Add prosody prediction in synthesize_e2e, test=tts

examples/csmsc/tts3_rhy/README.md

@@ -0,0 +1,75 @@
+([简体中文](./README_cn.md)|English)
+# This example mainly follows the FastSpeech2 with CSMSC
+This example contains code used to train a rhythm version of [Fastspeech2](https://arxiv.org/abs/2006.04558) model with [Chinese Standard Mandarin Speech Copus](https://www.data-baker.com/open_source.html).
+
+## Dataset
+### Download and Extract
+Download CSMSC from it's [Official Website](https://test.data-baker.com/data/index/TNtts/) and extract it to `~/datasets`. Then the dataset is in the directory `~/datasets/BZNSYP`.
+
+### Get MFA Result and Extract
+We use [MFA](https://github.com/MontrealCorpusTools/Montreal-Forced-Aligner) to get durations for fastspeech2.
+You can directly download the rhythm version of MFA result from here [baker_alignment_tone.zip](https://paddlespeech.bj.bcebos.com/Rhy_e2e/baker_alignment_tone.zip), or train your MFA model reference to [mfa example](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/other/mfa) of our repo.
+Remember in our repo, you should add `--rhy-with-duration` flag to obtain the rhythm information.
+
+## Get Started
+Assume the path to the dataset is `~/datasets/BZNSYP`.
+Assume the path to the MFA result of CSMSC is `./baker_alignment_tone`.
+Run the command below to
+1. **source path**.
+2. preprocess the dataset.
+3. train the model.
+4. synthesize wavs.
+    - synthesize waveform from `metadata.jsonl`.
+    - synthesize waveform from a text file.
+5. inference using the static model.
+```bash
+./run.sh
+```
+You can choose a range of stages you want to run, or set `stage` equal to `stop-stage` to use only one stage, for example, running the following command will only preprocess the dataset.
+```bash
+./run.sh --stage 0 --stop-stage 0
+```
+### Data Preprocessing
+```bash
+./local/preprocess.sh ${conf_path}
+```
+When it is done. A `dump` folder is created in the current directory. The structure of the dump folder is listed below.
+
+```text
+dump
+├── dev
+│   ├── norm
+│   └── raw
+├── phone_id_map.txt
+├── speaker_id_map.txt
+├── test
+│   ├── norm
+│   └── raw
+└── train
+    ├── energy_stats.npy
+    ├── norm
+    ├── pitch_stats.npy
+    ├── raw
+    └── speech_stats.npy
+```
+The dataset is split into 3 parts, namely `train`, `dev`, and` test`, each of which contains a `norm` and `raw` subfolder. The raw folder contains speech、pitch and energy features of each utterance, while the norm folder contains normalized ones. The statistics used to normalize features are computed from the training set, which is located in `dump/train/*_stats.npy`.
+
+Also, there is a `metadata.jsonl` in each subfolder. It is a table-like file that contains phones, text_lengths, speech_lengths, durations, the path of speech features, the path of pitch features, the path of energy features, speaker, and the id of each utterance.
+
+# More details can be refered to the example of FastSpeech2 with CSMSC(tts3)
+
+## Pretrained Model
+Pretrained FastSpeech2 model for end-to-end rhythm version:
+- [rhy_e2e_pretrain.zip](https://paddlespeech.bj.bcebos.com/Rhy_e2e/rhy_e2e_pretrain.zip)
+
+This FastSpeech2 checkpoint contains files listed below.
+```text
+rhy_e2e_pretrain
+├── default.yaml             # default config used to train fastspeech2
+├── phone_id_map.txt         # phone vocabulary file when training fastspeech2
+├── snapshot_iter_153000.pdz # model parameters and optimizer states
+├── durations.txt            # the intermediate output of preprocess.sh
+├── energy_stats.npy
+├── pitch_stats.npy
+└── speech_stats.npy         # statistics used to normalize spectrogram when training fastspeech2
+```

examples/csmsc/tts3_rhy/README_cn.md

@@ -0,0 +1,77 @@
+(简体中文|[English](./README.md))
+# 用 CSMSC 数据集训练 FastSpeech2 模型
+
+本用例包含用于训练 [Fastspeech2](https://arxiv.org/abs/2006.04558) 模型的代码，使用 [Chinese Standard Mandarin Speech Copus](https://www.data-baker.com/open_source.html) 数据集。
+
+## 数据集
+### 下载并解压
+从 [官方网站](https://test.data-baker.com/data/index/TNtts/) 下载数据集
+
+### 获取MFA结果并解压
+我们使用 [MFA](https://github.com/MontrealCorpusTools/Montreal-Forced-Aligner) 去获得 fastspeech2 的音素持续时间。
+你们可以从这里直接下载训练好的带节奏时长的 MFA 结果 [baker_alignment_tone.zip](https://paddlespeech.bj.bcebos.com/Rhy_e2e/baker_alignment_tone.zip), 或参考 [mfa example](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/other/mfa) 训练你自己的模型。
+利用 mfa repo 去训练自己的模型时，请添加 `--rhy-with-duration` 。
+
+## 开始
+假设数据集的路径是 `~/datasets/BZNSYP`.
+假设CSMSC的MFA结果路径为 `./baker_alignment_tone`.
+运行下面的命令会进行如下操作：
+
+1. **设置原路径**。
+2. 对数据集进行预处理。
+3. 训练模型
+4. 合成波形
+    - 从 `metadata.jsonl` 合成波形。
+    - 从文本文件合成波形。
+5. 使用静态模型进行推理。
+```bash
+./run.sh
+```
+您可以选择要运行的一系列阶段，或者将 `stage` 设置为 `stop-stage` 以仅使用一个阶段，例如，运行以下命令只会预处理数据集。
+```bash
+./run.sh --stage 0 --stop-stage 0
+```
+### 数据预处理
+```bash
+./local/preprocess.sh ${conf_path}
+```
+当它完成时。将在当前目录中创建 `dump` 文件夹。转储文件夹的结构如下所示。
+
+```text
+dump
+├── dev
+│   ├── norm
+│   └── raw
+├── phone_id_map.txt
+├── speaker_id_map.txt
+├── test
+│   ├── norm
+│   └── raw
+└── train
+    ├── energy_stats.npy
+    ├── norm
+    ├── pitch_stats.npy
+    ├── raw
+    └── speech_stats.npy
+```
+
+数据集分为三个部分，即 `train` 、 `dev` 和 `test` ，每个部分都包含一个 `norm` 和 `raw` 子文件夹。原始文件夹包含每个话语的语音、音调和能量特征，而 `norm` 文件夹包含规范化的特征。用于规范化特征的统计数据是从 `dump/train/*_stats.npy` 中的训练集计算出来的。
+
+此外，还有一个 `metadata.jsonl` 在每个子文件夹中。它是一个类似表格的文件，包含音素、文本长度、语音长度、持续时间、语音特征路径、音调特征路径、能量特征路径、说话人和每个话语的 id。
+
+# 更多训练细节请参考 example 下的 CSMSC(tts3)
+
+## 预训练模型
+预先训练的端到端带韵律预测的 FastSpeech2 模型：
+- [rhy_e2e_pretrain.zip](https://paddlespeech.bj.bcebos.com/Rhy_e2e/rhy_e2e_pretrain.zip)
+
+FastSpeech2检查点包含下列文件。
+```text
+fastspeech2_nosil_baker_ckpt_0.4
+├── default.yaml            # 用于训练 fastspeech2 的默认配置
+├── phone_id_map.txt        # 训练 fastspeech2 时的音素词汇文件
+├── snapshot_iter_153000.pdz # 模型参数和优化器状态
+├── durations.txt           # preprocess.sh的中间过程
+├── energy_stats.npy
+├── pitch_stats.npy
+└── speech_stats.npy        # 训练 fastspeech2 时用于规范化频谱图的统计数据

examples/csmsc/tts3_rhy/conf/default.yaml

@@ -0,0 +1 @@
+../../tts3/conf/default.yaml

examples/csmsc/tts3_rhy/local/preprocess.sh

@@ -0,0 +1 @@
+../../tts3/local/preprocess.sh

examples/csmsc/tts3_rhy/local/synthesize.sh

@@ -0,0 +1 @@
+../../tts3/local/synthesize.sh

examples/csmsc/tts3_rhy/local/synthesize_e2e.sh

@@ -0,0 +1,119 @@
+#!/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 \
+    FLAGS_fraction_of_gpu_memory_to_use=0.01 \
+    python3 ${BIN_DIR}/../synthesize_e2e.py \
+        --am=fastspeech2_csmsc \
+        --am_config=${config_path} \
+        --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \
+        --am_stat=dump/train/speech_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=${BIN_DIR}/../sentences.txt \
+        --output_dir=${train_output_path}/test_e2e \
+        --phones_dict=dump/phone_id_map.txt \
+        --inference_dir=${train_output_path}/inference \
+        --use_rhy=True
+fi
+
+# for more GAN Vocoders
+# multi band melgan
+if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
+    FLAGS_allocator_strategy=naive_best_fit \
+    FLAGS_fraction_of_gpu_memory_to_use=0.01 \
+    python3 ${BIN_DIR}/../synthesize_e2e.py \
+        --am=fastspeech2_csmsc \
+        --am_config=${config_path} \
+        --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \
+        --am_stat=dump/train/speech_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=${BIN_DIR}/../sentences.txt \
+        --output_dir=${train_output_path}/test_e2e \
+        --phones_dict=dump/phone_id_map.txt \
+        --inference_dir=${train_output_path}/inference \
+        --use_rhy=True
+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 \
+    FLAGS_fraction_of_gpu_memory_to_use=0.01 \
+    python3 ${BIN_DIR}/../synthesize_e2e.py \
+        --am=fastspeech2_csmsc \
+        --am_config=${config_path} \
+        --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \
+        --am_stat=dump/train/speech_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=${BIN_DIR}/../sentences.txt \
+        --output_dir=${train_output_path}/test_e2e \
+        --phones_dict=dump/phone_id_map.txt \
+        --use_rhy=True
+        # --inference_dir=${train_output_path}/inference
+fi
+
+# hifigan
+if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
+    echo "in hifigan syn_e2e"
+    FLAGS_allocator_strategy=naive_best_fit \
+    FLAGS_fraction_of_gpu_memory_to_use=0.01 \
+    python3 ${BIN_DIR}/../synthesize_e2e.py \
+        --am=fastspeech2_csmsc \
+        --am_config=${config_path} \
+        --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \
+        --am_stat=dump/train/speech_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=${BIN_DIR}/../sentences.txt \
+        --output_dir=${train_output_path}/test_e2e \
+        --phones_dict=dump/phone_id_map.txt \
+        --inference_dir=${train_output_path}/inference \
+        --use_rhy=True
+fi
+
+
+# wavernn
+if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
+    echo "in wavernn syn_e2e"
+    FLAGS_allocator_strategy=naive_best_fit \
+    FLAGS_fraction_of_gpu_memory_to_use=0.01 \
+    python3 ${BIN_DIR}/../synthesize_e2e.py \
+        --am=fastspeech2_csmsc \
+        --am_config=${config_path} \
+        --am_ckpt=${train_output_path}/checkpoints/${ckpt_name} \
+        --am_stat=dump/train/speech_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=${BIN_DIR}/../sentences.txt \
+        --output_dir=${train_output_path}/test_e2e \
+        --phones_dict=dump/phone_id_map.txt \
+        --inference_dir=${train_output_path}/inference \
+        --use_rhy=True
+fi

examples/csmsc/tts3_rhy/local/train.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+config_path=$1
+train_output_path=$2
+
+python3 ${BIN_DIR}/train.py \
+    --train-metadata=dump/train/norm/metadata.jsonl \
+    --dev-metadata=dump/dev/norm/metadata.jsonl \
+    --config=${config_path} \
+    --output-dir=${train_output_path} \
+    --ngpu=1 \
+    --phones-dict=dump/phone_id_map.txt

examples/csmsc/tts3_rhy/path.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+export MAIN_ROOT=`realpath ${PWD}/../../../`
+
+export PATH=${MAIN_ROOT}:${MAIN_ROOT}/utils:${PATH}
+export LC_ALL=C
+
+export PYTHONDONTWRITEBYTECODE=1
+# Use UTF-8 in Python to avoid UnicodeDecodeError when LC_ALL=C
+export PYTHONIOENCODING=UTF-8
+export PYTHONPATH=${MAIN_ROOT}:${PYTHONPATH}
+
+MODEL=fastspeech2
+export BIN_DIR=${MAIN_ROOT}/paddlespeech/t2s/exps/${MODEL}

examples/csmsc/tts3_rhy/run.sh

@@ -0,0 +1,38 @@
+#!/bin/bash
+
+set -e
+source path.sh
+
+gpus=0,1
+stage=0
+stop_stage=100
+
+conf_path=conf/default.yaml
+train_output_path=exp/default
+ckpt_name=snapshot_iter_153.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
+    ### please place the mfa result of rhythm here
+    ./local/preprocess.sh ${conf_path} || exit -1
+fi
+
+if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
+    # train model, all `ckpt` under `train_output_path/checkpoints/` dir
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/train.sh ${conf_path} ${train_output_path} || exit -1
+fi
+
+if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
+    # synthesize, vocoder is pwgan by default
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
+fi
+
+if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
+    # synthesize_e2e, vocoder is pwgan by default
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
+fi

paddlespeech/resource/pretrained_models.py

@@ -1658,3 +1658,17 @@
         },
     },
 }
+
+# ---------------------------------
+# ------------- Rhy_frontend ---------------
+# ---------------------------------
+rhy_frontend_models = {
+    'rhy_e2e': {
+        '1.0': {
+            'url':
+            'https://paddlespeech.bj.bcebos.com/Rhy_e2e/rhy_e2e_pretrain.zip',
+            'md5':
+            'd36566b835977ea05ffbd9c0210c8e3c',
+        },
+    },
+}

paddlespeech/t2s/exps/syn_utils.py

@@ -163,10 +163,13 @@ def get_test_dataset(test_metadata: List[Dict[str, Any]],
 # frontend
 def get_frontend(lang: str='zh',
                  phones_dict: Optional[os.PathLike]=None,
-                 tones_dict: Optional[os.PathLike]=None):
+                 tones_dict: Optional[os.PathLike]=None,
+                 use_rhy=False):
     if lang == 'zh':
         frontend = Frontend(
-            phone_vocab_path=phones_dict, tone_vocab_path=tones_dict)
+            phone_vocab_path=phones_dict,
+            tone_vocab_path=tones_dict,
+            use_rhy=use_rhy)
     elif lang == 'en':
         frontend = English(phone_vocab_path=phones_dict)
     elif lang == 'mix':