0.1.1.model.tar.gz -tar xzvf asr0_deepspeech2_aishell_ckpt_0.1.1.model.tar.gz +wget https://paddlespeech.bj.bcebos.com/s2t/aishell/asr0/asr0_deepspeech2_offline_aishell_ckpt_1.0.1.model.tar.gz +tar xzvf asr0_deepspeech2_offline_aishell_ckpt_1.0.1.model.tar.gz source path.sh # If you have process the data and get the manifest file， you can skip the following 2 steps bash local/data.sh --stage -1 --stop_stage -1 bash local/data.sh --stage 2 --stop_stage 2 -CUDA_VISIBLE_DEVICES= ./local/test.sh conf/deepspeech2.yaml exp/deepspeech2/checkpoints/avg_1 +CUDA_VISIBLE_DEVICES= ./local/test.sh conf/deepspeech2.yaml exp/deepspeech2/checkpoints/avg_10 ``` The performance of the released models are shown in [this](./RESULTS.md) ## Stage 4: Static graph model Export @@ -178,7 +185,7 @@ This stage is to transform dygraph to static graph. If you already have a dynamic graph model, you can run this script: ```bash source path.sh -./local/export.sh deepspeech2.yaml exp/deepspeech2/checkpoints/avg_1 exp/deepspeech2/checkpoints/avg_1.jit offline +./local/export.sh conf/deepspeech2.yaml exp/deepspeech2/checkpoints/avg_10 exp/deepspeech2/checkpoints/avg_10.jit ``` ## Stage 5: Static graph Model Testing Similar to stage 3, the static graph model can also be tested. @@ -190,7 +197,7 @@ Similar to stage 3, the static graph model can also be tested. ``` If you already have exported the static graph, you can run this script: ```bash -CUDA_VISIBLE_DEVICES= ./local/test_export.sh conf/deepspeech2.yaml exp/deepspeech2/checkpoints/avg_1.jit offline +CUDA_VISIBLE_DEVICES= ./local/test_export.sh conf/deepspeech2.yaml conf/tuning/decode.yaml exp/deepspeech2/checkpoints/avg_10.jit ``` ## Stage 6: Single Audio File Inference In some situations, you want to use the trained model to do the inference for the single audio file. You can use stage 5. The code is shown below @@ -202,8 +209,8 @@ if [ ${stage} -le 6 ] && [ ${stop_stage} -ge 6 ]; then ``` you can train the model by yourself, or you can download the pretrained model by the script below: ```bash -wget https://paddlespeech.bj.bcebos.com/s2t/aishell/asr0/asr0_deepspeech2_aishell_ckpt_0.1.1.model.tar.gz -tar xzvf asr0_deepspeech2_aishell_ckpt_0.1.1.model.tar.gz +wget https://paddlespeech.bj.bcebos.com/s2t/aishell/asr0/asr0_deepspeech2_offline_aishell_ckpt_1.0.1.model.tar.gz +tar asr0_deepspeech2_offline_aishell_ckpt_1.0.1.model.tar.gz ``` You can download the audio demo: ```bash @@ -211,5 +218,5 @@ wget -nc https://paddlespeech.bj.bcebos.com/datasets/single_wav/zh/demo_01_03.wa ``` You need to prepare an audio file or use the audio demo above, please confirm the sample rate of the audio is 16K. You can get the result of the audio demo by running the script below. ```bash -CUDA_VISIBLE_DEVICES= ./local/test_wav.sh conf/deepspeech2.yaml conf/tuning/decode.yaml exp/deepspeech2/checkpoints/avg_1 data/demo_01_03.wav +CUDA_VISIBLE_DEVICES= ./local/test_wav.sh conf/deepspeech2.yaml conf/tuning/decode.yaml exp/deepspeech2/checkpoints/avg_10 data/demo_01_03.wav ``` diff --git a/examples/ami/sd0/local/ami_prepare.py b/examples/ami/sd0/local/ami_prepare.py index 1f02afe00..e38eb7131 100644 --- a/examples/ami/sd0/local/ami_prepare.py +++ b/examples/ami/sd0/local/ami_prepare.py @@ -28,7 +28,8 @@ import xml.etree.ElementTree as et from ami_splits import get_AMI_split from dataio import load_pkl from dataio import save_pkl -from distutils.util import strtobool + +from paddlespeech.utils.argparse import strtobool logger = logging.getLogger(__name__) SAMPLERATE = 16000 diff --git a/examples/csmsc/tts2/local/inference_mlu.sh b/examples/csmsc/tts2/local/inference_mlu.sh new file mode 100755 index 000000000..d1bade84d --- /dev/null +++ b/examples/csmsc/tts2/local/inference_mlu.sh @@ -0,0 +1,33 @@ +#!/bin/bash + +train_output_path=$1 + +stage=0 +stop_stage=0 + +# for more GAN Vocoders +# multi band melgan +if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then + python3 ${BIN_DIR}/../inference.py \ + --inference_dir=${train_output_path}/inference \ + --am=speedyspeech_csmsc \ + --voc=mb_melgan_csmsc \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/pd_infer_out \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --device mlu +fi + +# hifigan +if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then + python3 ${BIN_DIR}/../inference.py \ + --inference_dir=${train_output_path}/inference \ + --am=speedyspeech_csmsc \ + --voc=hifigan_csmsc \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/pd_infer_out \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --device mlu +fi diff --git a/examples/csmsc/tts2/local/inference_npu.sh b/examples/csmsc/tts2/local/inference_npu.sh new file mode 100644 index 000000000..0746a0cdb --- /dev/null +++ b/examples/csmsc/tts2/local/inference_npu.sh @@ -0,0 +1,46 @@ +#!/bin/bash + +train_output_path=$1 + +stage=0 +stop_stage=0 + +# pwgan +if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then + python3 ${BIN_DIR}/../inference.py \ + --inference_dir=${train_output_path}/inference \ + --am=speedyspeech_csmsc \ + --voc=pwgan_csmsc \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/pd_infer_out \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --device npu +fi + +# for more GAN Vocoders +# multi band melgan +if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then + python3 ${BIN_DIR}/../inference.py \ + --inference_dir=${train_output_path}/inference \ + --am=speedyspeech_csmsc \ + --voc=mb_melgan_csmsc \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/pd_infer_out \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --device npu +fi + +# hifigan +if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then + python3 ${BIN_DIR}/../inference.py \ + --inference_dir=${train_output_path}/inference \ + --am=speedyspeech_csmsc \ + --voc=hifigan_csmsc \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/pd_infer_out \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --device npu +fi diff --git a/examples/csmsc/tts2/local/synthesize_e2e_mlu.sh b/examples/csmsc/tts2/local/synthesize_e2e_mlu.sh new file mode 100755 index 000000000..7ad2024ff --- /dev/null +++ b/examples/csmsc/tts2/local/synthesize_e2e_mlu.sh @@ -0,0 +1,99 @@ +#!/bin/bash + +config_path=$1 +train_output_path=$2 +ckpt_name=$3 + +stage=0 +stop_stage=0 + +# for more GAN Vocoders +# multi band melgan +if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize_e2e.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=mb_melgan_csmsc \ + --voc_config=mb_melgan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=mb_melgan_csmsc_ckpt_0.1.1/snapshot_iter_1000000.pdz\ + --voc_stat=mb_melgan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --lang=zh \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/test_e2e \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --ngpu=0 \ + --nmlu=1 +fi + +# the pretrained models haven't release now +# style melgan +# style melgan's Dygraph to Static Graph is not ready now +if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize_e2e.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=style_melgan_csmsc \ + --voc_config=style_melgan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=style_melgan_csmsc_ckpt_0.1.1/snapshot_iter_1500000.pdz \ + --voc_stat=style_melgan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --lang=zh \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/test_e2e \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --ngpu=0 \ + --nmlu=1 + # --inference_dir=${train_output_path}/inference +fi + +# hifigan +if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize_e2e.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=hifigan_csmsc \ + --voc_config=hifigan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=hifigan_csmsc_ckpt_0.1.1/snapshot_iter_2500000.pdz \ + --voc_stat=hifigan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --lang=zh \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/test_e2e \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --inference_dir=${train_output_path}/inference \ + --ngpu=0 \ + --nmlu=1 +fi + +# wavernn +if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then + echo "in wavernn syn_e2e" + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize_e2e.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=wavernn_csmsc \ + --voc_config=wavernn_csmsc_ckpt_0.2.0/default.yaml \ + --voc_ckpt=wavernn_csmsc_ckpt_0.2.0/snapshot_iter_400000.pdz \ + --voc_stat=wavernn_csmsc_ckpt_0.2.0/feats_stats.npy \ + --lang=zh \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/test_e2e \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --inference_dir=${train_output_path}/inference \ + --ngpu=0 \ + --nmlu=1 +fi diff --git a/examples/csmsc/tts2/local/synthesize_e2e_npu.sh b/examples/csmsc/tts2/local/synthesize_e2e_npu.sh new file mode 100755 index 000000000..1209a532b --- /dev/null +++ b/examples/csmsc/tts2/local/synthesize_e2e_npu.sh @@ -0,0 +1,124 @@ +#!/bin/bash + +config_path=$1 +train_output_path=$2 +ckpt_name=$3 + +stage=0 +stop_stage=0 + +# pwgan +if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize_e2e.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=pwgan_csmsc \ + --voc_config=pwg_baker_ckpt_0.4/pwg_default.yaml \ + --voc_ckpt=pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz \ + --voc_stat=pwg_baker_ckpt_0.4/pwg_stats.npy \ + --lang=zh \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/test_e2e \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --inference_dir=${train_output_path}/inference \ + --ngpu=0 \ + --nnpu=1 + + +fi + +# for more GAN Vocoders +# multi band melgan +if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize_e2e.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=mb_melgan_csmsc \ + --voc_config=mb_melgan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=mb_melgan_csmsc_ckpt_0.1.1/snapshot_iter_1000000.pdz\ + --voc_stat=mb_melgan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --lang=zh \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/test_e2e \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --inference_dir=${train_output_path}/inference \ + --ngpu=0 \ + --nnpu=1 +fi + +# the pretrained models haven't release now +# style melgan +# style melgan's Dygraph to Static Graph is not ready now +if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize_e2e.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=style_melgan_csmsc \ + --voc_config=style_melgan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=style_melgan_csmsc_ckpt_0.1.1/snapshot_iter_1500000.pdz \ + --voc_stat=style_melgan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --lang=zh \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/test_e2e \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --ngpu=0 \ + --nnpu=1 + # --inference_dir=${train_output_path}/inference +fi + +# hifigan +if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize_e2e.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=hifigan_csmsc \ + --voc_config=hifigan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=hifigan_csmsc_ckpt_0.1.1/snapshot_iter_2500000.pdz \ + --voc_stat=hifigan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --lang=zh \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/test_e2e \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --inference_dir=${train_output_path}/inference \ + --ngpu=0 \ + --nnpu=1 +fi + +# wavernn +if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then + echo "in wavernn syn_e2e" + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize_e2e.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=wavernn_csmsc \ + --voc_config=wavernn_csmsc_ckpt_0.2.0/default.yaml \ + --voc_ckpt=wavernn_csmsc_ckpt_0.2.0/snapshot_iter_400000.pdz \ + --voc_stat=wavernn_csmsc_ckpt_0.2.0/feats_stats.npy \ + --lang=zh \ + --text=${BIN_DIR}/../../assets/sentences.txt \ + --output_dir=${train_output_path}/test_e2e \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --inference_dir=${train_output_path}/inference \ + --ngpu=0 \ + --nnpu=1 +fi diff --git a/examples/csmsc/tts2/local/synthesize_mlu.sh b/examples/csmsc/tts2/local/synthesize_mlu.sh new file mode 100755 index 000000000..6c0b0b650 --- /dev/null +++ b/examples/csmsc/tts2/local/synthesize_mlu.sh @@ -0,0 +1,90 @@ +#!/bin/bash + +config_path=$1 +train_output_path=$2 +ckpt_name=$3 +stage=0 +stop_stage=0 + +# for more GAN Vocoders +# multi band melgan +if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=mb_melgan_csmsc \ + --voc_config=mb_melgan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=mb_melgan_csmsc_ckpt_0.1.1/snapshot_iter_1000000.pdz\ + --voc_stat=mb_melgan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --test_metadata=dump/test/norm/metadata.jsonl \ + --output_dir=${train_output_path}/test \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --ngpu=0 \ + --nmlu=1 +fi + +# style melgan +if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=style_melgan_csmsc \ + --voc_config=style_melgan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=style_melgan_csmsc_ckpt_0.1.1/snapshot_iter_1500000.pdz \ + --voc_stat=style_melgan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --test_metadata=dump/test/norm/metadata.jsonl \ + --output_dir=${train_output_path}/test \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --ngpu=0 \ + --nmlu=1 +fi + +# hifigan +if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then + echo "in hifigan syn" + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=hifigan_csmsc \ + --voc_config=hifigan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=hifigan_csmsc_ckpt_0.1.1/snapshot_iter_2500000.pdz \ + --voc_stat=hifigan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --test_metadata=dump/test/norm/metadata.jsonl \ + --output_dir=${train_output_path}/test \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --ngpu=0 \ + --nmlu=1 +fi + +# wavernn +if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then + echo "in wavernn syn" + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=wavernn_csmsc \ + --voc_config=wavernn_csmsc_ckpt_0.2.0/default.yaml \ + --voc_ckpt=wavernn_csmsc_ckpt_0.2.0/snapshot_iter_400000.pdz \ + --voc_stat=wavernn_csmsc_ckpt_0.2.0/feats_stats.npy \ + --test_metadata=dump/test/norm/metadata.jsonl \ + --output_dir=${train_output_path}/test \ + --tones_dict=dump/tone_id_map.txt \ + --phones_dict=dump/phone_id_map.txt \ + --ngpu=0 \ + --nmlu=1 +fi diff --git a/examples/csmsc/tts2/local/synthesize_npu.sh b/examples/csmsc/tts2/local/synthesize_npu.sh new file mode 100755 index 000000000..90fcef83d --- /dev/null +++ b/examples/csmsc/tts2/local/synthesize_npu.sh @@ -0,0 +1,110 @@ +#!/bin/bash + +config_path=$1 +train_output_path=$2 +ckpt_name=$3 +stage=0 +stop_stage=0 + +# pwgan +if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=pwgan_csmsc \ + --voc_config=pwg_baker_ckpt_0.4/pwg_default.yaml \ + --voc_ckpt=pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz \ + --voc_stat=pwg_baker_ckpt_0.4/pwg_stats.npy \ + --test_metadata=dump/test/norm/metadata.jsonl \ + --output_dir=${train_output_path}/test \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --ngpu=0 \ + --nnpu=1 +fi + +# for more GAN Vocoders +# multi band melgan +if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=mb_melgan_csmsc \ + --voc_config=mb_melgan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=mb_melgan_csmsc_ckpt_0.1.1/snapshot_iter_1000000.pdz\ + --voc_stat=mb_melgan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --test_metadata=dump/test/norm/metadata.jsonl \ + --output_dir=${train_output_path}/test \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --ngpu=0 \ + --nnpu=1 +fi + +# style melgan +if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=style_melgan_csmsc \ + --voc_config=style_melgan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=style_melgan_csmsc_ckpt_0.1.1/snapshot_iter_1500000.pdz \ + --voc_stat=style_melgan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --test_metadata=dump/test/norm/metadata.jsonl \ + --output_dir=${train_output_path}/test \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --ngpu=0 \ + --nnpu=1 +fi + +# hifigan +if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then + echo "in hifigan syn" + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=hifigan_csmsc \ + --voc_config=hifigan_csmsc_ckpt_0.1.1/default.yaml \ + --voc_ckpt=hifigan_csmsc_ckpt_0.1.1/snapshot_iter_2500000.pdz \ + --voc_stat=hifigan_csmsc_ckpt_0.1.1/feats_stats.npy \ + --test_metadata=dump/test/norm/metadata.jsonl \ + --output_dir=${train_output_path}/test \ + --phones_dict=dump/phone_id_map.txt \ + --tones_dict=dump/tone_id_map.txt \ + --ngpu=0 \ + --nnpu=1 +fi + +# wavernn +if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then + echo "in wavernn syn" + FLAGS_allocator_strategy=naive_best_fit \ + python3 ${BIN_DIR}/../synthesize.py \ + --am=speedyspeech_csmsc \ + --am_config=${config_path} \ + --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \ + --am_stat=dump/train/feats_stats.npy \ + --voc=wavernn_csmsc \ + --voc_config=wavernn_csmsc_ckpt_0.2.0/default.yaml \ + --voc_ckpt=wavernn_csmsc_ckpt_0.2.0/snapshot_iter_400000.pdz \ + --voc_stat=wavernn_csmsc_ckpt_0.2.0/feats_stats.npy \ + --test_metadata=dump/test/norm/metadata.jsonl \ + --output_dir=${train_output_path}/test \ + --tones_dict=dump/tone_id_map.txt \ + --phones_dict=dump/phone_id_map.txt \ + --ngpu=0 \ + --nnpu=1 +fi diff --git a/examples/csmsc/tts2/local/train_mlu.sh b/examples/csmsc/tts2/local/train_mlu.sh new file mode 100755 index 000000000..4c1486434 --- /dev/null +++ b/examples/csmsc/tts2/local/train_mlu.sh @@ -0,0 +1,16 @@ + +#!/bin/bash + +config_path=$1 +train_output_path=$2 +# export MLU_VISIBLE_DEVICES=8 +python ${BIN_DIR}/train.py \ + --train-metadata=dump/train/norm/metadata.jsonl \ + --dev-metadata=dump/dev/norm/metadata.jsonl \ + --config=${config_path} \ + --output-dir=${train_output_path} \ + --ngpu=0 \ + --nmlu=2 \ + --phones-dict=dump/phone_id_map.txt \ + --tones-dict=dump/tone_id_map.txt \ + --use-relative-path=True diff --git a/examples/csmsc/tts2/local/train_npu.sh b/examples/csmsc/tts2/local/train_npu.sh new file mode 100755 index 000000000..46243e155 --- /dev/null +++ b/examples/csmsc/tts2/local/train_npu.sh @@ -0,0 +1,16 @@ + +#!/bin/bash + +config_path=$1 +train_output_path=$2 + +python ${BIN_DIR}/train.py \ + --train-metadata=dump/train/norm/metadata.jsonl \ + --dev-metadata=dump/dev/norm/metadata.jsonl \ + --config=${config_path} \ + --output-dir=${train_output_path} \ + --ngpu=0 \ + --nnpu=1 \ + --phones-dict=dump/phone_id_map.txt \ + --tones-dict=dump/tone_id_map.txt \ + --use-relative-path=True diff --git a/examples/csmsc/tts2/run_mlu.sh b/examples/csmsc/tts2/run_mlu.sh new file mode 100755 index 000000000..848e54077 --- /dev/null +++ b/examples/csmsc/tts2/run_mlu.sh @@ -0,0 +1,76 @@ +#!/bin/bash + +set -e +source path.sh +export CUSTOM_DEVICE_BLACK_LIST=elementwise_max +mlus=0 +stage=0 +stop_stage=100 + +conf_path=conf/default.yaml +train_output_path=exp/default +ckpt_name=snapshot_iter_30600.pdz + +# with the following command, you can choose the stage range you want to run +# such as `./run.sh --stage 0 --stop-stage 0` +# this can not be mixed use with `$1`, `$2` ... +source ${MAIN_ROOT}/utils/parse_options.sh || exit 1 + +if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then + # prepare data + ./local/preprocess.sh ${conf_path} || exit -1 +fi + +if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then + # train model, all `ckpt` under `train_output_path/checkpoints/` dir + FLAGS_selected_mlus=${mlus} ./local/train_mlu.sh ${conf_path} ${train_output_path} || exit -1 +fi + +if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then + # synthesize, vocoder is pwgan by default + FLAGS_selected_mlus=${mlus} ./local/synthesize_mlu.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1 +fi + +if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then + # synthesize_e2e, vocoder is pwgan by default + FLAGS_selected_mlus=${mlus} ./local/synthesize_e2e_mlu.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1 +fi + +if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then + # inference with static model + FLAGS_selected_mlus=${mlus} ./local/inference_mlu.sh ${train_output_path} || exit -1 +fi + +# paddle2onnx, please make sure the static models are in ${train_output_path}/inference first +# we have only tested the following models so far +if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then + # install paddle2onnx + pip install paddle2onnx --upgrade + ./local/paddle2onnx.sh ${train_output_path} inference inference_onnx speedyspeech_csmsc + # considering the balance between speed and quality, we recommend that you use hifigan as vocoder + ./local/paddle2onnx.sh ${train_output_path} inference inference_onnx pwgan_csmsc + # ./local/paddle2onnx.sh ${train_output_path} inference inference_onnx mb_melgan_csmsc + # ./local/paddle2onnx.sh ${train_output_path} inference inference_onnx hifigan_csmsc +fi + +# inference with onnxruntime +if [ ${stage} -le 6 ] && [ ${stop_stage} -ge 6 ]; then + ./local/ort_predict.sh ${train_output_path} +fi + +# must run after stage 3 (which stage generated static models) +if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then + ./local/export2lite.sh ${train_output_path} inference pdlite speedyspeech_csmsc x86 + ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_csmsc x86 + # ./local/export2lite.sh ${train_output_path} inference pdlite mb_melgan_csmsc x86 + # ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_csmsc x86 +fi + +if [ ${stage} -le 8 ] && [ ${stop_stage} -ge 8 ]; then + CUDA_VISIBLE_DEVICES=${gpus} ./local/lite_predict.sh ${train_output_path} || exit -1 +fi + +# PTQ_static +if [ ${stage} -le 9 ] && [ ${stop_stage} -ge 9 ]; then + CUDA_VISIBLE_DEVICES=${gpus} ./local/PTQ_static.sh ${train_output_path} speedyspeech_csmsc || exit -1 +fi diff --git a/examples/csmsc/tts2/run_npu.sh b/examples/csmsc/tts2/run_npu.sh new file mode 100644 index 000000000..f36c93f74 --- /dev/null +++ b/examples/csmsc/tts2/run_npu.sh @@ -0,0 +1,42 @@ +#!/bin/bash + +set -e +source path.sh + +npus=0 +stage=0 +stop_stage=100 + +conf_path=conf/default.yaml +train_output_path=exp/default +ckpt_name=snapshot_iter_76.pdz + +# with the following command, you can choose the stage range you want to run +# such as `./run_xpu.sh --stage 0 --stop-stage 0` +# this can not be mixed use with `$1`, `$2` ... +source ${MAIN_ROOT}/utils/parse_options.sh || exit 1 + +if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then + # prepare data + ./local/preprocess.sh ${conf_path} || exit -1 +fi + +if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then + # train model, all `ckpt` under `train_output_path/checkpoints/` dir + FLAGS_selected_npus=${npus} ./local/train_npu.sh ${conf_path} ${train_output_path} || exit -1 +fi + +if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then + # synthesize, vocoder is pwgan by default + FLAGS_selected_npus=${npus} ./local/synthesize_npu.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1 +fi + +if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then + # synthesize_e2e, vocoder is pwgan by default + FLAGS_selected_npus=${npus} ./local/synthesize_e2e_npu.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1 +fi + +if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then + # inference with static model + FLAGS_selected_npus=${npus} ./local/inference_npu.sh ${train_output_path} || exit -1 +fi diff --git a/paddlespeech/audio/transform/perturb.py b/paddlespeech/audio/transform/perturb.py index 0825caec8..757a2f1bf 100644 --- a/paddlespeech/audio/transform/perturb.py +++ b/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): diff --git a/paddlespeech/audio/utils/tensor_utils.py b/paddlespeech/audio/utils/tensor_utils.py index 93883c94d..43dcac0ac 100644 --- a/paddlespeech/audio/utils/tensor_utils.py +++ b/paddlespeech/audio/utils/tensor_utils.py @@ -248,7 +248,7 @@ def st_reverse_pad_list(ys_pad: paddle.Tensor, # >>> tensor([[ 2, 1, 0], # >>> [ 2, 1, 0], # >>> [ 0, -1, -2]]) - index = index * seq_mask + index = index * seq_mask.astype(index.dtype) # >>> index # >>> tensor([[2, 1, 0], diff --git a/paddlespeech/cli/ssl/infer.py b/paddlespeech/cli/ssl/infer.py index 9b4b02803..33cdf7637 100644 --- a/paddlespeech/cli/ssl/infer.py +++ b/paddlespeech/cli/ssl/infer.py @@ -119,6 +119,7 @@ class SSLExecutor(BaseExecutor): '--verbose', action='store_true', help='Increase logger verbosity of current task.') + self.last_call_params = None def _init_from_path(self, model_type: str=None, @@ -453,6 +454,23 @@ class SSLExecutor(BaseExecutor): Python API to call an executor. """ + current_call_params = { + "model": model, + "task": task, + "lang": lang, + "sample_rate": sample_rate, + "config": config, + "ckpt_path": ckpt_path, + "decode_method": decode_method, + "force_yes": force_yes, + "rtf": rtf, + "device": device + } + if self.last_call_params is not None and self.last_call_params != current_call_params and hasattr( + self, 'model'): + del self.model + self.last_call_params = current_call_params + audio_file = os.path.abspath(audio_file) paddle.set_device(device) self._init_from_path(model, task, lang, sample_rate, config, diff --git a/paddlespeech/cls/exps/panns/deploy/predict.py b/paddlespeech/cls/exps/panns/deploy/predict.py index f14b44215..1dd0fb531 100644 --- a/paddlespeech/cls/exps/panns/deploy/predict.py +++ b/paddlespeech/cls/exps/panns/deploy/predict.py @@ -17,14 +17,14 @@ import os import numpy as np from paddle import inference from paddle.audio.datasets import ESC50 -from paddle.audio.features import MelSpectrogram +from paddle.audio.features import LogMelSpectrogram from paddleaudio.backends import soundfile_load as load_audio from scipy.special import softmax # yapf: disable parser = argparse.ArgumentParser() parser.add_argument("--model_dir", type=str, required=True, default="./export", help="The directory to static model.") -parser.add_argument('--device', choices=['cpu', 'gpu', 'xpu'], default="gpu", help="Select which device to train model, defaults to gpu.") +parser.add_argument('--device', choices=['cpu', 'gpu', 'xpu', 'gcu'], default="gpu", help="Select which device to train model, defaults to gpu.") parser.add_argument("--wav", type=str, required=True, help="Audio file to infer.") parser.add_argument("--batch_size", type=int, default=1, help="Batch size per GPU/CPU for training.") parser.add_argument('--use_tensorrt', type=eval, default=False, choices=[True, False], help='Enable to use tensorrt to speed up.') @@ -53,7 +53,10 @@ def extract_features(files: str, **kwargs): pad_width = max_length - len(waveforms[i]) waveforms[i] = np.pad(waveforms[i], pad_width=(0, pad_width)) - feat = MelSpectrogram(waveforms[i], sr, **kwargs).transpose() + feature_extractor = LogMelSpectrogram(sr, **kwargs) + feat = feature_extractor(paddle.to_tensor(waveforms[i])) + feat = paddle.transpose(feat, perm=[1, 0]).unsqueeze(0) + feats.append(feat) return np.stack(feats, axis=0) diff --git a/paddlespeech/s2t/decoders/recog_bin.py b/paddlespeech/s2t/decoders/recog_bin.py index 37b49f3a0..829b2b4a7 100644 --- a/paddlespeech/s2t/decoders/recog_bin.py +++ b/paddlespeech/s2t/decoders/recog_bin.py @@ -21,7 +21,8 @@ import sys import configargparse import numpy as np -from distutils.util import strtobool + +from paddlespeech.utils.argparse import strtobool def get_parser(): diff --git a/paddlespeech/s2t/exps/deepspeech2/bin/export.py b/paddlespeech/s2t/exps/deepspeech2/bin/export.py index 07228e98b..762aa6f2c 100644 --- a/paddlespeech/s2t/exps/deepspeech2/bin/export.py +++ b/paddlespeech/s2t/exps/deepspeech2/bin/export.py @@ -32,9 +32,6 @@ def main(config, args): if __name__ == "__main__": parser = default_argument_parser() - # save jit model to - parser.add_argument( - "--export_path", type=str, help="path of the jit model to save") args = parser.parse_args() print_arguments(args) diff --git a/paddlespeech/s2t/exps/deepspeech2/bin/test.py b/paddlespeech/s2t/exps/deepspeech2/bin/test.py index a8e20ff93..0839cf453 100644 --- a/paddlespeech/s2t/exps/deepspeech2/bin/test.py +++ b/paddlespeech/s2t/exps/deepspeech2/bin/test.py @@ -32,9 +32,6 @@ def main(config, args): if __name__ == "__main__": parser = default_argument_parser() - # save asr result to - parser.add_argument( - "--result_file", type=str, help="path of save the asr result") args = parser.parse_args() print_arguments(args, globals()) diff --git a/paddlespeech/s2t/exps/deepspeech2/bin/test_export.py b/paddlespeech/s2t/exps/deepspeech2/bin/test_export.py index 1e07aa800..71ffa6613 100644 --- a/paddlespeech/s2t/exps/deepspeech2/bin/test_export.py +++ b/paddlespeech/s2t/exps/deepspeech2/bin/test_export.py @@ -32,12 +32,6 @@ def main(config, args): if __name__ == "__main__": parser = default_argument_parser() - # save asr result to - parser.add_argument( - "--result_file", type=str, help="path of save the asr result") - #load jit model from - parser.add_argument( - "--export_path", type=str, help="path of the jit model to save") parser.add_argument( "--enable-auto-log", action="store_true", help="use auto log") args = parser.parse_args() diff --git a/paddlespeech/s2t/exps/deepspeech2/bin/test_wav.py b/paddlespeech/s2t/exps/deepspeech2/bin/test_wav.py index 32a583b6a..d087405d5 100644 --- a/paddlespeech/s2t/exps/deepspeech2/bin/test_wav.py +++ b/paddlespeech/s2t/exps/deepspeech2/bin/test_wav.py @@ -171,10 +171,6 @@ def main(config, args): if __name__ == "__main__": parser = default_argument_parser() - parser.add_argument("--audio_file", type=str, help='audio file path') - # save asr result to - parser.add_argument( - "--result_file", type=str, help="path of save the asr result") args = parser.parse_args() print_arguments(args, globals()) if not os.path.isfile(args.audio_file): diff --git a/paddlespeech/s2t/exps/deepspeech2/model.py b/paddlespeech/s2t/exps/deepspeech2/model.py index d007a9e39..710757115 100644 --- a/paddlespeech/s2t/exps/deepspeech2/model.py +++ b/paddlespeech/s2t/exps/deepspeech2/model.py @@ -335,7 +335,12 @@ class DeepSpeech2Tester(DeepSpeech2Trainer): self.test_loader, self.config, self.args.checkpoint_path) infer_model.eval() static_model = infer_model.export() - logger.info(f"Export code: {static_model.forward.code}") + try: + logger.info(f"Export code: {static_model.forward.code}") + except: + logger.info( + f"Fail to print Export code, static_model.forward.code can not be run." + ) paddle.jit.save(static_model, self.args.export_path) diff --git a/paddlespeech/s2t/exps/whisper/test_wav.py b/paddlespeech/s2t/exps/whisper/test_wav.py index e04eec4f2..d9c32a406 100644 --- a/paddlespeech/s2t/exps/whisper/test_wav.py +++ b/paddlespeech/s2t/exps/whisper/test_wav.py @@ -27,6 +27,7 @@ from paddlespeech.s2t.models.whisper import transcribe from paddlespeech.s2t.models.whisper import Whisper from paddlespeech.s2t.training.cli import default_argument_parser from paddlespeech.s2t.utils.log import Log +from paddlespeech.utils.argparse import strtobool logger = Log(__name__).getlog() @@ -103,10 +104,7 @@ if __name__ == "__main__": parser.add_argument( "--audio_file", type=str, help="path of the input audio file") parser.add_argument( - "--debug", - type=distutils.util.strtobool, - default=False, - help="for debug.") + "--debug", type=strtobool, default=False, help="for debug.") args = parser.parse_args() config = CfgNode(new_allowed=True) diff --git a/paddlespeech/s2t/io/reader.py b/paddlespeech/s2t/io/reader.py index 5e018befb..d433a643f 100644 --- a/paddlespeech/s2t/io/reader.py +++ b/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): diff --git a/paddlespeech/s2t/models/wav2vec2/wav2vec2_ASR.py b/paddlespeech/s2t/models/wav2vec2/wav2vec2_ASR.py index a3744d340..64195defc 100755 --- a/paddlespeech/s2t/models/wav2vec2/wav2vec2_ASR.py +++ b/paddlespeech/s2t/models/wav2vec2/wav2vec2_ASR.py @@ -188,7 +188,7 @@ class Wav2vec2ASR(nn.Layer): x_lens = x.shape[1] ctc_probs = self.ctc.log_softmax(x) # (B, maxlen, vocab_size) topk_prob, topk_index = ctc_probs.topk(1, axis=2) # (B, maxlen, 1) - topk_index = topk_index.view([batch_size, x_lens]) # (B, maxlen) + topk_index = topk_index.reshape([batch_size, x_lens]) # (B, maxlen) hyps = [hyp.tolist() for hyp in topk_index] hyps = [remove_duplicates_and_blank(hyp) for hyp in hyps] diff --git a/paddlespeech/s2t/training/cli.py b/paddlespeech/s2t/training/cli.py index 741b95dff..ded2aff9f 100644 --- a/paddlespeech/s2t/training/cli.py +++ b/paddlespeech/s2t/training/cli.py @@ -16,6 +16,8 @@ import argparse import distutils from yacs.config import CfgNode +from paddlespeech.utils.argparse import strtobool + class ExtendAction(argparse.Action): """ @@ -73,7 +75,7 @@ def default_argument_parser(parser=None): '--conf', type=open, action=LoadFromFile, help="config file.") parser.add_argument( "--debug", - type=distutils.util.strtobool, + type=strtobool, default=False, help="logging with debug mode.") parser.add_argument( diff --git a/paddlespeech/s2t/utils/cli_utils.py b/paddlespeech/s2t/utils/cli_utils.py index ccb0d3c97..ab93723b4 100644 --- a/paddlespeech/s2t/utils/cli_utils.py +++ b/paddlespeech/s2t/utils/cli_utils.py @@ -16,11 +16,12 @@ import sys from collections.abc import Sequence import numpy -from distutils.util import strtobool as dist_strtobool + +from paddlespeech.utils.argparse import strtobool as dist_strtobool def strtobool(x): - # distutils.util.strtobool returns integer, but it's confusing, + # paddlespeech.utils.argparse.strtobool returns integer, but it's confusing, return bool(dist_strtobool(x)) diff --git a/paddlespeech/server/engine/tts/online/onnx/tts_engine.py b/paddlespeech/server/engine/tts/online/onnx/tts_engine.py index 9dd31a08b..c78414163 100644 --- a/paddlespeech/server/engine/tts/online/onnx/tts_engine.py +++ b/paddlespeech/server/engine/tts/online/onnx/tts_engine.py @@ -76,8 +76,8 @@ class TTSServerExecutor(TTSExecutor): version=None, # default version ) self.am_res_path = self.task_resource.res_dir - self.am_ckpt = os.path.join( - self.am_res_path, self.task_resource.res_dict['ckpt'][0]) + self.am_ckpt = os.path.join(self.am_res_path, + self.task_resource.res_dict['ckpt']) # must have phones_dict in acoustic self.phones_dict = os.path.join( self.am_res_path, @@ -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}") diff --git a/paddlespeech/t2s/exps/inference.py b/paddlespeech/t2s/exps/inference.py index 8a5269825..3edc4b63b 100644 --- a/paddlespeech/t2s/exps/inference.py +++ b/paddlespeech/t2s/exps/inference.py @@ -112,7 +112,7 @@ def parse_args(): parser.add_argument( "--device", default="gpu", - choices=["gpu", "cpu", "xpu"], + choices=["gpu", "cpu", "xpu", "npu", "mlu", "gcu"], help="Device selected for inference.", ) parser.add_argument('--cpu_threads', type=int, default=1) diff --git a/paddlespeech/t2s/exps/speedyspeech/train.py b/paddlespeech/t2s/exps/speedyspeech/train.py index c90090daa..b1916fbc4 100644 --- a/paddlespeech/t2s/exps/speedyspeech/train.py +++ b/paddlespeech/t2s/exps/speedyspeech/train.py @@ -45,15 +45,20 @@ def train_sp(args, config): # decides device type and whether to run in parallel # setup running environment correctly world_size = paddle.distributed.get_world_size() - if (not paddle.is_compiled_with_cuda()) or args.ngpu == 0: - if (not paddle.is_compiled_with_xpu()) or args.nxpu == 0: - paddle.set_device("cpu") - else: - paddle.set_device("xpu") - else: + if paddle.is_compiled_with_cuda() and args.ngpu > 0: paddle.set_device("gpu") if world_size > 1: paddle.distributed.init_parallel_env() + elif paddle.is_compiled_with_xpu() and args.nxpu > 0: + paddle.device.set_device("xpu") + elif args.nnpu > 0: + paddle.device.set_device("npu") + if world_size > 1: + paddle.distributed.init_parallel_env() + elif args.nmlu > 0: + paddle.device.set_device("mlu") + else: + paddle.set_device("cpu") # set the random seed, it is a must for multiprocess training seed_everything(config.seed) @@ -191,9 +196,25 @@ def main(): "--nxpu", type=int, default=0, - help="if nxpu == 0 and ngpu == 0, use cpu.") + help="if wish to use xpu, set ngpu == 0 and nxpu > 0, otherwise use gpu, npu, mlu or cpu." + ) parser.add_argument( - "--ngpu", type=int, default=1, help="if ngpu == 0, use cpu or xpu") + "--nnpu", + type=int, + default=0, + help="if wish to use npu, set ngpu == 0 and nnpu > 0, otherwise use gpu, xpu, mlu or cpu." + ) + parser.add_argument( + "--nmlu", + type=int, + default=1, + help="if wish to use npu, set ngpu == 0 and nmlu > 0, otherwise use gpu, xpu, npu or cpu." + ) + parser.add_argument( + "--ngpu", + type=int, + default=1, + help="if wish to use gpu, set ngpu > 0, otherwise use xpu, npu or cpu.") parser.add_argument( "--use-relative-path", diff --git a/paddlespeech/t2s/exps/syn_utils.py b/paddlespeech/t2s/exps/syn_utils.py index 9a07df64d..d29dd8110 100644 --- a/paddlespeech/t2s/exps/syn_utils.py +++ b/paddlespeech/t2s/exps/syn_utils.py @@ -591,7 +591,8 @@ def get_predictor( config = inference.Config( str(Path(model_dir) / model_file), str(Path(model_dir) / params_file)) - config.enable_memory_optim() + if paddle.__version__ <= "2.5.2" and paddle.__version__ != "0.0.0": + config.enable_memory_optim() config.switch_ir_optim(True) if device == "gpu": config.enable_use_gpu(100, device_id) diff --git a/paddlespeech/t2s/exps/synthesize.py b/paddlespeech/t2s/exps/synthesize.py index e7cf7850e..b159725e2 100644 --- a/paddlespeech/t2s/exps/synthesize.py +++ b/paddlespeech/t2s/exps/synthesize.py @@ -219,12 +219,28 @@ def parse_args(): ) # other parser.add_argument( - "--ngpu", type=int, default=1, help="if ngpu == 0, use cpu or xpu.") + "--ngpu", + type=int, + default=1, + help="if wish to use gpu, set ngpu > 0, otherwise use xpu, npu, mlu or cpu." + ) parser.add_argument( "--nxpu", type=int, default=0, - help="if wish to use xpu, set ngpu == 0 and nxpu > 0, and if ngpu == 0 and nxpu == 0, use cpu." + help="if wish to use xpu, set ngpu == 0 and nxpu > 0, otherwise use gpu, npu, mlu or cpu." + ) + parser.add_argument( + "--nnpu", + type=int, + default=0, + help="if wish to use npu, set ngpu == 0 and nnpu > 0, otherwise use gpu, xpu, mlu or cpu." + ) + parser.add_argument( + "--nmlu", + type=int, + default=0, + help="if wish to use xpu, set ngpu == 0 and nmlu > 0, otherwise use gpu, xpu, npu or cpu." ) parser.add_argument("--test_metadata", type=str, help="test metadata.") parser.add_argument("--output_dir", type=str, help="output dir.") @@ -245,10 +261,16 @@ def main(): paddle.set_device("gpu") elif args.nxpu > 0: paddle.set_device("xpu") - elif args.ngpu == 0 and args.nxpu == 0: + elif args.nnpu > 0: + paddle.set_device("npu") + elif args.nmlu > 0: + paddle.set_device("mlu") + elif args.ngpu == 0 and args.nxpu == 0 and args.nnpu == 0 and args.nmlu == 0: paddle.set_device("cpu") else: - print("ngpu or nxpu should >= 0 !") + print( + "one of ngpu, nxpu, nnpu or nmlu should be greater than 0 or all of them equal to 0" + ) evaluate(args) diff --git a/paddlespeech/t2s/exps/synthesize_e2e.py b/paddlespeech/t2s/exps/synthesize_e2e.py index c63a5fbe9..08a14b315 100644 --- a/paddlespeech/t2s/exps/synthesize_e2e.py +++ b/paddlespeech/t2s/exps/synthesize_e2e.py @@ -299,12 +299,28 @@ def parse_args(): default=None, help="dir to save inference models") parser.add_argument( - "--ngpu", type=int, default=1, help="if ngpu == 0, use cpu or xpu.") + "--ngpu", + type=int, + default=1, + help="if wish to use gpu, set ngpu > 0, otherwise use xpu, npu, mlu or cpu." + ) parser.add_argument( "--nxpu", type=int, default=0, - help="if wish to use xpu, set ngpu == 0 and nxpu > 0, and if ngpu == 0 and nxpu == 0, use cpu." + help="if wish to use xpu, set ngpu == 0 and nxpu > 0, otherwise use gpu, npu, mlu or cpu." + ) + parser.add_argument( + "--nnpu", + type=int, + default=0, + help="if wish to use npu, set ngpu == 0 and nnpu > 0, otherwise use gpu, xpu, mlu or cpu." + ) + parser.add_argument( + "--nmlu", + type=int, + default=0, + help="if wish to use xpu, set ngpu == 0 and nmlu > 0, otherwise use gpu, xpu, npu or cpu." ) parser.add_argument( "--text", @@ -339,10 +355,16 @@ def main(): paddle.set_device("gpu") elif args.nxpu > 0: paddle.set_device("xpu") - elif args.ngpu == 0 and args.nxpu == 0: + elif args.nnpu > 0: + paddle.set_device("npu") + elif args.nmlu > 0: + paddle.set_device("mlu") + elif args.ngpu == 0 and args.nxpu == 0 and args.nnpu == 0 and args.nmlu == 0: paddle.set_device("cpu") else: - print("ngpu or nxpu should >= 0 !") + print( + "one of ngpu, nxpu, nnpu or nmlu should be greater than 0 or all of them equal to 0" + ) evaluate(args) diff --git a/paddlespeech/t2s/frontend/tone_sandhi.py b/paddlespeech/t2s/frontend/tone_sandhi.py index 690f69aa2..3558064cd 100644 --- a/paddlespeech/t2s/frontend/tone_sandhi.py +++ b/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][ - 0] == seg[i + 1][0] and seg[i - 1][1] == "v": - if i - 1 < len(new_seg): - new_seg[i - - 1][0] = new_seg[i - 1][0] + "一" + new_seg[i - 1][0] - else: - new_seg.append([word, pos]) - new_seg.append([seg[i + 1][0], pos]) + if skip_next: + skip_next = False + continue + 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": + new_seg[-1] = (new_seg[-1][0] + "一" + seg[i + 1][0], new_seg[-1][1]) + skip_next = True else: - if i - 2 >= 0 and seg[i - 1][0] == "一" and seg[i - 2][ - 0] == word and pos == "v": - continue - else: - new_seg.append([word, pos]) + 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 diff --git a/paddlespeech/t2s/frontend/zh_normalization/num.py b/paddlespeech/t2s/frontend/zh_normalization/num.py index 8a54d3e63..c56563e56 100644 --- a/paddlespeech/t2s/frontend/zh_normalization/num.py +++ b/paddlespeech/t2s/frontend/zh_normalization/num.py @@ -28,7 +28,7 @@ UNITS = OrderedDict({ 8: '亿', }) -COM_QUANTIFIERS = '(封|艘|把|目|套|段|人|所|朵|匹|张|座|回|场|尾|条|个|首|阙|阵|网|炮|顶|丘|棵|只|支|袭|辆|挑|担|颗|壳|窠|曲|墙|群|腔|砣|座|客|贯|扎|捆|刀|令|打|手|罗|坡|山|岭|江|溪|钟|队|单|双|对|出|口|头|脚|板|跳|枝|件|贴|针|线|管|名|位|身|堂|课|本|页|家|户|层|丝|毫|厘|分|钱|两|斤|担|铢|石|钧|锱|忽|(千|毫|微)克|毫|厘|(公)分|分|寸|尺|丈|里|寻|常|铺|程|(千|分|厘|毫|微)米|米|撮|勺|合|升|斗|石|盘|碗|碟|叠|桶|笼|盆|盒|杯|钟|斛|锅|簋|篮|盘|桶|罐|瓶|壶|卮|盏|箩|箱|煲|啖|袋|钵|年|月|日|季|刻|时|周|天|秒|分|小时|旬|纪|岁|世|更|夜|春|夏|秋|冬|代|伏|辈|丸|泡|粒|颗|幢|堆|条|根|支|道|面|片|张|颗|块|元|(亿|千万|百万|万|千|百)|(亿|千万|百万|万|千|百|美|)元|(亿|千万|百万|万|千|百|十|)吨|(亿|千万|百万|万|千|百|)块|角|毛|分)' +COM_QUANTIFIERS = '(封|艘|把|目|套|段|人|所|朵|匹|张|座|回|场|尾|条|个|首|阙|阵|网|炮|顶|丘|棵|只|支|袭|辆|挑|担|颗|壳|窠|曲|墙|群|腔|砣|座|客|贯|扎|捆|刀|令|打|手|罗|坡|山|岭|江|溪|钟|队|单|双|对|出|口|头|脚|板|跳|枝|件|贴|针|线|管|名|位|身|堂|课|本|页|家|户|层|丝|毫|厘|分|钱|两|斤|担|铢|石|钧|锱|忽|(千|毫|微)克|毫|厘|(公)分|分|寸|尺|丈|里|寻|常|铺|程|(千|分|厘|毫|微)米|米|撮|勺|合|升|斗|石|盘|碗|碟|叠|桶|笼|盆|盒|杯|钟|斛|锅|簋|篮|盘|桶|罐|瓶|壶|卮|盏|箩|箱|煲|啖|袋|钵|年|月|日|季|刻|时|周|天|秒|分|小时|旬|纪|岁|世|更|夜|春|夏|秋|冬|代|伏|辈|丸|泡|粒|颗|幢|堆|条|根|支|道|面|片|张|颗|块|元|(亿|千万|百万|万|千|百)|(亿|千万|百万|万|千|百|美|)元|(亿|千万|百万|万|千|百|十|)吨|(亿|千万|百万|万|千|百|)块|角|毛|分|(公(里|引|丈|尺|寸|分|釐)))' # 分数表达式 RE_FRAC = re.compile(r'(-?)(\d+)/(\d+)') diff --git a/paddlespeech/t2s/frontend/zh_normalization/quantifier.py b/paddlespeech/t2s/frontend/zh_normalization/quantifier.py index 598030e43..6790d7eab 100644 --- a/paddlespeech/t2s/frontend/zh_normalization/quantifier.py +++ b/paddlespeech/t2s/frontend/zh_normalization/quantifier.py @@ -35,7 +35,9 @@ measure_dict = { "ml": "毫升", "m": "米", "mm": "毫米", - "s": "秒" + "s": "秒", + "h": "小时", + "mg": "毫克" } diff --git a/paddlespeech/t2s/modules/pqmf.py b/paddlespeech/t2s/modules/pqmf.py index 7b42409d8..cbdd6e375 100644 --- a/paddlespeech/t2s/modules/pqmf.py +++ b/paddlespeech/t2s/modules/pqmf.py @@ -17,7 +17,7 @@ import numpy as np import paddle import paddle.nn.functional as F from paddle import nn -from scipy.signal import kaiser +from scipy.signal.windows import kaiser def design_prototype_filter(taps=62, cutoff_ratio=0.142, beta=9.0): diff --git a/paddlespeech/t2s/training/trainer.py b/paddlespeech/t2s/training/trainer.py index 7f1b17de2..7631ef350 100644 --- a/paddlespeech/t2s/training/trainer.py +++ b/paddlespeech/t2s/training/trainer.py @@ -164,9 +164,10 @@ class Trainer(object): self.updater. batch_size) + "avg_ips: {:.5f} sequences/sec,".format( self.updater.batch_size / avg_batch_cost) - max_mem_reserved_str = f" max_mem_reserved: {paddle.device.cuda.max_memory_reserved()} B" - max_mem_allocated_str = f" max_mem_allocated: {paddle.device.cuda.max_memory_allocated()} B" - msg += max_mem_reserved_str + "," + max_mem_allocated_str + if paddle.device.is_compiled_with_cuda(): + max_mem_reserved_str = f" max_mem_reserved: {paddle.device.cuda.max_memory_reserved() // (1024 ** 2)} MB" + max_mem_allocated_str = f" max_mem_allocated: {paddle.device.cuda.max_memory_allocated() // (1024 ** 2)} MB" + msg += max_mem_reserved_str + "," + max_mem_allocated_str logger.info(msg) diff --git a/paddlespeech/utils/argparse.py b/paddlespeech/utils/argparse.py index aad3801ea..3ebefb826 100644 --- a/paddlespeech/utils/argparse.py +++ b/paddlespeech/utils/argparse.py @@ -18,7 +18,9 @@ from typing import Text import distutils -__all__ = ["print_arguments", "add_arguments", "get_commandline_args"] +__all__ = [ + "print_arguments", "add_arguments", "get_commandline_args", "strtobool" +] def get_commandline_args(): @@ -80,6 +82,27 @@ def print_arguments(args, info=None): print("-----------------------------------------------------------") +def strtobool(value): + """Convert a string value to an integer boolean (1 for True, 0 for False). + + The function recognizes the following strings as True (case insensitive): + - "yes" + - "true" + - "1" + + All other values are considered False. + + NOTE: After Python 3.10, the distutils module, particularly distutils.util, has been partially deprecated. To maintain compatibility with existing code, the strtobool function implemented here. + """ + if isinstance(value, bool): + return int(value) + value = value.strip().lower() + if value in ('yes', 'true', '1'): + return 1 + else: + return 0 + + def add_arguments(argname, type, default, help, argparser, **kwargs): """Add argparse's argument. @@ -91,7 +114,7 @@ def add_arguments(argname, type, default, help, argparser, **kwargs): add_argument("name", str, "Jonh", "User name.", parser) args = parser.parse_args() """ - type = distutils.util.strtobool if type == bool else type + type = strtobool if type == bool else type argparser.add_argument( "--" + argname, default=default, diff --git a/paddlespeech/vector/cluster/diarization.py b/paddlespeech/vector/cluster/diarization.py index b47b3f248..1de055c85 100644 --- a/paddlespeech/vector/cluster/diarization.py +++ b/paddlespeech/vector/cluster/diarization.py @@ -24,7 +24,6 @@ import warnings import numpy as np import scipy import sklearn -from distutils.util import strtobool from scipy import linalg from scipy import sparse from scipy.sparse.csgraph import connected_components @@ -34,6 +33,8 @@ from sklearn.cluster import SpectralClustering from sklearn.cluster._kmeans import k_means from sklearn.neighbors import kneighbors_graph +from paddlespeech.utils.argparse import strtobool + def _graph_connected_component(graph, node_id): """ diff --git a/setup.py b/setup.py index af7c4dc3d..927f7edb6 100644 --- a/setup.py +++ b/setup.py @@ -43,21 +43,23 @@ base = [ # paddleaudio align with librosa==0.8.1, which need numpy==1.23.x "numpy==1.23.5", "librosa==0.8.1", - "scipy>=1.4.0", + "scipy>=1.4.0, <=1.12.0", "loguru", - "matplotlib", + "matplotlib<=3.8.4", "nara_wpe", "onnxruntime>=1.11.0", - "opencc", + "opencc==1.1.6", "opencc-python-reimplemented", "pandas", "paddleaudio>=1.1.0", "paddlenlp>=2.4.8", + "paddlepaddle-gpu==2.5.1", "paddleslim>=2.3.4", "ppdiffusers>=0.9.0", "paddlespeech_feat", "praatio>=5.0.0, <=5.1.1", "prettytable", + "pydantic>=1.10.14, <2.0", "pypinyin<=0.44.0", "pypinyin-dict", "python-dateutil", diff --git a/tests/test_tipc/prepare.sh b/tests/test_tipc/prepare.sh index e57feda0a..7d4dd8b16 100755 --- a/tests/test_tipc/prepare.sh +++ b/tests/test_tipc/prepare.sh @@ -35,6 +35,8 @@ if [[ ${MODE} = "benchmark_train" ]];then pip install setuptools_scm #-i https://pypi.tuna.tsinghua.edu.cn/simple pip install . #-i https://pypi.tuna.tsinghua.edu.cn/simple pip install jsonlines + pip install -U scipy==1.12.0 # 高版本数据处理部分报错 + pip install -U matplotlib==3.7.1 # 高版本报错cannot import name 'get_cmap' from 'matplotlib.cm' pip list cd - if [[ ${model_name} == "conformer" ]]; then diff --git a/tests/unit/asr/deepspeech2_model_test.py b/tests/unit/asr/deepspeech2_model_test.py index 5835445d2..fd42192ea 100644 --- a/tests/unit/asr/deepspeech2_model_test.py +++ b/tests/unit/asr/deepspeech2_model_test.py @@ -48,7 +48,7 @@ class TestDeepSpeech2Model(unittest.TestCase): num_rnn_layers=3, rnn_size=1024, use_gru=False, - share_rnn_weights=False, ) + rnn_direction="forward", ) loss = model(self.audio, self.audio_len, self.text, self.text_len) self.assertEqual(loss.numel(), 1) @@ -60,7 +60,7 @@ class TestDeepSpeech2Model(unittest.TestCase): num_rnn_layers=3, rnn_size=1024, use_gru=True, - share_rnn_weights=False, ) + rnn_direction="forward", ) loss = model(self.audio, self.audio_len, self.text, self.text_len) self.assertEqual(loss.numel(), 1) @@ -72,7 +72,7 @@ class TestDeepSpeech2Model(unittest.TestCase): num_rnn_layers=3, rnn_size=1024, use_gru=False, - share_rnn_weights=True, ) + rnn_direction="bidirect", ) loss = model(self.audio, self.audio_len, self.text, self.text_len) self.assertEqual(loss.numel(), 1) @@ -84,7 +84,7 @@ class TestDeepSpeech2Model(unittest.TestCase): num_rnn_layers=3, rnn_size=1024, use_gru=True, - share_rnn_weights=True, ) + rnn_direction="bidirect", ) loss = model(self.audio, self.audio_len, self.text, self.text_len) self.assertEqual(loss.numel(), 1) @@ -96,7 +96,7 @@ class TestDeepSpeech2Model(unittest.TestCase): num_rnn_layers=3, rnn_size=1024, use_gru=False, - share_rnn_weights=False, ) + rnn_direction="forward", ) loss = model(self.audio, self.audio_len, self.text, self.text_len) self.assertEqual(loss.numel(), 1) diff --git a/tests/unit/asr/deepspeech2_online_model_test.py b/tests/unit/asr/deepspeech2_online_model_test.py index f23c49263..f7ea87b12 100644 --- a/tests/unit/asr/deepspeech2_online_model_test.py +++ b/tests/unit/asr/deepspeech2_online_model_test.py @@ -19,11 +19,11 @@ import numpy as np import paddle from paddle import inference -from paddlespeech.s2t.models.ds2_online import DeepSpeech2InferModelOnline -from paddlespeech.s2t.models.ds2_online import DeepSpeech2ModelOnline +from paddlespeech.s2t.models.ds2 import DeepSpeech2InferModel +from paddlespeech.s2t.models.ds2 import DeepSpeech2Model -class TestDeepSpeech2ModelOnline(unittest.TestCase): +class TestDeepSpeech2Model(unittest.TestCase): def setUp(self): paddle.set_device('cpu') @@ -45,7 +45,7 @@ class TestDeepSpeech2ModelOnline(unittest.TestCase): self.text_len = paddle.to_tensor(text_len, dtype='int64') def test_ds2_1(self): - model = DeepSpeech2ModelOnline( + model = DeepSpeech2Model( feat_size=self.feat_dim, dict_size=10, num_conv_layers=2, @@ -58,7 +58,7 @@ class TestDeepSpeech2ModelOnline(unittest.TestCase): self.assertEqual(loss.numel(), 1) def test_ds2_2(self): - model = DeepSpeech2ModelOnline( + model = DeepSpeech2Model( feat_size=self.feat_dim, dict_size=10, num_conv_layers=2, @@ -71,7 +71,7 @@ class TestDeepSpeech2ModelOnline(unittest.TestCase): self.assertEqual(loss.numel(), 1) def test_ds2_3(self): - model = DeepSpeech2ModelOnline( + model = DeepSpeech2Model( feat_size=self.feat_dim, dict_size=10, num_conv_layers=2, @@ -84,7 +84,7 @@ class TestDeepSpeech2ModelOnline(unittest.TestCase): self.assertEqual(loss.numel(), 1) def test_ds2_4(self): - model = DeepSpeech2ModelOnline( + model = DeepSpeech2Model( feat_size=self.feat_dim, dict_size=10, num_conv_layers=2, @@ -97,7 +97,7 @@ class TestDeepSpeech2ModelOnline(unittest.TestCase): self.assertEqual(loss.numel(), 1) def test_ds2_5(self): - model = DeepSpeech2ModelOnline( + model = DeepSpeech2Model( feat_size=self.feat_dim, dict_size=10, num_conv_layers=2, @@ -110,7 +110,7 @@ class TestDeepSpeech2ModelOnline(unittest.TestCase): self.assertEqual(loss.numel(), 1) def test_ds2_6(self): - model = DeepSpeech2ModelOnline( + model = DeepSpeech2Model( feat_size=self.feat_dim, dict_size=10, num_conv_layers=2, @@ -125,7 +125,7 @@ class TestDeepSpeech2ModelOnline(unittest.TestCase): def test_ds2_7(self): use_gru = False - model = DeepSpeech2ModelOnline( + model = DeepSpeech2Model( feat_size=self.feat_dim, dict_size=10, num_conv_layers=2, @@ -156,7 +156,7 @@ class TestDeepSpeech2ModelOnline(unittest.TestCase): def test_ds2_8(self): use_gru = True - model = DeepSpeech2ModelOnline( + model = DeepSpeech2Model( feat_size=self.feat_dim, dict_size=10, num_conv_layers=2, @@ -191,7 +191,7 @@ class TestDeepSpeech2StaticModelOnline(unittest.TestCase): export_prefix = "exp/deepspeech2_online/checkpoints/test_export" if not os.path.exists(os.path.dirname(export_prefix)): os.makedirs(os.path.dirname(export_prefix), mode=0o755) - infer_model = DeepSpeech2InferModelOnline( + infer_model = DeepSpeech2InferModel( feat_size=161, dict_size=4233, num_conv_layers=2, diff --git a/tests/unit/ci.sh b/tests/unit/ci.sh new file mode 100644 index 000000000..72b4678d6 --- /dev/null +++ b/tests/unit/ci.sh @@ -0,0 +1,36 @@ +function main(){ + set -ex + speech_ci_path=`pwd` + + echo "Start asr" + cd ${speech_ci_path}/asr + bash deepspeech2_online_model_test.sh + python error_rate_test.py + python mask_test.py + python reverse_pad_list.py + echo "End asr" + + echo "Start TTS" + cd ${speech_ci_path}/tts + python test_data_table.py + python test_enfrontend.py + python test_mixfrontend.py + echo "End TTS" + + echo "Start Vector" + cd ${speech_ci_path}/vector + python test_augment.py + echo "End Vector" + + echo "Start cli" + cd ${speech_ci_path}/cli + bash test_cli.sh + echo "End cli" + + echo "Start server" + cd ${speech_ci_path}/server/offline + bash test_server_client.sh + echo "End server" +} + +main diff --git a/tests/unit/cli/test_cli.sh b/tests/unit/cli/test_cli.sh index a7f7d11e4..3903e6597 100755 --- a/tests/unit/cli/test_cli.sh +++ b/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" diff --git a/tests/unit/doc/test_cli.md b/tests/unit/doc/test_cli.md new file mode 100644 index 000000000..34a0c016a --- /dev/null +++ b/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。 diff --git a/tests/unit/server/offline/test_server_client.sh b/tests/unit/server/offline/test_server_client.sh index dc52609c5..29bdd4032 100644 --- a/tests/unit/server/offline/test_server_client.sh +++ b/tests/unit/server/offline/test_server_client.sh @@ -1,5 +1,7 @@ #!/bin/bash # bash test_server_client.sh +## require lsof to get server pid +## apt-get install -y lsof StartService(){ # Start service diff --git a/tests/unit/server/online/tts/test_server/test_http_client.py b/tests/unit/server/online/tts/test_server/test_http_client.py index 3174e85e2..685c5ca9c 100644 --- a/tests/unit/server/online/tts/test_server/test_http_client.py +++ b/tests/unit/server/online/tts/test_server/test_http_client.py @@ -48,7 +48,7 @@ if __name__ == "__main__": parser.add_argument( "--text", type=str, - default="../../../../../../paddlespeech/t2s/exps/csmsc_test.txt", + default="../../../../../../paddlespeech/t2s/assets/csmsc_test.txt", help="text to synthesize, a 'utt_id sentence' pair per line") parser.add_argument('--spk_id', type=int, default=0, help='Speaker id') parser.add_argument('--speed', type=float, default=1.0, help='Audio speed') diff --git a/tests/unit/tts/test_data_table.py b/tests/unit/tts/test_data_table.py index 3ff5bc1af..773942a2e 100644 --- a/tests/unit/tts/test_data_table.py +++ b/tests/unit/tts/test_data_table.py @@ -11,7 +11,7 @@ # 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. -from paddlespeech.t2s.datasets.data_tabel import DataTable +from paddlespeech.t2s.datasets.data_table import DataTable def test_audio_dataset(): diff --git a/tools/Dockerfile b/tools/Dockerfile new file mode 100644 index 000000000..b2f2b3125 --- /dev/null +++ b/tools/Dockerfile @@ -0,0 +1,4 @@ +FROM registry.baidubce.com/paddlepaddle/paddle:latest-dev-cuda11.2-cudnn8.2-trt8.0-gcc82 +RUN apt-get update -y +RUN apt-get -y install libsndfile1 lsof +RUN pip3.8 install pytest-runner diff --git a/tools/extras/install_openblas.sh b/tools/extras/install_openblas.sh index 91b6444b8..9e2564e0a 100755 --- a/tools/extras/install_openblas.sh +++ b/tools/extras/install_openblas.sh @@ -19,18 +19,18 @@ fi tarball=OpenBLAS-$OPENBLAS_VERSION.tar.gz -rm -rf xianyi-OpenBLAS-* OpenBLAS OpenBLAS-*.tar.gz +rm -rf OpenMathLib-OpenBLAS-* OpenBLAS OpenBLAS-*.tar.gz if [ -d "$DOWNLOAD_DIR" ]; then cp -p "$DOWNLOAD_DIR/$tarball" . else - url=$($WGET -qO- "https://api.github.com/repos/xianyi/OpenBLAS/releases/tags/v${OPENBLAS_VERSION}" | python -c 'import sys,json;print(json.load(sys.stdin)["tarball_url"])') + url=$($WGET -qO- "https://api.github.com/repos/OpenMathLib/OpenBLAS/releases/tags/v${OPENBLAS_VERSION}" | python -c 'import sys,json;print(json.load(sys.stdin)["tarball_url"])') test -n "$url" $WGET -t3 -nv -O $tarball "$url" fi tar xzf $tarball -mv xianyi-OpenBLAS-* OpenBLAS +mv OpenMathLib-OpenBLAS-* OpenBLAS make PREFIX=$(pwd)/OpenBLAS/install USE_LOCKING=1 USE_THREAD=0 -C OpenBLAS all install if [ $? -eq 0 ]; then diff --git a/tools/pre_commit.sh b/tools/pre_commit.sh new file mode 100644 index 000000000..3a179782c --- /dev/null +++ b/tools/pre_commit.sh @@ -0,0 +1,54 @@ +set +x + +# use pre-commit 2.17 +if ! [[ $(pre-commit --version) == *"2.17.0"* ]]; then + pip install pre-commit==2.17.0 1>nul +fi + +# Install clang-format before git commit to avoid repeat installation due to +# pre-commit multi-thread running. +readonly VERSION="13.0.0" +version=$(clang-format -version) +if ! [[ $(python -V 2>&1 | awk '{print $2}' | awk -F '.' '{print $1$2}') -ge 36 ]]; then + echo "clang-format installation by pip need python version great equal 3.6, + please change the default python to higher version." + exit 1 +fi + +diff_files=$(git diff --name-only --diff-filter=ACMR ${BRANCH}) +num_diff_files=$(echo "$diff_files" | wc -l) +echo -e "diff files between pr and ${BRANCH}:\n${diff_files}" + +echo "Checking code style by pre-commit ..." +pre-commit run --files ${diff_files};check_error=$? + +if test ! -z "$(git diff)"; then + echo -e '\n************************************************************************************' + echo -e "These files have been formatted by code format hook. You should use pre-commit to \ +format them before git push." + echo -e '************************************************************************************\n' + git diff 2>&1 +fi + +echo -e '\n************************************************************************************' +if [ ${check_error} != 0 ];then + echo "Your PR code style check failed." + echo "Please install pre-commit locally and set up git hook scripts:" + echo "" + echo " pip install pre-commit==2.17.0" + echo " pre-commit install" + echo "" + if [[ $num_diff_files -le 100 ]];then + echo "Then, run pre-commit to check codestyle issues in your PR:" + echo "" + echo " pre-commit run --files" $(echo ${diff_files} | tr "\n" " ") + echo "" + fi + echo "For more information, please refer to our codestyle check guide:" + echo "https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/dev_guides/git_guides/codestyle_check_guide_cn.html" +else + echo "Your PR code style check passed." +fi +echo -e '************************************************************************************\n' + +exit ${check_error} diff --git a/utils/DER.py b/utils/DER.py index 59bcbec47..d12620199 100755 --- a/utils/DER.py +++ b/utils/DER.py @@ -28,7 +28,8 @@ import re import subprocess import numpy as np -from distutils.util import strtobool + +from paddlespeech.utils.argparse import strtobool FILE_IDS = re.compile(r"(?<=Speaker Diarization for).+(?=\*\*\*)") SCORED_SPEAKER_TIME = re.compile(r"(?<=SCORED SPEAKER TIME =)[\d.]+") diff --git a/utils/addjson.py b/utils/addjson.py index e1be7ab31..f90f7afab 100755 --- a/utils/addjson.py +++ b/utils/addjson.py @@ -11,9 +11,10 @@ import json import logging import sys -from distutils.util import strtobool from espnet.utils.cli_utils import get_commandline_args +from paddlespeech.utils.argparse import strtobool + is_python2 = sys.version_info[0] == 2 diff --git a/utils/apply-cmvn.py b/utils/apply-cmvn.py index cf91bdfcd..872d69608 100755 --- a/utils/apply-cmvn.py +++ b/utils/apply-cmvn.py @@ -4,13 +4,13 @@ import logging 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 from paddlespeech.s2t.utils.cli_writers import file_writer_helper +from paddlespeech.utils.argparse import strtobool def get_parser(): diff --git a/utils/compute-cmvn-stats.py b/utils/compute-cmvn-stats.py index 276bcd36e..763347ce8 100755 --- a/utils/compute-cmvn-stats.py +++ b/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 diff --git a/utils/copy-feats.py b/utils/copy-feats.py index dc7a70b45..8f38dc8ee 100755 --- a/utils/copy-feats.py +++ b/utils/copy-feats.py @@ -2,13 +2,12 @@ import argparse 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 from paddlespeech.s2t.utils.cli_writers import file_writer_helper +from paddlespeech.utils.argparse import strtobool def get_parser(): diff --git a/utils/feat-to-shape.py b/utils/feat-to-shape.py index bbc9242f4..e5e014ded 100755 --- a/utils/feat-to-shape.py +++ b/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 diff --git a/utils/merge_scp2json.py b/utils/merge_scp2json.py index 99db6bac8..98f6cae84 100755 --- a/utils/merge_scp2json.py +++ b/utils/merge_scp2json.py @@ -7,9 +7,8 @@ import logging import sys from io import open -from distutils.util import strtobool - from paddlespeech.s2t.utils.cli_utils import get_commandline_args +from paddlespeech.utils.argparse import strtobool PY2 = sys.version_info[0] == 2 sys.stdin = codecs.getreader("utf-8")(sys.stdin if PY2 else sys.stdin.buffer)