Merge branch 'PaddlePaddle:develop' into No9tts3

examples/csmsc/tts0/README.md

@@ -99,8 +99,10 @@ pwg_baker_ckpt_0.4
 ```
 `./local/synthesize.sh` calls `${BIN_DIR}/../synthesize.py`, which can synthesize waveform from `metadata.jsonl`.
 ```bash
-CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh ${conf_path} ${train_output_path} ${ckpt_name}
+CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name}
 ```
+`--stage` controls the vocoder model during synthesis, which can use stage `0-4` to select the vocoder to use {`pwgan`, `multi band melgan`, `style melgan`, ` hifigan`, `wavernn`}
+
 ```text
 usage: synthesize.py [-h]
                      [--am {speedyspeech_csmsc,fastspeech2_csmsc,fastspeech2_ljspeech,fastspeech2_aishell3,fastspeech2_vctk,tacotron2_csmsc,tacotron2_ljspeech,tacotron2_aishell3}]
@@ -146,9 +148,12 @@ optional arguments:
                         output dir.
 ```
 `./local/synthesize_e2e.sh` calls `${BIN_DIR}/../synthesize_e2e.py`, which can synthesize waveform from text file.
+
 ```bash
-CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh ${conf_path} ${train_output_path} ${ckpt_name}
+CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name}
 ```
+`--stage` controls the vocoder model during synthesis, which can use stage `0,1,3,4` to select the vocoder to use{`pwgan`, `multi band melgan`, `hifigan`, `wavernn`}
+
 ```text
 usage: synthesize_e2e.py [-h]
                          [--am {speedyspeech_csmsc,speedyspeech_aishell3,fastspeech2_csmsc,fastspeech2_ljspeech,fastspeech2_aishell3,fastspeech2_vctk,tacotron2_csmsc,tacotron2_ljspeech}]

examples/csmsc/tts0/run.sh

@@ -27,13 +27,15 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
 fi
 
 if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
-    # synthesize, vocoder is pwgan
-    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
+    # synthesize, vocoder is pwgan by default stage 0
+    # stage 1-4 to select the vocoder to use {multi band melgan, style melgan, hifigan, wavernn}
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
 fi
 
 if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
-    # synthesize_e2e, vocoder is pwgan
-    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
+    # synthesize_e2e, vocoder is pwgan by default stage 0
+    # stage 1,3,4 to select the vocoder to use {multi band melgan, hifigan, wavernn}
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
 fi
 
 if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then

examples/csmsc/tts2/README.md

@@ -116,8 +116,10 @@ pwg_baker_ckpt_0.4
 ```
 `./local/synthesize.sh` calls `${BIN_DIR}/../synthesize.py`, which can synthesize waveform from `metadata.jsonl`.
 ```bash
-CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh ${conf_path} ${train_output_path} ${ckpt_name}
+CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name}
 ```
+`--stage` controls the vocoder model during synthesis, which can use stage `0-4` to select the vocoder to use {`pwgan`, `multi band melgan`, `style melgan`, `hifigan`, `wavernn`}
+
 ```text
 usage: synthesize.py [-h]
                      [--am {speedyspeech_csmsc,fastspeech2_csmsc,fastspeech2_ljspeech,fastspeech2_aishell3,fastspeech2_vctk,tacotron2_csmsc,tacotron2_ljspeech,tacotron2_aishell3}]
@@ -164,8 +166,10 @@ optional arguments:
 ```
 `./local/synthesize_e2e.sh` calls `${BIN_DIR}/../synthesize_e2e.py`, which can synthesize waveform from text file.
 ```bash
-CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh ${conf_path} ${train_output_path} ${ckpt_name}
+CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name}
 ```
+`--stage` controls the vocoder model during synthesis, which can use stage `0,1,3,4` to select the vocoder to use {`pwgan`, `multi band melgan`, `hifigan`, `wavernn`}
+
 ```text
 usage: synthesize_e2e.py [-h]
                          [--am {speedyspeech_csmsc,speedyspeech_aishell3,fastspeech2_csmsc,fastspeech2_ljspeech,fastspeech2_aishell3,fastspeech2_vctk,tacotron2_csmsc,tacotron2_ljspeech}]

examples/csmsc/tts2/run.sh

@@ -27,13 +27,15 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
 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} ${ckpt_name} || exit -1
+    # synthesize, vocoder is pwgan by default stage 0
+    # use stage 1-4 to select the vocoder to use {multi band melgan, style melgan, hifigan, wavernn}
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh --stage 0 ${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} ${ckpt_name} || exit -1
+    # synthesize_e2e, vocoder is pwgan by default stage 0
+    # use stage 1,3,4 to select the vocoder to use {multi band melgan, hifigan, wavernn}
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
 fi
 
 if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then

examples/csmsc/tts3_rhy/run.sh

@@ -28,11 +28,13 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
 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} ${ckpt_name} || exit -1
+    # synthesize, vocoder is pwgan by default stage 0
+    # use stage 1-4 to select the vocoder to use {multi band melgan, style melgan, hifigan, wavernn}
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh --stage 0 ${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} ${ckpt_name} || exit -1
+    # synthesize_e2e, vocoder is pwgan by default stage 0
+    # use stage 1,3,4 to select the vocoder to use {multi band melgan, hifigan, wavernn}
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
 fi

examples/csmsc/voc1/local/synthesize_e2e.sh

@@ -8,13 +8,13 @@ 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 \
+    --am_config=fastspeech2_nosil_baker_ckpt_0.4/default.yaml \
+    --am_ckpt=fastspeech2_nosil_baker_ckpt_0.4/snapshot_iter_76000.pdz \
+    --am_stat=fastspeech2_nosil_baker_ckpt_0.4/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 \
+    --voc_config=${config_path} \
+    --voc_ckpt=${train_output_path}/checkpoints/${ckpt_name} \
+    --voc_stat=dump/train/feats_stats.npy \
     --lang=zh \
     --text=${BIN_DIR}/../../assets/sentences.txt \
     --output_dir=${train_output_path}/test_e2e \

examples/opencpop/svs1/README.md

@@ -118,6 +118,8 @@ pwgan_opencpop_ckpt_1.4.0.zip
 ```bash
 CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh ${conf_path} ${train_output_path} ${ckpt_name}
 ```
+use `pwgan` model as vocoder.
+
 ```text
 usage: synthesize.py [-h]
                      [--am {diffsinger_opencpop}]
@@ -170,8 +172,10 @@ optional arguments:
 `local/pinyin_to_phone.txt` comes from the readme of the opencpop dataset, indicating the mapping from pinyin to phonemes in opencpop.
 
 ```bash
-CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh ${conf_path} ${train_output_path} ${ckpt_name}
+CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name}
 ```
+`--stage` controls the vocoder model during synthesis, which can be `0` or `1`, use `pwgan` or `hifigan` model as vocoder.
+
 ```text
 usage: synthesize_e2e.py [-h]
                          [--am {speedyspeech_csmsc,speedyspeech_aishell3,fastspeech2_csmsc,fastspeech2_ljspeech,fastspeech2_aishell3,fastspeech2_vctk,tacotron2_csmsc,tacotron2_ljspeech}]

examples/opencpop/svs1/README_cn.md

@@ -121,6 +121,8 @@ pwgan_opencpop_ckpt_1.4.0.zip
 ```bash
 CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh ${conf_path} ${train_output_path} ${ckpt_name}
 ```
+使用 `pwgan` 模型作为声码器。
+
 ```text
 usage: synthesize.py [-h]
                      [--am {diffsinger_opencpop}]
@@ -173,8 +175,10 @@ optional arguments:
 `local/pinyin_to_phone.txt`来源于opencpop数据集中的README，表示opencpop中拼音到音素的映射。
 
 ```bash
-CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh ${conf_path} ${train_output_path} ${ckpt_name}
+CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name}
 ```
+`--stage` 用于选择合成时使用的声码器模型，取值为 `0` 或 `1`，分别对应使用 `pwgan` 或 `hifigan` 模型作为声码器。
+
 ```text
 usage: synthesize_e2e.py [-h]
                          [--am {speedyspeech_csmsc,speedyspeech_aishell3,fastspeech2_csmsc,fastspeech2_ljspeech,fastspeech2_aishell3,fastspeech2_vctk,tacotron2_csmsc,tacotron2_ljspeech}]

examples/opencpop/svs1/run.sh

@@ -32,6 +32,6 @@ if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
 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} ${ckpt_name} || exit -1
+    # synthesize_e2e, vocoder is pwgan by default, stage 1 will use hifigan as vocoder
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
 fi