Merge branch 'PaddlePaddle:develop' into develop

68bd85a6 · WongLaw · GitHub · fab5b3a3 · 5e714ecb · 68bd85a6
116 changed file
--- a/demos/speech_server/conf/application.yaml
+++ b/demos/speech_server/conf/application.yaml
@@ -61,7 +61,7 @@ tts_python:
    phones_dict: 
    tones_dict: 
    speaker_dict: 
-    spk_id: 0
+

    # voc (vocoder) choices=['pwgan_csmsc', 'pwgan_ljspeech', 'pwgan_aishell3',
    #                        'pwgan_vctk', 'mb_melgan_csmsc', 'style_melgan_csmsc',
@@ -87,7 +87,7 @@ tts_inference:
    phones_dict: 
    tones_dict: 
    speaker_dict: 
-    spk_id: 0
+

    am_predictor_conf:
        device:  # set 'gpu:id' or 'cpu'

--- a/demos/streaming_tts_server/conf/tts_online_application.yaml
+++ b/demos/streaming_tts_server/conf/tts_online_application.yaml
@@ -29,7 +29,7 @@ tts_online:
    phones_dict: 
    tones_dict: 
    speaker_dict: 
-    spk_id: 0
+    

    # voc (vocoder) choices=['mb_melgan_csmsc, hifigan_csmsc']
    # Both mb_melgan_csmsc and hifigan_csmsc support streaming voc inference
@@ -70,7 +70,6 @@ tts_online-onnx:
    phones_dict: 
    tones_dict: 
    speaker_dict: 
-    spk_id: 0
    am_sample_rate: 24000
    am_sess_conf:
        device: "cpu" # set 'gpu:id' or 'cpu'

--- a/demos/streaming_tts_server/conf/tts_online_ws_application.yaml
+++ b/demos/streaming_tts_server/conf/tts_online_ws_application.yaml
@@ -29,7 +29,7 @@ tts_online:
    phones_dict: 
    tones_dict: 
    speaker_dict: 
-    spk_id: 0
+        

    # voc (vocoder) choices=['mb_melgan_csmsc, hifigan_csmsc']
    # Both mb_melgan_csmsc and hifigan_csmsc support streaming voc inference
@@ -70,7 +70,6 @@ tts_online-onnx:
    phones_dict: 
    tones_dict: 
    speaker_dict: 
-    spk_id: 0
    am_sample_rate: 24000
    am_sess_conf:
        device: "cpu" # set 'gpu:id' or 'cpu'

--- a/docs/source/api/paddlespeech.audio.rst
+++ b/docs/source/api/paddlespeech.audio.rst
@@ -20,4 +20,7 @@ Subpackages
   paddlespeech.audio.io
   paddlespeech.audio.metric
   paddlespeech.audio.sox_effects
+   paddlespeech.audio.streamdata
+   paddlespeech.audio.text
+   paddlespeech.audio.transform
   paddlespeech.audio.utils
--- a/docs/source/api/paddlespeech.audio.streamdata.autodecode.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.autodecode.rst
+paddlespeech.audio.streamdata.autodecode module
+===============================================
+
+.. automodule:: paddlespeech.audio.streamdata.autodecode
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.cache.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.cache.rst
+paddlespeech.audio.streamdata.cache module
+==========================================
+
+.. automodule:: paddlespeech.audio.streamdata.cache
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.compat.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.compat.rst
+paddlespeech.audio.streamdata.compat module
+===========================================
+
+.. automodule:: paddlespeech.audio.streamdata.compat
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.extradatasets.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.extradatasets.rst
+paddlespeech.audio.streamdata.extradatasets module
+==================================================
+
+.. automodule:: paddlespeech.audio.streamdata.extradatasets
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.filters.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.filters.rst
+paddlespeech.audio.streamdata.filters module
+============================================
+
+.. automodule:: paddlespeech.audio.streamdata.filters
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.gopen.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.gopen.rst
+paddlespeech.audio.streamdata.gopen module
+==========================================
+
+.. automodule:: paddlespeech.audio.streamdata.gopen
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.handlers.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.handlers.rst
+paddlespeech.audio.streamdata.handlers module
+=============================================
+
+.. automodule:: paddlespeech.audio.streamdata.handlers
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.mix.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.mix.rst
+paddlespeech.audio.streamdata.mix module
+========================================
+
+.. automodule:: paddlespeech.audio.streamdata.mix
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.paddle_utils.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.paddle_utils.rst
+paddlespeech.audio.streamdata.paddle\_utils module
+==================================================
+
+.. automodule:: paddlespeech.audio.streamdata.paddle_utils
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.pipeline.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.pipeline.rst
+paddlespeech.audio.streamdata.pipeline module
+=============================================
+
+.. automodule:: paddlespeech.audio.streamdata.pipeline
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.rst
+paddlespeech.audio.streamdata package
+=====================================
+
+.. automodule:: paddlespeech.audio.streamdata
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Submodules
+----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.audio.streamdata.autodecode
+   paddlespeech.audio.streamdata.cache
+   paddlespeech.audio.streamdata.compat
+   paddlespeech.audio.streamdata.extradatasets
+   paddlespeech.audio.streamdata.filters
+   paddlespeech.audio.streamdata.gopen
+   paddlespeech.audio.streamdata.handlers
+   paddlespeech.audio.streamdata.mix
+   paddlespeech.audio.streamdata.paddle_utils
+   paddlespeech.audio.streamdata.pipeline
+   paddlespeech.audio.streamdata.shardlists
+   paddlespeech.audio.streamdata.tariterators
+   paddlespeech.audio.streamdata.utils
+   paddlespeech.audio.streamdata.writer
--- a/docs/source/api/paddlespeech.audio.streamdata.shardlists.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.shardlists.rst
+paddlespeech.audio.streamdata.shardlists module
+===============================================
+
+.. automodule:: paddlespeech.audio.streamdata.shardlists
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.tariterators.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.tariterators.rst
+paddlespeech.audio.streamdata.tariterators module
+=================================================
+
+.. automodule:: paddlespeech.audio.streamdata.tariterators
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.utils.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.utils.rst
+paddlespeech.audio.streamdata.utils module
+==========================================
+
+.. automodule:: paddlespeech.audio.streamdata.utils
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.streamdata.writer.rst
+++ b/docs/source/api/paddlespeech.audio.streamdata.writer.rst
+paddlespeech.audio.streamdata.writer module
+===========================================
+
+.. automodule:: paddlespeech.audio.streamdata.writer
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.text.rst
+++ b/docs/source/api/paddlespeech.audio.text.rst
+paddlespeech.audio.text package
+===============================
+
+.. automodule:: paddlespeech.audio.text
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Submodules
+----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.audio.text.text_featurizer
+   paddlespeech.audio.text.utility
--- a/docs/source/api/paddlespeech.audio.text.text_featurizer.rst
+++ b/docs/source/api/paddlespeech.audio.text.text_featurizer.rst
+paddlespeech.audio.text.text\_featurizer module
+===============================================
+
+.. automodule:: paddlespeech.audio.text.text_featurizer
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.text.utility.rst
+++ b/docs/source/api/paddlespeech.audio.text.utility.rst
+paddlespeech.audio.text.utility module
+======================================
+
+.. automodule:: paddlespeech.audio.text.utility
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.transform.add_deltas.rst
+++ b/docs/source/api/paddlespeech.audio.transform.add_deltas.rst
+paddlespeech.audio.transform.add\_deltas module
+===============================================
+
+.. automodule:: paddlespeech.audio.transform.add_deltas
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.transform.channel_selector.rst
+++ b/docs/source/api/paddlespeech.audio.transform.channel_selector.rst
+paddlespeech.audio.transform.channel\_selector module
+=====================================================
+
+.. automodule:: paddlespeech.audio.transform.channel_selector
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.transform.cmvn.rst
+++ b/docs/source/api/paddlespeech.audio.transform.cmvn.rst
+paddlespeech.audio.transform.cmvn module
+========================================
+
+.. automodule:: paddlespeech.audio.transform.cmvn
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.transform.functional.rst
+++ b/docs/source/api/paddlespeech.audio.transform.functional.rst
+paddlespeech.audio.transform.functional module
+==============================================
+
+.. automodule:: paddlespeech.audio.transform.functional
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.transform.perturb.rst
+++ b/docs/source/api/paddlespeech.audio.transform.perturb.rst
+paddlespeech.audio.transform.perturb module
+===========================================
+
+.. automodule:: paddlespeech.audio.transform.perturb
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.transform.rst
+++ b/docs/source/api/paddlespeech.audio.transform.rst
+paddlespeech.audio.transform package
+====================================
+
+.. automodule:: paddlespeech.audio.transform
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Submodules
+----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.audio.transform.add_deltas
+   paddlespeech.audio.transform.channel_selector
+   paddlespeech.audio.transform.cmvn
+   paddlespeech.audio.transform.functional
+   paddlespeech.audio.transform.perturb
+   paddlespeech.audio.transform.spec_augment
+   paddlespeech.audio.transform.spectrogram
+   paddlespeech.audio.transform.transform_interface
+   paddlespeech.audio.transform.transformation
+   paddlespeech.audio.transform.wpe
--- a/docs/source/api/paddlespeech.audio.transform.spec_augment.rst
+++ b/docs/source/api/paddlespeech.audio.transform.spec_augment.rst
+paddlespeech.audio.transform.spec\_augment module
+=================================================
+
+.. automodule:: paddlespeech.audio.transform.spec_augment
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.transform.spectrogram.rst
+++ b/docs/source/api/paddlespeech.audio.transform.spectrogram.rst
+paddlespeech.audio.transform.spectrogram module
+===============================================
+
+.. automodule:: paddlespeech.audio.transform.spectrogram
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.transform.transform_interface.rst
+++ b/docs/source/api/paddlespeech.audio.transform.transform_interface.rst
+paddlespeech.audio.transform.transform\_interface module
+========================================================
+
+.. automodule:: paddlespeech.audio.transform.transform_interface
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.transform.transformation.rst
+++ b/docs/source/api/paddlespeech.audio.transform.transformation.rst
+paddlespeech.audio.transform.transformation module
+==================================================
+
+.. automodule:: paddlespeech.audio.transform.transformation
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.transform.wpe.rst
+++ b/docs/source/api/paddlespeech.audio.transform.wpe.rst
+paddlespeech.audio.transform.wpe module
+=======================================
+
+.. automodule:: paddlespeech.audio.transform.wpe
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.utils.check_kwargs.rst
+++ b/docs/source/api/paddlespeech.audio.utils.check_kwargs.rst
+paddlespeech.audio.utils.check\_kwargs module
+=============================================
+
+.. automodule:: paddlespeech.audio.utils.check_kwargs
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.utils.dynamic_import.rst
+++ b/docs/source/api/paddlespeech.audio.utils.dynamic_import.rst
+paddlespeech.audio.utils.dynamic\_import module
+===============================================
+
+.. automodule:: paddlespeech.audio.utils.dynamic_import
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.audio.utils.rst
+++ b/docs/source/api/paddlespeech.audio.utils.rst
@@ -12,8 +12,11 @@ Submodules
 .. toctree::
   :maxdepth: 4

+   paddlespeech.audio.utils.check_kwargs
   paddlespeech.audio.utils.download
+   paddlespeech.audio.utils.dynamic_import
   paddlespeech.audio.utils.error
   paddlespeech.audio.utils.log
   paddlespeech.audio.utils.numeric
+   paddlespeech.audio.utils.tensor_utils
   paddlespeech.audio.utils.time
--- a/docs/source/api/paddlespeech.audio.utils.tensor_utils.rst
+++ b/docs/source/api/paddlespeech.audio.utils.tensor_utils.rst
+paddlespeech.audio.utils.tensor\_utils module
+=============================================
+
+.. automodule:: paddlespeech.audio.utils.tensor_utils
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.kws.exps.mdtc.collate.rst
+++ b/docs/source/api/paddlespeech.kws.exps.mdtc.collate.rst
+paddlespeech.kws.exps.mdtc.collate module
+=========================================
+
+.. automodule:: paddlespeech.kws.exps.mdtc.collate
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.kws.exps.mdtc.compute_det.rst
+++ b/docs/source/api/paddlespeech.kws.exps.mdtc.compute_det.rst
+paddlespeech.kws.exps.mdtc.compute\_det module
+==============================================
+
+.. automodule:: paddlespeech.kws.exps.mdtc.compute_det
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.kws.exps.mdtc.plot_det_curve.rst
+++ b/docs/source/api/paddlespeech.kws.exps.mdtc.plot_det_curve.rst
+paddlespeech.kws.exps.mdtc.plot\_det\_curve module
+==================================================
+
+.. automodule:: paddlespeech.kws.exps.mdtc.plot_det_curve
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.kws.exps.mdtc.rst
+++ b/docs/source/api/paddlespeech.kws.exps.mdtc.rst
+paddlespeech.kws.exps.mdtc package
+==================================
+
+.. automodule:: paddlespeech.kws.exps.mdtc
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Submodules
+----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.kws.exps.mdtc.collate
+   paddlespeech.kws.exps.mdtc.compute_det
+   paddlespeech.kws.exps.mdtc.plot_det_curve
+   paddlespeech.kws.exps.mdtc.score
+   paddlespeech.kws.exps.mdtc.train
--- a/docs/source/api/paddlespeech.kws.exps.mdtc.score.rst
+++ b/docs/source/api/paddlespeech.kws.exps.mdtc.score.rst
+paddlespeech.kws.exps.mdtc.score module
+=======================================
+
+.. automodule:: paddlespeech.kws.exps.mdtc.score
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.kws.exps.mdtc.train.rst
+++ b/docs/source/api/paddlespeech.kws.exps.mdtc.train.rst
+paddlespeech.kws.exps.mdtc.train module
+=======================================
+
+.. automodule:: paddlespeech.kws.exps.mdtc.train
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.kws.exps.rst
+++ b/docs/source/api/paddlespeech.kws.exps.rst
+paddlespeech.kws.exps package
+=============================
+
+.. automodule:: paddlespeech.kws.exps
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Subpackages
+-----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.kws.exps.mdtc
--- a/docs/source/api/paddlespeech.kws.rst
+++ b/docs/source/api/paddlespeech.kws.rst
@@ -12,4 +12,5 @@ Subpackages
 .. toctree::
   :maxdepth: 4

+   paddlespeech.kws.exps
   paddlespeech.kws.models
--- a/docs/source/api/paddlespeech.resource.model_alias.rst
+++ b/docs/source/api/paddlespeech.resource.model_alias.rst
+paddlespeech.resource.model\_alias module
+=========================================
+
+.. automodule:: paddlespeech.resource.model_alias
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.resource.pretrained_models.rst
+++ b/docs/source/api/paddlespeech.resource.pretrained_models.rst
+paddlespeech.resource.pretrained\_models module
+===============================================
+
+.. automodule:: paddlespeech.resource.pretrained_models
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.resource.resource.rst
+++ b/docs/source/api/paddlespeech.resource.resource.rst
+paddlespeech.resource.resource module
+=====================================
+
+.. automodule:: paddlespeech.resource.resource
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.resource.rst
+++ b/docs/source/api/paddlespeech.resource.rst
+paddlespeech.resource package
+=============================
+
+.. automodule:: paddlespeech.resource
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Submodules
+----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.resource.model_alias
+   paddlespeech.resource.pretrained_models
+   paddlespeech.resource.resource
--- a/docs/source/api/paddlespeech.rst
+++ b/docs/source/api/paddlespeech.rst
@@ -16,8 +16,10 @@ Subpackages
   paddlespeech.cli
   paddlespeech.cls
   paddlespeech.kws
+   paddlespeech.resource
   paddlespeech.s2t
   paddlespeech.server
   paddlespeech.t2s
   paddlespeech.text
+   paddlespeech.utils
   paddlespeech.vector
--- a/docs/source/api/paddlespeech.s2t.rst
+++ b/docs/source/api/paddlespeech.s2t.rst
@@ -19,5 +19,4 @@ Subpackages
   paddlespeech.s2t.models
   paddlespeech.s2t.modules
   paddlespeech.s2t.training
-   paddlespeech.s2t.transform
   paddlespeech.s2t.utils
--- a/docs/source/api/paddlespeech.server.utils.rst
+++ b/docs/source/api/paddlespeech.server.utils.rst
@@ -18,7 +18,6 @@ Submodules
   paddlespeech.server.utils.config
   paddlespeech.server.utils.errors
   paddlespeech.server.utils.exception
-   paddlespeech.server.utils.log
   paddlespeech.server.utils.onnx_infer
   paddlespeech.server.utils.paddle_predictor
   paddlespeech.server.utils.util

--- a/docs/source/api/paddlespeech.t2s.datasets.rst
+++ b/docs/source/api/paddlespeech.t2s.datasets.rst
@@ -19,4 +19,5 @@ Submodules
   paddlespeech.t2s.datasets.get_feats
   paddlespeech.t2s.datasets.ljspeech
   paddlespeech.t2s.datasets.preprocess_utils
+   paddlespeech.t2s.datasets.sampler
   paddlespeech.t2s.datasets.vocoder_batch_fn
--- a/docs/source/api/paddlespeech.t2s.datasets.sampler.rst
+++ b/docs/source/api/paddlespeech.t2s.datasets.sampler.rst
+paddlespeech.t2s.datasets.sampler module
+========================================
+
+.. automodule:: paddlespeech.t2s.datasets.sampler
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.ernie_sat.align.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.ernie_sat.align.rst
+paddlespeech.t2s.exps.ernie\_sat.align module
+=============================================
+
+.. automodule:: paddlespeech.t2s.exps.ernie_sat.align
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.ernie_sat.normalize.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.ernie_sat.normalize.rst
+paddlespeech.t2s.exps.ernie\_sat.normalize module
+=================================================
+
+.. automodule:: paddlespeech.t2s.exps.ernie_sat.normalize
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.ernie_sat.preprocess.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.ernie_sat.preprocess.rst
+paddlespeech.t2s.exps.ernie\_sat.preprocess module
+==================================================
+
+.. automodule:: paddlespeech.t2s.exps.ernie_sat.preprocess
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.ernie_sat.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.ernie_sat.rst
+paddlespeech.t2s.exps.ernie\_sat package
+========================================
+
+.. automodule:: paddlespeech.t2s.exps.ernie_sat
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Submodules
+----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.t2s.exps.ernie_sat.align
+   paddlespeech.t2s.exps.ernie_sat.normalize
+   paddlespeech.t2s.exps.ernie_sat.preprocess
+   paddlespeech.t2s.exps.ernie_sat.synthesize
+   paddlespeech.t2s.exps.ernie_sat.synthesize_e2e
+   paddlespeech.t2s.exps.ernie_sat.train
+   paddlespeech.t2s.exps.ernie_sat.utils
--- a/docs/source/api/paddlespeech.t2s.exps.ernie_sat.synthesize.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.ernie_sat.synthesize.rst
+paddlespeech.t2s.exps.ernie\_sat.synthesize module
+==================================================
+
+.. automodule:: paddlespeech.t2s.exps.ernie_sat.synthesize
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.ernie_sat.synthesize_e2e.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.ernie_sat.synthesize_e2e.rst
+paddlespeech.t2s.exps.ernie\_sat.synthesize\_e2e module
+=======================================================
+
+.. automodule:: paddlespeech.t2s.exps.ernie_sat.synthesize_e2e
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.ernie_sat.train.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.ernie_sat.train.rst
+paddlespeech.t2s.exps.ernie\_sat.train module
+=============================================
+
+.. automodule:: paddlespeech.t2s.exps.ernie_sat.train
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.ernie_sat.utils.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.ernie_sat.utils.rst
+paddlespeech.t2s.exps.ernie\_sat.utils module
+=============================================
+
+.. automodule:: paddlespeech.t2s.exps.ernie_sat.utils
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.fastspeech2.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.fastspeech2.rst
@@ -16,3 +16,4 @@ Submodules
   paddlespeech.t2s.exps.fastspeech2.normalize
   paddlespeech.t2s.exps.fastspeech2.preprocess
   paddlespeech.t2s.exps.fastspeech2.train
+   paddlespeech.t2s.exps.fastspeech2.vc2_infer
--- a/docs/source/api/paddlespeech.t2s.exps.fastspeech2.vc2_infer.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.fastspeech2.vc2_infer.rst
+paddlespeech.t2s.exps.fastspeech2.vc2\_infer module
+===================================================
+
+.. automodule:: paddlespeech.t2s.exps.fastspeech2.vc2_infer
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.rst
@@ -12,11 +12,13 @@ Subpackages
 .. toctree::
   :maxdepth: 4

+   paddlespeech.t2s.exps.ernie_sat
   paddlespeech.t2s.exps.fastspeech2
   paddlespeech.t2s.exps.gan_vocoder
   paddlespeech.t2s.exps.speedyspeech
   paddlespeech.t2s.exps.tacotron2
   paddlespeech.t2s.exps.transformer_tts
+   paddlespeech.t2s.exps.vits
   paddlespeech.t2s.exps.waveflow
   paddlespeech.t2s.exps.wavernn

@@ -31,6 +33,7 @@ Submodules
   paddlespeech.t2s.exps.ort_predict
   paddlespeech.t2s.exps.ort_predict_e2e
   paddlespeech.t2s.exps.ort_predict_streaming
+   paddlespeech.t2s.exps.stream_play_tts
   paddlespeech.t2s.exps.syn_utils
   paddlespeech.t2s.exps.synthesize
   paddlespeech.t2s.exps.synthesize_e2e

--- a/docs/source/api/paddlespeech.t2s.exps.stream_play_tts.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.stream_play_tts.rst
+paddlespeech.t2s.exps.stream\_play\_tts module
+==============================================
+
+.. automodule:: paddlespeech.t2s.exps.stream_play_tts
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.vits.normalize.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.vits.normalize.rst
+paddlespeech.t2s.exps.vits.normalize module
+===========================================
+
+.. automodule:: paddlespeech.t2s.exps.vits.normalize
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.vits.preprocess.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.vits.preprocess.rst
+paddlespeech.t2s.exps.vits.preprocess module
+============================================
+
+.. automodule:: paddlespeech.t2s.exps.vits.preprocess
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.vits.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.vits.rst
+paddlespeech.t2s.exps.vits package
+==================================
+
+.. automodule:: paddlespeech.t2s.exps.vits
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Submodules
+----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.t2s.exps.vits.normalize
+   paddlespeech.t2s.exps.vits.preprocess
+   paddlespeech.t2s.exps.vits.synthesize
+   paddlespeech.t2s.exps.vits.synthesize_e2e
+   paddlespeech.t2s.exps.vits.train
+   paddlespeech.t2s.exps.vits.voice_cloning
--- a/docs/source/api/paddlespeech.t2s.exps.vits.synthesize.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.vits.synthesize.rst
+paddlespeech.t2s.exps.vits.synthesize module
+============================================
+
+.. automodule:: paddlespeech.t2s.exps.vits.synthesize
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.vits.synthesize_e2e.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.vits.synthesize_e2e.rst
+paddlespeech.t2s.exps.vits.synthesize\_e2e module
+=================================================
+
+.. automodule:: paddlespeech.t2s.exps.vits.synthesize_e2e
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.vits.train.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.vits.train.rst
+paddlespeech.t2s.exps.vits.train module
+=======================================
+
+.. automodule:: paddlespeech.t2s.exps.vits.train
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.exps.vits.voice_cloning.rst
+++ b/docs/source/api/paddlespeech.t2s.exps.vits.voice_cloning.rst
+paddlespeech.t2s.exps.vits.voice\_cloning module
+================================================
+
+.. automodule:: paddlespeech.t2s.exps.vits.voice_cloning
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.frontend.g2pw.dataset.rst
+++ b/docs/source/api/paddlespeech.t2s.frontend.g2pw.dataset.rst
+paddlespeech.t2s.frontend.g2pw.dataset module
+=============================================
+
+.. automodule:: paddlespeech.t2s.frontend.g2pw.dataset
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.frontend.g2pw.onnx_api.rst
+++ b/docs/source/api/paddlespeech.t2s.frontend.g2pw.onnx_api.rst
+paddlespeech.t2s.frontend.g2pw.onnx\_api module
+===============================================
+
+.. automodule:: paddlespeech.t2s.frontend.g2pw.onnx_api
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.frontend.g2pw.rst
+++ b/docs/source/api/paddlespeech.t2s.frontend.g2pw.rst
+paddlespeech.t2s.frontend.g2pw package
+======================================
+
+.. automodule:: paddlespeech.t2s.frontend.g2pw
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Submodules
+----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.t2s.frontend.g2pw.dataset
+   paddlespeech.t2s.frontend.g2pw.onnx_api
+   paddlespeech.t2s.frontend.g2pw.utils
--- a/docs/source/api/paddlespeech.t2s.frontend.g2pw.utils.rst
+++ b/docs/source/api/paddlespeech.t2s.frontend.g2pw.utils.rst
+paddlespeech.t2s.frontend.g2pw.utils module
+===========================================
+
+.. automodule:: paddlespeech.t2s.frontend.g2pw.utils
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.frontend.mix_frontend.rst
+++ b/docs/source/api/paddlespeech.t2s.frontend.mix_frontend.rst
+paddlespeech.t2s.frontend.mix\_frontend module
+==============================================
+
+.. automodule:: paddlespeech.t2s.frontend.mix_frontend
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.frontend.rst
+++ b/docs/source/api/paddlespeech.t2s.frontend.rst
@@ -12,6 +12,7 @@ Subpackages
 .. toctree::
   :maxdepth: 4

+   paddlespeech.t2s.frontend.g2pw
   paddlespeech.t2s.frontend.normalizer
   paddlespeech.t2s.frontend.zh_normalization

@@ -23,6 +24,7 @@ Submodules

   paddlespeech.t2s.frontend.arpabet
   paddlespeech.t2s.frontend.generate_lexicon
+   paddlespeech.t2s.frontend.mix_frontend
   paddlespeech.t2s.frontend.phonectic
   paddlespeech.t2s.frontend.punctuation
   paddlespeech.t2s.frontend.tone_sandhi

--- a/docs/source/api/paddlespeech.t2s.models.ernie_sat.ernie_sat.rst
+++ b/docs/source/api/paddlespeech.t2s.models.ernie_sat.ernie_sat.rst
+paddlespeech.t2s.models.ernie\_sat.ernie\_sat module
+====================================================
+
+.. automodule:: paddlespeech.t2s.models.ernie_sat.ernie_sat
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.models.ernie_sat.ernie_sat_updater.rst
+++ b/docs/source/api/paddlespeech.t2s.models.ernie_sat.ernie_sat_updater.rst
+paddlespeech.t2s.models.ernie\_sat.ernie\_sat\_updater module
+=============================================================
+
+.. automodule:: paddlespeech.t2s.models.ernie_sat.ernie_sat_updater
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.models.ernie_sat.rst
+++ b/docs/source/api/paddlespeech.t2s.models.ernie_sat.rst
@@ -12,4 +12,5 @@ Submodules
 .. toctree::
   :maxdepth: 4

-   paddlespeech.t2s.models.ernie_sat.mlm
+   paddlespeech.t2s.models.ernie_sat.ernie_sat
+   paddlespeech.t2s.models.ernie_sat.ernie_sat_updater
--- a/docs/source/api/paddlespeech.t2s.models.vits.monotonic_align.core.rst
+++ b/docs/source/api/paddlespeech.t2s.models.vits.monotonic_align.core.rst
+paddlespeech.t2s.models.vits.monotonic\_align.core module
+=========================================================
+
+.. automodule:: paddlespeech.t2s.models.vits.monotonic_align.core
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.t2s.models.vits.monotonic_align.rst
+++ b/docs/source/api/paddlespeech.t2s.models.vits.monotonic_align.rst
+paddlespeech.t2s.models.vits.monotonic\_align package
+=====================================================
+
+.. automodule:: paddlespeech.t2s.models.vits.monotonic_align
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Submodules
+----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.t2s.models.vits.monotonic_align.core
+   paddlespeech.t2s.models.vits.monotonic_align.setup
--- a/docs/source/api/paddlespeech.t2s.models.vits.monotonic_align.setup.rst
+++ b/docs/source/api/paddlespeech.t2s.models.vits.monotonic_align.setup.rst
+paddlespeech.t2s.models.vits.monotonic\_align.setup module
+==========================================================
+
+.. automodule:: paddlespeech.t2s.models.vits.monotonic_align.setup
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.utils.dynamic_import.rst
+++ b/docs/source/api/paddlespeech.utils.dynamic_import.rst
+paddlespeech.utils.dynamic\_import module
+=========================================
+
+.. automodule:: paddlespeech.utils.dynamic_import
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.utils.env.rst
+++ b/docs/source/api/paddlespeech.utils.env.rst
+paddlespeech.utils.env module
+=============================
+
+.. automodule:: paddlespeech.utils.env
+   :members:
+   :undoc-members:
+   :show-inheritance:
--- a/docs/source/api/paddlespeech.utils.rst
+++ b/docs/source/api/paddlespeech.utils.rst
+paddlespeech.utils package
+==========================
+
+.. automodule:: paddlespeech.utils
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Submodules
+----------
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlespeech.utils.dynamic_import
+   paddlespeech.utils.env
--- a/docs/source/index.rst
+++ b/docs/source/index.rst
@@ -74,8 +74,10 @@ Contents
   paddlespeech.cli <api/paddlespeech.cli>
   paddlespeech.cls <api/paddlespeech.cls>
   paddlespeech.kws <api/paddlespeech.kws>
+   paddlespeech.resource <api/paddlespeech.resource>
   paddlespeech.s2t <api/paddlespeech.s2t>
   paddlespeech.server <api/paddlespeech.server>
   paddlespeech.t2s <api/paddlespeech.t2s>
   paddlespeech.text <api/paddlespeech.text>
+   paddlespeech.utils <api/ppaddlespeech.utils>
   paddlespeech.vector <api/paddlespeech.vector>
--- a/docs/source/tts/tts_papers.md
+++ b/docs/source/tts/tts_papers.md
@@ -5,6 +5,7 @@
 - [Disambiguation of Chinese Polyphones in an End-to-End Framework with Semantic Features Extracted by Pre-trained BERT](https://www1.se.cuhk.edu.hk/~hccl/publications/pub/201909_INTERSPEECH_DongyangDAI.pdf)
 - [Polyphone Disambiguation in Mandarin Chinese with Semi-Supervised Learning](https://www.isca-speech.org/archive/pdfs/interspeech_2021/shi21d_interspeech.pdf)
  * github: https://github.com/PaperMechanica/SemiPPL
+- [WikipediaHomographData](https://github.com/google-research-datasets/WikipediaHomographData)
 ### Text Normalization
 #### English
 - [applenob/text_normalization](https://github.com/applenob/text_normalization)

--- a/examples/aishell3/ernie_sat/README.md
+++ b/examples/aishell3/ernie_sat/README.md
-# ERNIE-SAT with AISHELL3 dataset
+# ERNIE-SAT with VCTK dataset
+ERNIE-SAT speech-text joint pretraining framework, which achieves SOTA results in cross-lingual multi-speaker speech synthesis and cross-lingual speech editing tasks, It can be applied to a series of scenarios such as Speech Editing, personalized Speech Synthesis, and Voice Cloning.

-ERNIE-SAT 是可以同时处理中英文的跨语言的语音-语言跨模态大模型，其在语音编辑、个性化语音合成以及跨语言的语音合成等多个任务取得了领先效果。可以应用于语音编辑、个性化合成、语音克隆、同传翻译等一系列场景，该项目供研究使用。
-
-## 模型框架
-ERNIE-SAT 中我们提出了两项创新：
- 在预训练过程中将中英双语对应的音素作为输入，实现了跨语言、个性化的软音素映射
- 采用语言和语音的联合掩码学习实现了语言和语音的对齐
+## Model Framework
+In ERNIE-SAT, we propose two innovations:
+- In the pretraining process, the phonemes corresponding to Chinese and English are used as input to achieve cross-language and personalized soft phoneme mapping
+- The joint mask learning of speech and text is used to realize the alignment of speech and text

 <p align="center">
    <img src="https://user-images.githubusercontent.com/24568452/186110814-1b9c6618-a0ab-4c0c-bb3d-3d860b0e8cc2.png" />

--- a/examples/aishell3_vctk/ernie_sat/README.md
+++ b/examples/aishell3_vctk/ernie_sat/README.md
-# ERNIE-SAT with AISHELL3 and VCTK dataset
+# ERNIE-SAT with VCTK dataset
+ERNIE-SAT speech-text joint pretraining framework, which achieves SOTA results in cross-lingual multi-speaker speech synthesis and cross-lingual speech editing tasks, It can be applied to a series of scenarios such as Speech Editing, personalized Speech Synthesis, and Voice Cloning.

-ERNIE-SAT 是可以同时处理中英文的跨语言的语音-语言跨模态大模型，其在语音编辑、个性化语音合成以及跨语言的语音合成等多个任务取得了领先效果。可以应用于语音编辑、个性化合成、语音克隆、同传翻译等一系列场景，该项目供研究使用。
-
-## 模型框架
-ERNIE-SAT 中我们提出了两项创新：
- 在预训练过程中将中英双语对应的音素作为输入，实现了跨语言、个性化的软音素映射
- 采用语言和语音的联合掩码学习实现了语言和语音的对齐
+## Model Framework
+In ERNIE-SAT, we propose two innovations:
+- In the pretraining process, the phonemes corresponding to Chinese and English are used as input to achieve cross-language and personalized soft phoneme mapping
+- The joint mask learning of speech and text is used to realize the alignment of speech and text

 <p align="center">
    <img src="https://user-images.githubusercontent.com/24568452/186110814-1b9c6618-a0ab-4c0c-bb3d-3d860b0e8cc2.png" />

--- a/examples/other/tts_finetune/tts3/README.md
+++ b/examples/other/tts_finetune/tts3/README.md
@@ -75,6 +75,15 @@ When "Prepare" done. The structure of the current directory is listed below.

 ```

+### Set finetune.yaml
+`finetune.yaml` contains some configurations for fine-tuning. You can try various options to fine better result.
+Arguments:
+  - `batch_size`: finetune batch size. Default: -1, means 64 which same to pretrained model
+  - `learning_rate`: learning rate. Default: 0.0001
+  - `num_snapshots`: number of save models. Default: -1, means 5 which same to pretrained model
+  - `frozen_layers`: frozen layers. must be a list. If you don't want to frozen any layer, set []. 
+
+

 ## Get Started
 Run the command below to

--- a/examples/other/tts_finetune/tts3/finetune.py
+++ b/examples/other/tts_finetune/tts3/finetune.py
@@ -14,6 +14,7 @@
 import argparse
 import os
 from pathlib import Path
+from typing import List
 from typing import Union

 import yaml
@@ -21,10 +22,10 @@ from local.check_oov import get_check_result
 from local.extract import extract_feature
 from local.label_process import get_single_label
 from local.prepare_env import generate_finetune_env
+from local.train import train_sp
 from paddle import distributed as dist
 from yacs.config import CfgNode

-from paddlespeech.t2s.exps.fastspeech2.train import train_sp
 from utils.gen_duration_from_textgrid import gen_duration_from_textgrid

 DICT_EN = 'tools/aligner/cmudict-0.7b'
@@ -38,15 +39,24 @@ os.environ['PATH'] = MFA_PATH + '/:' + os.environ['PATH']


 class TrainArgs():
-    def __init__(self, ngpu, config_file, dump_dir: Path, output_dir: Path):
+    def __init__(self,
+                 ngpu,
+                 config_file,
+                 dump_dir: Path,
+                 output_dir: Path,
+                 frozen_layers: List[str]):
+        # config: fastspeech2 config file.
        self.config = str(config_file)
        self.train_metadata = str(dump_dir / "train/norm/metadata.jsonl")
        self.dev_metadata = str(dump_dir / "dev/norm/metadata.jsonl")
+        # model output dir.
        self.output_dir = str(output_dir)
        self.ngpu = ngpu
        self.phones_dict = str(dump_dir / "phone_id_map.txt")
        self.speaker_dict = str(dump_dir / "speaker_id_map.txt")
        self.voice_cloning = False
+        # frozen layers
+        self.frozen_layers = frozen_layers


 def get_mfa_result(
@@ -122,12 +132,11 @@ if __name__ == '__main__':
        "--ngpu", type=int, default=2, help="if ngpu=0, use cpu.")

    parser.add_argument("--epoch", type=int, default=100, help="finetune epoch")
-
    parser.add_argument(
-        "--batch_size",
-        type=int,
-        default=-1,
-        help="batch size, default -1 means same as pretrained model")
+        "--finetune_config",
+        type=str,
+        default="./finetune.yaml",
+        help="Path to finetune config file")

    args = parser.parse_args()

@@ -147,8 +156,14 @@ if __name__ == '__main__':
    with open(config_file) as f:
        config = CfgNode(yaml.safe_load(f))
    config.max_epoch = config.max_epoch + args.epoch
-    if args.batch_size > 0:
-        config.batch_size = args.batch_size
+
+    with open(args.finetune_config) as f2:
+        finetune_config = CfgNode(yaml.safe_load(f2))
+    config.batch_size = finetune_config.batch_size if finetune_config.batch_size > 0 else config.batch_size
+    config.optimizer.learning_rate = finetune_config.learning_rate if finetune_config.learning_rate > 0 else config.optimizer.learning_rate
+    config.num_snapshots = finetune_config.num_snapshots if finetune_config.num_snapshots > 0 else config.num_snapshots
+    frozen_layers = finetune_config.frozen_layers
+    assert type(frozen_layers) == list, "frozen_layers should be set a list."

    if args.lang == 'en':
        lexicon_file = DICT_EN
@@ -158,6 +173,13 @@ if __name__ == '__main__':
        mfa_phone_file = MFA_PHONE_ZH
    else:
        print('please input right lang!!')
+
+    print(f"finetune max_epoch: {config.max_epoch}")
+    print(f"finetune batch_size: {config.batch_size}")
+    print(f"finetune learning_rate: {config.optimizer.learning_rate}")
+    print(f"finetune num_snapshots: {config.num_snapshots}")
+    print(f"finetune frozen_layers: {frozen_layers}")
+
    am_phone_file = pretrained_model_dir / "phone_id_map.txt"
    label_file = input_dir / "labels.txt"

@@ -181,7 +203,8 @@ if __name__ == '__main__':
    generate_finetune_env(output_dir, pretrained_model_dir)

    # create a new args for training
-    train_args = TrainArgs(args.ngpu, config_file, dump_dir, output_dir)
+    train_args = TrainArgs(args.ngpu, config_file, dump_dir, output_dir,
+                           frozen_layers)

    # finetune models
    # dispatch

--- a/examples/other/tts_finetune/tts3/finetune.yaml
+++ b/examples/other/tts_finetune/tts3/finetune.yaml
+###########################################################
+#                 PARAS SETTING               #
+###########################################################
+# Set to -1 to indicate that the parameter is the same as the pretrained model configuration
+
+batch_size: -1
+learning_rate: 0.0001     # learning rate
+num_snapshots: -1
+
+# frozen_layers should be a list
+# if you don't need to freeze, set frozen_layers to []
+frozen_layers: ["encoder", "duration_predictor"]
--- a/examples/other/tts_finetune/tts3/local/extract.py
+++ b/examples/other/tts_finetune/tts3/local/extract.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import logging
-import math
 import os
 from operator import itemgetter
 from pathlib import Path
@@ -211,9 +210,9 @@ def extract_feature(duration_file: str,
    mel_extractor, pitch_extractor, energy_extractor = get_extractor(config)

    wav_files = sorted(list((input_dir).rglob("*.wav")))
-    # split data into 3 sections, train: 80%, dev: 10%, test: 10%
-    num_train = math.ceil(len(wav_files) * 0.8)
-    num_dev = math.ceil(len(wav_files) * 0.1)
+    # split data into 3 sections, train: len(wav_files) - 2, dev: 1, test: 1
+    num_train = len(wav_files) - 2
+    num_dev = 1
    print(num_train, num_dev)

    train_wav_files = wav_files[:num_train]

--- a/examples/other/tts_finetune/tts3/local/train.py
+++ b/examples/other/tts_finetune/tts3/local/train.py
+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import logging
+import os
+import shutil
+from pathlib import Path
+from typing import List
+
+import jsonlines
+import numpy as np
+import paddle
+from paddle import DataParallel
+from paddle import distributed as dist
+from paddle.io import DataLoader
+from paddle.io import DistributedBatchSampler
+
+from paddlespeech.t2s.datasets.am_batch_fn import fastspeech2_multi_spk_batch_fn
+from paddlespeech.t2s.datasets.am_batch_fn import fastspeech2_single_spk_batch_fn
+from paddlespeech.t2s.datasets.data_table import DataTable
+from paddlespeech.t2s.models.fastspeech2 import FastSpeech2
+from paddlespeech.t2s.models.fastspeech2 import FastSpeech2Evaluator
+from paddlespeech.t2s.models.fastspeech2 import FastSpeech2Updater
+from paddlespeech.t2s.training.extensions.snapshot import Snapshot
+from paddlespeech.t2s.training.extensions.visualizer import VisualDL
+from paddlespeech.t2s.training.optimizer import build_optimizers
+from paddlespeech.t2s.training.seeding import seed_everything
+from paddlespeech.t2s.training.trainer import Trainer
+
+
+def freeze_layer(model, layers: List[str]):
+    """freeze layers
+
+    Args:
+        layers (List[str]): frozen layers
+    """
+    for layer in layers:
+        for param in eval("model." + layer + ".parameters()"):
+            param.trainable = False
+
+
+def train_sp(args, config):
+    # decides device type and whether to run in parallel
+    # setup running environment correctly
+    if (not paddle.is_compiled_with_cuda()) or args.ngpu == 0:
+        paddle.set_device("cpu")
+    else:
+        paddle.set_device("gpu")
+    world_size = paddle.distributed.get_world_size()
+    if world_size > 1:
+        paddle.distributed.init_parallel_env()
+
+    # set the random seed, it is a must for multiprocess training
+    seed_everything(config.seed)
+
+    print(
+        f"rank: {dist.get_rank()}, pid: {os.getpid()}, parent_pid: {os.getppid()}",
+    )
+    fields = [
+        "text", "text_lengths", "speech", "speech_lengths", "durations",
+        "pitch", "energy"
+    ]
+    converters = {"speech": np.load, "pitch": np.load, "energy": np.load}
+    spk_num = None
+    if args.speaker_dict is not None:
+        print("multiple speaker fastspeech2!")
+        collate_fn = fastspeech2_multi_spk_batch_fn
+        with open(args.speaker_dict, 'rt') as f:
+            spk_id = [line.strip().split() for line in f.readlines()]
+        spk_num = len(spk_id)
+        fields += ["spk_id"]
+    elif args.voice_cloning:
+        print("Training voice cloning!")
+        collate_fn = fastspeech2_multi_spk_batch_fn
+        fields += ["spk_emb"]
+        converters["spk_emb"] = np.load
+    else:
+        print("single speaker fastspeech2!")
+        collate_fn = fastspeech2_single_spk_batch_fn
+    print("spk_num:", spk_num)
+
+    # dataloader has been too verbose
+    logging.getLogger("DataLoader").disabled = True
+
+    # construct dataset for training and validation
+    with jsonlines.open(args.train_metadata, 'r') as reader:
+        train_metadata = list(reader)
+    train_dataset = DataTable(
+        data=train_metadata,
+        fields=fields,
+        converters=converters, )
+    with jsonlines.open(args.dev_metadata, 'r') as reader:
+        dev_metadata = list(reader)
+    dev_dataset = DataTable(
+        data=dev_metadata,
+        fields=fields,
+        converters=converters, )
+
+    # collate function and dataloader
+
+    train_sampler = DistributedBatchSampler(
+        train_dataset,
+        batch_size=config.batch_size,
+        shuffle=True,
+        drop_last=True)
+
+    print("samplers done!")
+
+    train_dataloader = DataLoader(
+        train_dataset,
+        batch_sampler=train_sampler,
+        collate_fn=collate_fn,
+        num_workers=config.num_workers)
+
+    dev_dataloader = DataLoader(
+        dev_dataset,
+        shuffle=False,
+        drop_last=False,
+        batch_size=config.batch_size,
+        collate_fn=collate_fn,
+        num_workers=config.num_workers)
+    print("dataloaders done!")
+
+    with open(args.phones_dict, "r") as f:
+        phn_id = [line.strip().split() for line in f.readlines()]
+    vocab_size = len(phn_id)
+    print("vocab_size:", vocab_size)
+
+    odim = config.n_mels
+    model = FastSpeech2(
+        idim=vocab_size, odim=odim, spk_num=spk_num, **config["model"])
+
+    # freeze layer
+    if args.frozen_layers != []:
+        freeze_layer(model, args.frozen_layers)
+
+    if world_size > 1:
+        model = DataParallel(model)
+    print("model done!")
+
+    optimizer = build_optimizers(model, **config["optimizer"])
+    print("optimizer done!")
+
+    output_dir = Path(args.output_dir)
+    output_dir.mkdir(parents=True, exist_ok=True)
+    if dist.get_rank() == 0:
+        config_name = args.config.split("/")[-1]
+        # copy conf to output_dir
+        shutil.copyfile(args.config, output_dir / config_name)
+
+    updater = FastSpeech2Updater(
+        model=model,
+        optimizer=optimizer,
+        dataloader=train_dataloader,
+        output_dir=output_dir,
+        **config["updater"])
+
+    trainer = Trainer(updater, (config.max_epoch, 'epoch'), output_dir)
+
+    evaluator = FastSpeech2Evaluator(
+        model, dev_dataloader, output_dir=output_dir, **config["updater"])
+
+    if dist.get_rank() == 0:
+        trainer.extend(evaluator, trigger=(1, "epoch"))
+        trainer.extend(VisualDL(output_dir), trigger=(1, "iteration"))
+    trainer.extend(
+        Snapshot(max_size=config.num_snapshots), trigger=(1, 'epoch'))
+    trainer.run()
--- a/examples/other/tts_finetune/tts3/run.sh
+++ b/examples/other/tts_finetune/tts3/run.sh
@@ -10,11 +10,12 @@ mfa_dir=./mfa_result
 dump_dir=./dump
 output_dir=./exp/default
 lang=zh
-ngpu=2
+ngpu=1
+finetune_config=./finetune.yaml

-ckpt=snapshot_iter_96600
+ckpt=snapshot_iter_96699

-gpus=0,1
+gpus=1
 CUDA_VISIBLE_DEVICES=${gpus}
 stage=0
 stop_stage=100
@@ -35,7 +36,8 @@ if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
        --output_dir=${output_dir} \
        --lang=${lang} \
        --ngpu=${ngpu} \
-        --epoch=100
+        --epoch=100 \
+        --finetune_config=${finetune_config}
 fi


@@ -54,7 +56,7 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
        --voc_stat=pretrained_models/hifigan_aishell3_ckpt_0.2.0/feats_stats.npy \
        --lang=zh \
        --text=${BIN_DIR}/../sentences.txt \
-        --output_dir=./test_e2e \
+        --output_dir=./test_e2e/ \
        --phones_dict=${dump_dir}/phone_id_map.txt \
        --speaker_dict=${dump_dir}/speaker_id_map.txt \
        --spk_id=0 

--- a/examples/vctk/ernie_sat/README.md
+++ b/examples/vctk/ernie_sat/README.md
 # ERNIE-SAT with VCTK dataset
+ERNIE-SAT speech-text joint pretraining framework, which achieves SOTA results in cross-lingual multi-speaker speech synthesis and cross-lingual speech editing tasks, It can be applied to a series of scenarios such as Speech Editing, personalized Speech Synthesis, and Voice Cloning.

-ERNIE-SAT 是可以同时处理中英文的跨语言的语音-语言跨模态大模型，其在语音编辑、个性化语音合成以及跨语言的语音合成等多个任务取得了领先效果。可以应用于语音编辑、个性化合成、语音克隆、同传翻译等一系列场景，该项目供研究使用。
-
-## 模型框架
-ERNIE-SAT 中我们提出了两项创新：
- 在预训练过程中将中英双语对应的音素作为输入，实现了跨语言、个性化的软音素映射
- 采用语言和语音的联合掩码学习实现了语言和语音的对齐
+## Model Framework
+In ERNIE-SAT, we propose two innovations:
+- In the pretraining process, the phonemes corresponding to Chinese and English are used as input to achieve cross-language and personalized soft phoneme mapping
+- The joint mask learning of speech and text is used to realize the alignment of speech and text

 <p align="center">
    <img src="https://user-images.githubusercontent.com/24568452/186110814-1b9c6618-a0ab-4c0c-bb3d-3d860b0e8cc2.png" />

--- a/examples/wenetspeech/asr1/RESULTS.md
+++ b/examples/wenetspeech/asr1/RESULTS.md
@@ -46,3 +46,10 @@ Pretrain model from https://paddlespeech.bj.bcebos.com/s2t/wenetspeech/asr1/asr1
 | conformer | 32.52 M | conf/chunk_conformer.yaml | spec_aug  | aishell1 | ctc_greedy_search | 16 | 0.078918 |  
 | conformer | 32.52 M | conf/chunk_conformer.yaml | spec_aug  | aishell1 | ctc_prefix_beam_search | 16 | 0.079080 |  
 | conformer | 32.52 M | conf/chunk_conformer.yaml | spec_aug  | aishell1 | attention_rescoring | 16 | 0.054401 |
+
+| Model | Params | Config | Augmentation| Test set | Decode method | Chunk Size | CER |  
+| --- | --- | --- | --- | --- | --- | --- | --- |
+| conformer | 32.52 M | conf/chunk_conformer.yaml | spec_aug  | aishell1 | attention | -1 | 0.050767 |  
+| conformer | 32.52 M | conf/chunk_conformer.yaml | spec_aug  | aishell1 | ctc_greedy_search | -1 | 0.061884 |  
+| conformer | 32.52 M | conf/chunk_conformer.yaml | spec_aug  | aishell1 | ctc_prefix_beam_search | -1 | 0.062056 |  
+| conformer | 32.52 M | conf/chunk_conformer.yaml | spec_aug  | aishell1 | attention_rescoring | -1 |  0.052110 |
--- a/paddlespeech/server/engine/engine_warmup.py
+++ b/paddlespeech/server/engine/engine_warmup.py
@@ -27,8 +27,10 @@ def warm_up(engine_and_type: str, warm_up_time: int=3) -> bool:
            sentence = "您好，欢迎使用语音合成服务。"
        elif tts_engine.lang == 'en':
            sentence = "Hello and welcome to the speech synthesis service."
+        elif tts_engine.lang == 'mix':
+            sentence = "您好，欢迎使用TTS多语种服务。"
        else:
-            logger.error("tts engine only support lang: zh or en.")
+            logger.error("tts engine only support lang: zh or en or mix.")
            sys.exit(-1)

        if engine_and_type == "tts_python":

--- a/paddlespeech/t2s/frontend/g2pw/__init__.py
+++ b/paddlespeech/t2s/frontend/g2pw/__init__.py
-from paddlespeech.t2s.frontend.g2pw.onnx_api import G2PWOnnxConverter
+from .onnx_api import G2PWOnnxConverter
--- a/paddlespeech/t2s/frontend/g2pw/dataset.py
+++ b/paddlespeech/t2s/frontend/g2pw/dataset.py
@@ -15,6 +15,10 @@
 Credits
    This code is modified from https://github.com/GitYCC/g2pW
 """
+from typing import Dict
+from typing import List
+from typing import Tuple
+
 import numpy as np

 from paddlespeech.t2s.frontend.g2pw.utils import tokenize_and_map
@@ -23,22 +27,17 @@ ANCHOR_CHAR = '▁'


 def prepare_onnx_input(tokenizer,
-                       labels,
-                       char2phonemes,
-                       chars,
-                       texts,
-                       query_ids,
-                       phonemes=None,
-                       pos_tags=None,
-                       use_mask=False,
-                       use_char_phoneme=False,
-                       use_pos=False,
-                       window_size=None,
-                       max_len=512):
+                       labels: List[str],
+                       char2phonemes: Dict[str, List[int]],
+                       chars: List[str],
+                       texts: List[str],
+                       query_ids: List[int],
+                       use_mask: bool=False,
+                       window_size: int=None,
+                       max_len: int=512) -> Dict[str, np.array]:
    if window_size is not None:
-        truncated_texts, truncated_query_ids = _truncate_texts(window_size,
-                                                               texts, query_ids)
-
+        truncated_texts, truncated_query_ids = _truncate_texts(
+            window_size=window_size, texts=texts, query_ids=query_ids)
    input_ids = []
    token_type_ids = []
    attention_masks = []
@@ -51,13 +50,19 @@ def prepare_onnx_input(tokenizer,
        query_id = (truncated_query_ids if window_size else query_ids)[idx]

        try:
-            tokens, text2token, token2text = tokenize_and_map(tokenizer, text)
+            tokens, text2token, token2text = tokenize_and_map(
+                tokenizer=tokenizer, text=text)
        except Exception:
            print(f'warning: text "{text}" is invalid')
            return {}

        text, query_id, tokens, text2token, token2text = _truncate(
-            max_len, text, query_id, tokens, text2token, token2text)
+            max_len=max_len,
+            text=text,
+            query_id=query_id,
+            tokens=tokens,
+            text2token=text2token,
+            token2text=token2text)

        processed_tokens = ['[CLS]'] + tokens + ['[SEP]']

@@ -91,7 +96,8 @@ def prepare_onnx_input(tokenizer,
    return outputs


-def _truncate_texts(window_size, texts, query_ids):
+def _truncate_texts(window_size: int, texts: List[str],
+                    query_ids: List[int]) -> Tuple[List[str], List[int]]:
    truncated_texts = []
    truncated_query_ids = []
    for text, query_id in zip(texts, query_ids):
@@ -105,7 +111,12 @@ def _truncate_texts(window_size, texts, query_ids):
    return truncated_texts, truncated_query_ids


-def _truncate(max_len, text, query_id, tokens, text2token, token2text):
+def _truncate(max_len: int,
+              text: str,
+              query_id: int,
+              tokens: List[str],
+              text2token: List[int],
+              token2text: List[Tuple[int]]):
    truncate_len = max_len - 2
    if len(tokens) <= truncate_len:
        return (text, query_id, tokens, text2token, token2text)
@@ -132,18 +143,8 @@ def _truncate(max_len, text, query_id, tokens, text2token, token2text):
    ], [(s - start, e - start) for s, e in token2text[token_start:token_end]])


-def prepare_data(sent_path, lb_path=None):
-    raw_texts = open(sent_path).read().rstrip().split('\n')
-    query_ids = [raw.index(ANCHOR_CHAR) for raw in raw_texts]
-    texts = [raw.replace(ANCHOR_CHAR, '') for raw in raw_texts]
-    if lb_path is None:
-        return texts, query_ids
-    else:
-        phonemes = open(lb_path).read().rstrip().split('\n')
-        return texts, query_ids, phonemes
-
-
-def get_phoneme_labels(polyphonic_chars):
+def get_phoneme_labels(polyphonic_chars: List[List[str]]
+                       ) -> Tuple[List[str], Dict[str, List[int]]]:
    labels = sorted(list(set([phoneme for char, phoneme in polyphonic_chars])))
    char2phonemes = {}
    for char, phoneme in polyphonic_chars:
@@ -153,7 +154,8 @@ def get_phoneme_labels(polyphonic_chars):
    return labels, char2phonemes


-def get_char_phoneme_labels(polyphonic_chars):
+def get_char_phoneme_labels(polyphonic_chars: List[List[str]]
+                            ) -> Tuple[List[str], Dict[str, List[int]]]:
    labels = sorted(
        list(set([f'{char} {phoneme}' for char, phoneme in polyphonic_chars])))
    char2phonemes = {}

--- a/paddlespeech/t2s/frontend/g2pw/onnx_api.py
+++ b/paddlespeech/t2s/frontend/g2pw/onnx_api.py
@@ -17,6 +17,10 @@ Credits
 """
 import json
 import os
+from typing import Any
+from typing import Dict
+from typing import List
+from typing import Tuple

 import numpy as np
 import onnxruntime
@@ -37,7 +41,8 @@ from paddlespeech.utils.env import MODEL_HOME
 model_version = '1.1'


-def predict(session, onnx_input, labels):
+def predict(session, onnx_input: Dict[str, Any],
+            labels: List[str]) -> Tuple[List[str], List[float]]:
    all_preds = []
    all_confidences = []
    probs = session.run([], {
@@ -61,10 +66,10 @@ def predict(session, onnx_input, labels):

 class G2PWOnnxConverter:
    def __init__(self,
-                 model_dir=MODEL_HOME,
-                 style='bopomofo',
-                 model_source=None,
-                 enable_non_tradional_chinese=False):
+                 model_dir: os.PathLike=MODEL_HOME,
+                 style: str='bopomofo',
+                 model_source: str=None,
+                 enable_non_tradional_chinese: bool=False):
        uncompress_path = download_and_decompress(
            g2pw_onnx_models['G2PWModel'][model_version], model_dir)

@@ -76,7 +81,8 @@ class G2PWOnnxConverter:
            os.path.join(uncompress_path, 'g2pW.onnx'),
            sess_options=sess_options)
        self.config = load_config(
-            os.path.join(uncompress_path, 'config.py'), use_default=True)
+            config_path=os.path.join(uncompress_path, 'config.py'),
+            use_default=True)

        self.model_source = model_source if model_source else self.config.model_source
        self.enable_opencc = enable_non_tradional_chinese
@@ -103,9 +109,9 @@ class G2PWOnnxConverter:
            .strip().split('\n')
        ]
        self.labels, self.char2phonemes = get_char_phoneme_labels(
-            self.polyphonic_chars
+            polyphonic_chars=self.polyphonic_chars
        ) if self.config.use_char_phoneme else get_phoneme_labels(
-            self.polyphonic_chars)
+            polyphonic_chars=self.polyphonic_chars)

        self.chars = sorted(list(self.char2phonemes.keys()))

@@ -146,7 +152,7 @@ class G2PWOnnxConverter:
        if self.enable_opencc:
            self.cc = OpenCC('s2tw')

-    def _convert_bopomofo_to_pinyin(self, bopomofo):
+    def _convert_bopomofo_to_pinyin(self, bopomofo: str) -> str:
        tone = bopomofo[-1]
        assert tone in '12345'
        component = self.bopomofo_convert_dict.get(bopomofo[:-1])
@@ -156,7 +162,7 @@ class G2PWOnnxConverter:
            print(f'Warning: "{bopomofo}" cannot convert to pinyin')
            return None

-    def __call__(self, sentences):
+    def __call__(self, sentences: List[str]) -> List[List[str]]:
        if isinstance(sentences, str):
            sentences = [sentences]

@@ -169,23 +175,25 @@ class G2PWOnnxConverter:
            sentences = translated_sentences

        texts, query_ids, sent_ids, partial_results = self._prepare_data(
-            sentences)
+            sentences=sentences)
        if len(texts) == 0:
            # sentences no polyphonic words
            return partial_results

        onnx_input = prepare_onnx_input(
-            self.tokenizer,
-            self.labels,
-            self.char2phonemes,
-            self.chars,
-            texts,
-            query_ids,
+            tokenizer=self.tokenizer,
+            labels=self.labels,
+            char2phonemes=self.char2phonemes,
+            chars=self.chars,
+            texts=texts,
+            query_ids=query_ids,
            use_mask=self.config.use_mask,
-            use_char_phoneme=self.config.use_char_phoneme,
            window_size=None)

-        preds, confidences = predict(self.session_g2pW, onnx_input, self.labels)
+        preds, confidences = predict(
+            session=self.session_g2pW,
+            onnx_input=onnx_input,
+            labels=self.labels)
        if self.config.use_char_phoneme:
            preds = [pred.split(' ')[1] for pred in preds]

@@ -195,7 +203,9 @@ class G2PWOnnxConverter:

        return results

-    def _prepare_data(self, sentences):
+    def _prepare_data(
+            self, sentences: List[str]
+    ) -> Tuple[List[str], List[int], List[int], List[List[str]]]:
        texts, query_ids, sent_ids, partial_results = [], [], [], []
        for sent_id, sent in enumerate(sentences):
            # pypinyin works well for Simplified Chinese than Traditional Chinese

--- a/paddlespeech/t2s/frontend/g2pw/utils.py
+++ b/paddlespeech/t2s/frontend/g2pw/utils.py
@@ -15,10 +15,11 @@
 Credits
    This code is modified from https://github.com/GitYCC/g2pW
 """
+import os
 import re


-def wordize_and_map(text):
+def wordize_and_map(text: str):
    words = []
    index_map_from_text_to_word = []
    index_map_from_word_to_text = []
@@ -54,8 +55,8 @@ def wordize_and_map(text):
    return words, index_map_from_text_to_word, index_map_from_word_to_text


-def tokenize_and_map(tokenizer, text):
-    words, text2word, word2text = wordize_and_map(text)
+def tokenize_and_map(tokenizer, text: str):
+    words, text2word, word2text = wordize_and_map(text=text)

    tokens = []
    index_map_from_token_to_text = []
@@ -82,7 +83,7 @@ def tokenize_and_map(tokenizer, text):
    return tokens, index_map_from_text_to_token, index_map_from_token_to_text


-def _load_config(config_path):
+def _load_config(config_path: os.PathLike):
    import importlib.util
    spec = importlib.util.spec_from_file_location('__init__', config_path)
    config = importlib.util.module_from_spec(spec)
@@ -130,7 +131,7 @@ default_config_dict = {
 }


-def load_config(config_path, use_default=False):
+def load_config(config_path: os.PathLike, use_default: bool=False):
    config = _load_config(config_path)
    if use_default:
        for attr, val in default_config_dict.items():

--- a/paddlespeech/t2s/models/ernie_sat/ernie_sat.py
+++ b/paddlespeech/t2s/models/ernie_sat/ernie_sat.py
@@ -71,31 +71,53 @@ class MLMEncoder(nn.Layer):
    """Conformer encoder module.

    Args:
-        idim (int): Input dimension.
-        attention_dim (int): Dimension of attention.
-        attention_heads (int): The number of heads of multi head attention.
-        linear_units (int): The number of units of position-wise feed forward.
-        num_blocks (int): The number of decoder blocks.
-        dropout_rate (float): Dropout rate.
-        positional_dropout_rate (float): Dropout rate after adding positional encoding.
-        attention_dropout_rate (float): Dropout rate in attention.
-        input_layer (Union[str, paddle.nn.Layer]): Input layer type.
-        normalize_before (bool): Whether to use layer_norm before the first block.
-        concat_after (bool): Whether to concat attention layer's input and output.
+        idim (int): 
+            Input dimension.
+        attention_dim (int): 
+            Dimension of attention.
+        attention_heads (int): 
+            The number of heads of multi head attention.
+        linear_units (int): 
+            The number of units of position-wise feed forward.
+        num_blocks (int): 
+            The number of decoder blocks.
+        dropout_rate (float): 
+            Dropout rate.
+        positional_dropout_rate (float): 
+            Dropout rate after adding positional encoding.
+        attention_dropout_rate (float): 
+            Dropout rate in attention.
+        input_layer (Union[str, paddle.nn.Layer]): 
+            Input layer type.
+        normalize_before (bool): 
+            Whether to use layer_norm before the first block.
+        concat_after (bool): 
+            Whether to concat attention layer's input and output.
            if True, additional linear will be applied.
            i.e. x -> x + linear(concat(x, att(x)))
            if False, no additional linear will be applied. i.e. x -> x + att(x)
-        positionwise_layer_type (str): "linear", "conv1d", or "conv1d-linear".
-        positionwise_conv_kernel_size (int): Kernel size of positionwise conv1d layer.
-        macaron_style (bool): Whether to use macaron style for positionwise layer.
-        pos_enc_layer_type (str): Encoder positional encoding layer type.
-        selfattention_layer_type (str): Encoder attention layer type.
-        activation_type (str): Encoder activation function type.
-        use_cnn_module (bool): Whether to use convolution module.
-        zero_triu (bool): Whether to zero the upper triangular part of attention matrix.
-        cnn_module_kernel (int): Kernerl size of convolution module.
-        padding_idx (int): Padding idx for input_layer=embed.
-        stochastic_depth_rate (float): Maximum probability to skip the encoder layer.
+        positionwise_layer_type (str): 
+            "linear", "conv1d", or "conv1d-linear".
+        positionwise_conv_kernel_size (int): 
+            Kernel size of positionwise conv1d layer.
+        macaron_style (bool): 
+            Whether to use macaron style for positionwise layer.
+        pos_enc_layer_type (str): 
+            Encoder positional encoding layer type.
+        selfattention_layer_type (str): 
+            Encoder attention layer type.
+        activation_type (str): 
+            Encoder activation function type.
+        use_cnn_module (bool): 
+            Whether to use convolution module.
+        zero_triu (bool): 
+            Whether to zero the upper triangular part of attention matrix.
+        cnn_module_kernel (int): 
+            Kernerl size of convolution module.
+        padding_idx (int): 
+            Padding idx for input_layer=embed.
+        stochastic_depth_rate (float): 
+            Maximum probability to skip the encoder layer.

    """

@@ -320,12 +342,16 @@ class MLMDecoder(MLMEncoder):
        """Encode input sequence.

        Args:
-            xs (paddle.Tensor): Input tensor (#batch, time, idim).
-            masks (paddle.Tensor): Mask tensor (#batch, time).
+            xs (paddle.Tensor): 
+                Input tensor (#batch, time, idim).
+            masks (paddle.Tensor): 
+                Mask tensor (#batch, time).

        Returns:
-            paddle.Tensor: Output tensor (#batch, time, attention_dim).
-            paddle.Tensor: Mask tensor (#batch, time).
+            paddle.Tensor: 
+                Output tensor (#batch, time, attention_dim).
+            paddle.Tensor: 
+                Mask tensor (#batch, time).

        """
        xs = self.embed(xs)
@@ -392,19 +418,27 @@ class MLM(nn.Layer):
            use_teacher_forcing: bool=True, ) -> List[paddle.Tensor]:
        '''
        Args:
-            speech (paddle.Tensor): input speech (1, Tmax, D).
-            text (paddle.Tensor): input text (1, Tmax2).
-            masked_pos (paddle.Tensor): masked position of input speech (1, Tmax)
-            speech_mask (paddle.Tensor): mask of speech (1, 1, Tmax).
-            text_mask (paddle.Tensor): mask of text (1, 1, Tmax2).
-            speech_seg_pos (paddle.Tensor): n-th phone of each mel, 0<=n<=Tmax2 (1, Tmax).
-            text_seg_pos (paddle.Tensor): n-th phone of each phone, 0<=n<=Tmax2 (1, Tmax2).
-            span_bdy (List[int]): masked mel boundary of input speech (2,)
-            use_teacher_forcing (bool): whether to use teacher forcing
+            speech (paddle.Tensor): 
+                input speech (1, Tmax, D).
+            text (paddle.Tensor): 
+                input text (1, Tmax2).
+            masked_pos (paddle.Tensor): 
+                masked position of input speech (1, Tmax)
+            speech_mask (paddle.Tensor): 
+                mask of speech (1, 1, Tmax).
+            text_mask (paddle.Tensor): 
+                mask of text (1, 1, Tmax2).
+            speech_seg_pos (paddle.Tensor): 
+                n-th phone of each mel, 0<=n<=Tmax2 (1, Tmax).
+            text_seg_pos (paddle.Tensor): 
+                n-th phone of each phone, 0<=n<=Tmax2 (1, Tmax2).
+            span_bdy (List[int]): 
+                masked mel boundary of input speech (2,)
+            use_teacher_forcing (bool): 
+                whether to use teacher forcing
        Returns:
            List[Tensor]:
-                eg:
-                [Tensor(shape=[1, 181, 80]), Tensor(shape=[80, 80]), Tensor(shape=[1, 67, 80])]
+                eg: [Tensor(shape=[1, 181, 80]), Tensor(shape=[80, 80]), Tensor(shape=[1, 67, 80])]
        '''

        z_cache = None

--- a/paddlespeech/t2s/models/vits/duration_predictor.py
+++ b/paddlespeech/t2s/models/vits/duration_predictor.py
@@ -48,12 +48,18 @@ class StochasticDurationPredictor(nn.Layer):
            global_channels: int=-1, ):
        """Initialize StochasticDurationPredictor module.
        Args:
-            channels (int): Number of channels.
-            kernel_size (int): Kernel size.
-            dropout_rate (float): Dropout rate.
-            flows (int): Number of flows.
-            dds_conv_layers (int): Number of conv layers in DDS conv.
-            global_channels (int): Number of global conditioning channels.
+            channels (int):
+                Number of channels.
+            kernel_size (int):
+                Kernel size.
+            dropout_rate (float):
+                Dropout rate.
+            flows (int):
+                Number of flows.
+            dds_conv_layers (int):
+                Number of conv layers in DDS conv.
+            global_channels (int):
+                Number of global conditioning channels.
        """
        super().__init__()

@@ -108,14 +114,21 @@ class StochasticDurationPredictor(nn.Layer):
            noise_scale: float=1.0, ) -> paddle.Tensor:
        """Calculate forward propagation.
        Args:
-            x (Tensor): Input tensor (B, channels, T_text).
-            x_mask (Tensor): Mask tensor (B, 1, T_text).
-            w (Optional[Tensor]): Duration tensor (B, 1, T_text).
-            g (Optional[Tensor]): Global conditioning tensor (B, channels, 1)
-            inverse (bool): Whether to inverse the flow.
-            noise_scale (float): Noise scale value.
+            x (Tensor):
+                Input tensor (B, channels, T_text).
+            x_mask (Tensor):
+                Mask tensor (B, 1, T_text).
+            w (Optional[Tensor]):
+                Duration tensor (B, 1, T_text).
+            g (Optional[Tensor]):
+                Global conditioning tensor (B, channels, 1)
+            inverse (bool):
+                Whether to inverse the flow.
+            noise_scale (float):
+                Noise scale value.
        Returns:
-            Tensor: If not inverse, negative log-likelihood (NLL) tensor (B,).
+            Tensor: 
+                If not inverse, negative log-likelihood (NLL) tensor (B,).
                If inverse, log-duration tensor (B, 1, T_text).
        """
        # stop gradient

--- a/paddlespeech/t2s/models/vits/flow.py
+++ b/paddlespeech/t2s/models/vits/flow.py
@@ -34,11 +34,15 @@ class FlipFlow(nn.Layer):
                ) -> Union[paddle.Tensor, Tuple[paddle.Tensor, paddle.Tensor]]:
        """Calculate forward propagation.
        Args:
-            x (Tensor): Input tensor (B, channels, T).
-            inverse (bool): Whether to inverse the flow.
+            x (Tensor):
+                Input tensor (B, channels, T).
+            inverse (bool):
+                Whether to inverse the flow.
        Returns:
-            Tensor: Flipped tensor (B, channels, T).
-            Tensor: Log-determinant tensor for NLL (B,) if not inverse.
+            Tensor:
+                Flipped tensor (B, channels, T).
+            Tensor:
+                Log-determinant tensor for NLL (B,) if not inverse.
        """
        x = paddle.flip(x, [1])
        if not inverse:
@@ -60,13 +64,19 @@ class LogFlow(nn.Layer):
                ) -> Union[paddle.Tensor, Tuple[paddle.Tensor, paddle.Tensor]]:
        """Calculate forward propagation.
        Args:
-            x (Tensor): Input tensor (B, channels, T).
-            x_mask (Tensor): Mask tensor (B, 1, T).
-            inverse (bool): Whether to inverse the flow.
-            eps (float): Epsilon for log.
+            x (Tensor):
+                Input tensor (B, channels, T).
+            x_mask (Tensor):
+                Mask tensor (B, 1, T).
+            inverse (bool):
+                Whether to inverse the flow.
+            eps (float):
+                Epsilon for log.
        Returns:
-            Tensor: Output tensor (B, channels, T).
-            Tensor: Log-determinant tensor for NLL (B,) if not inverse.
+            Tensor:
+                Output tensor (B, channels, T).
+            Tensor:
+                Log-determinant tensor for NLL (B,) if not inverse.
        """
        if not inverse:
            y = paddle.log(paddle.clip(x, min=eps)) * x_mask
@@ -83,7 +93,8 @@ class ElementwiseAffineFlow(nn.Layer):
    def __init__(self, channels: int):
        """Initialize ElementwiseAffineFlow module.
        Args:
-            channels (int): Number of channels.
+            channels (int):
+                Number of channels.
        """
        super().__init__()
        self.channels = channels
@@ -107,12 +118,17 @@ class ElementwiseAffineFlow(nn.Layer):
                ) -> Union[paddle.Tensor, Tuple[paddle.Tensor, paddle.Tensor]]:
        """Calculate forward propagation.
        Args:
-            x (Tensor): Input tensor (B, channels, T).
-            x_mask (Tensor): Mask tensor (B, 1, T).
-            inverse (bool): Whether to inverse the flow.
+            x (Tensor):
+                Input tensor (B, channels, T).
+            x_mask (Tensor):
+                Mask tensor (B, 1, T).
+            inverse (bool):
+                Whether to inverse the flow.
        Returns:
-            Tensor: Output tensor (B, channels, T).
-            Tensor: Log-determinant tensor for NLL (B,) if not inverse.
+            Tensor:
+                Output tensor (B, channels, T).
+            Tensor:
+                Log-determinant tensor for NLL (B,) if not inverse.
        """
        if not inverse:
            y = self.m + paddle.exp(self.logs) * x
@@ -157,11 +173,16 @@ class DilatedDepthSeparableConv(nn.Layer):
            eps: float=1e-5, ):
        """Initialize DilatedDepthSeparableConv module.
        Args:
-            channels (int): Number of channels.
-            kernel_size (int): Kernel size.
-            layers (int): Number of layers.
-            dropout_rate (float): Dropout rate.
-            eps (float): Epsilon for layer norm.
+            channels (int):
+                Number of channels.
+            kernel_size (int):
+                Kernel size.
+            layers (int):
+                Number of layers.
+            dropout_rate (float):
+                Dropout rate.
+            eps (float):
+                Epsilon for layer norm.
        """
        super().__init__()

@@ -198,11 +219,15 @@ class DilatedDepthSeparableConv(nn.Layer):
                g: Optional[paddle.Tensor]=None) -> paddle.Tensor:
        """Calculate forward propagation.
        Args:
-            x (Tensor): Input tensor (B, in_channels, T).
-            x_mask (Tensor): Mask tensor (B, 1, T).
-            g (Optional[Tensor]): Global conditioning tensor (B, global_channels, 1).
+            x (Tensor):
+                Input tensor (B, in_channels, T).
+            x_mask (Tensor):
+                Mask tensor (B, 1, T).
+            g (Optional[Tensor]):
+                Global conditioning tensor (B, global_channels, 1).
        Returns:
-            Tensor: Output tensor (B, channels, T).
+            Tensor:
+                Output tensor (B, channels, T).
        """
        if g is not None:
            x = x + g
@@ -225,12 +250,18 @@ class ConvFlow(nn.Layer):
            tail_bound: float=5.0, ):
        """Initialize ConvFlow module.
        Args:
-            in_channels (int): Number of input channels.
-            hidden_channels (int): Number of hidden channels.
-            kernel_size (int): Kernel size.
-            layers (int): Number of layers.
-            bins (int): Number of bins.
-            tail_bound (float): Tail bound value.
+            in_channels (int):
+                Number of input channels.
+            hidden_channels (int):
+                Number of hidden channels.
+            kernel_size (int):
+                Kernel size.
+            layers (int):
+                Number of layers.
+            bins (int):
+                Number of bins.
+            tail_bound (float):
+                Tail bound value.
        """
        super().__init__()
        self.half_channels = in_channels // 2
@@ -275,13 +306,19 @@ class ConvFlow(nn.Layer):
    ) -> Union[paddle.Tensor, Tuple[paddle.Tensor, paddle.Tensor]]:
        """Calculate forward propagation.
        Args:
-            x (Tensor): Input tensor (B, channels, T).
-            x_mask (Tensor): Mask tensor (B, 1, T).
-            g (Optional[Tensor]): Global conditioning tensor (B, channels, 1).
-            inverse (bool): Whether to inverse the flow.
+            x (Tensor):
+                Input tensor (B, channels, T).
+            x_mask (Tensor):
+                Mask tensor (B, 1, T).
+            g (Optional[Tensor]):
+                Global conditioning tensor (B, channels, 1).
+            inverse (bool):
+                Whether to inverse the flow.
        Returns:
-            Tensor: Output tensor (B, channels, T).
-            Tensor: Log-determinant tensor for NLL (B,) if not inverse.
+            Tensor:
+                Output tensor (B, channels, T).
+            Tensor:
+                Log-determinant tensor for NLL (B,) if not inverse.
        """
        xa, xb = x.split(2, 1)
        h = self.input_conv(xa)

--- a/paddlespeech/t2s/models/vits/generator.py
+++ b/paddlespeech/t2s/models/vits/generator.py
@@ -97,81 +97,104 @@ class VITSGenerator(nn.Layer):
            stochastic_duration_predictor_dds_conv_layers: int=3, ):
        """Initialize VITS generator module.
        Args:
-            vocabs (int): Input vocabulary size.
-            aux_channels (int): Number of acoustic feature channels.
-            hidden_channels (int): Number of hidden channels.
-            spks (Optional[int]): Number of speakers. If set to > 1, assume that the
+            vocabs (int):
+                Input vocabulary size.
+            aux_channels (int):
+                Number of acoustic feature channels.
+            hidden_channels (int):
+                Number of hidden channels.
+            spks (Optional[int]):
+                Number of speakers. If set to > 1, assume that the
                sids will be provided as the input and use sid embedding layer.
-            langs (Optional[int]): Number of languages. If set to > 1, assume that the
+            langs (Optional[int]):
+                Number of languages. If set to > 1, assume that the
                lids will be provided as the input and use sid embedding layer.
-            spk_embed_dim (Optional[int]): Speaker embedding dimension. If set to > 0,
+            spk_embed_dim (Optional[int]):
+                Speaker embedding dimension. If set to > 0,
                assume that spembs will be provided as the input.
-            global_channels (int): Number of global conditioning channels.
-            segment_size (int): Segment size for decoder.
-            text_encoder_attention_heads (int): Number of heads in conformer block
-                of text encoder.
-            text_encoder_ffn_expand (int): Expansion ratio of FFN in conformer block
-                of text encoder.
-            text_encoder_blocks (int): Number of conformer blocks in text encoder.
-            text_encoder_positionwise_layer_type (str): Position-wise layer type in
-                conformer block of text encoder.
-            text_encoder_positionwise_conv_kernel_size (int): Position-wise convolution
-                kernel size in conformer block of text encoder. Only used when the
-                above layer type is conv1d or conv1d-linear.
-            text_encoder_positional_encoding_layer_type (str): Positional encoding layer
-                type in conformer block of text encoder.
-            text_encoder_self_attention_layer_type (str): Self-attention layer type in
-                conformer block of text encoder.
-            text_encoder_activation_type (str): Activation function type in conformer
-                block of text encoder.
-            text_encoder_normalize_before (bool): Whether to apply layer norm before
-                self-attention in conformer block of text encoder.
-            text_encoder_dropout_rate (float): Dropout rate in conformer block of
-                text encoder.
-            text_encoder_positional_dropout_rate (float): Dropout rate for positional
-                encoding in conformer block of text encoder.
-            text_encoder_attention_dropout_rate (float): Dropout rate for attention in
-                conformer block of text encoder.
-            text_encoder_conformer_kernel_size (int): Conformer conv kernel size. It
-                will be used when only use_conformer_conv_in_text_encoder = True.
-            use_macaron_style_in_text_encoder (bool): Whether to use macaron style FFN
-                in conformer block of text encoder.
-            use_conformer_conv_in_text_encoder (bool): Whether to use covolution in
-                conformer block of text encoder.
-            decoder_kernel_size (int): Decoder kernel size.
-            decoder_channels (int): Number of decoder initial channels.
-            decoder_upsample_scales (List[int]): List of upsampling scales in decoder.
-            decoder_upsample_kernel_sizes (List[int]): List of kernel size for
-                upsampling layers in decoder.
-            decoder_resblock_kernel_sizes (List[int]): List of kernel size for resblocks
-                in decoder.
-            decoder_resblock_dilations (List[List[int]]): List of list of dilations for
-                resblocks in decoder.
-            use_weight_norm_in_decoder (bool): Whether to apply weight normalization in
-                decoder.
-            posterior_encoder_kernel_size (int): Posterior encoder kernel size.
-            posterior_encoder_layers (int): Number of layers of posterior encoder.
-            posterior_encoder_stacks (int): Number of stacks of posterior encoder.
-            posterior_encoder_base_dilation (int): Base dilation of posterior encoder.
-            posterior_encoder_dropout_rate (float): Dropout rate for posterior encoder.
-            use_weight_norm_in_posterior_encoder (bool): Whether to apply weight
-                normalization in posterior encoder.
-            flow_flows (int): Number of flows in flow.
-            flow_kernel_size (int): Kernel size in flow.
-            flow_base_dilation (int): Base dilation in flow.
-            flow_layers (int): Number of layers in flow.
-            flow_dropout_rate (float): Dropout rate in flow
-            use_weight_norm_in_flow (bool): Whether to apply weight normalization in
-                flow.
-            use_only_mean_in_flow (bool): Whether to use only mean in flow.
-            stochastic_duration_predictor_kernel_size (int): Kernel size in stochastic
-                duration predictor.
-            stochastic_duration_predictor_dropout_rate (float): Dropout rate in
-                stochastic duration predictor.
-            stochastic_duration_predictor_flows (int): Number of flows in stochastic
-                duration predictor.
-            stochastic_duration_predictor_dds_conv_layers (int): Number of DDS conv
-                layers in stochastic duration predictor.
+            global_channels (int):
+                Number of global conditioning channels.
+            segment_size (int):
+                Segment size for decoder.
+            text_encoder_attention_heads (int):
+                Number of heads in conformer block of text encoder.
+            text_encoder_ffn_expand (int): 
+                Expansion ratio of FFN in conformer block of text encoder.
+            text_encoder_blocks (int):
+                Number of conformer blocks in text encoder.
+            text_encoder_positionwise_layer_type (str):
+                Position-wise layer type in conformer block of text encoder.
+            text_encoder_positionwise_conv_kernel_size (int):
+                Position-wise convolution kernel size in conformer block of text encoder. 
+                Only used when the above layer type is conv1d or conv1d-linear.
+            text_encoder_positional_encoding_layer_type (str):
+                Positional encoding layer type in conformer block of text encoder.
+            text_encoder_self_attention_layer_type (str):
+                Self-attention layer type in conformer block of text encoder.
+            text_encoder_activation_type (str):
+                Activation function type in conformer block of text encoder.
+            text_encoder_normalize_before (bool): 
+                Whether to apply layer norm before self-attention in conformer block of text encoder.
+            text_encoder_dropout_rate (float):
+                Dropout rate in conformer block of text encoder.
+            text_encoder_positional_dropout_rate (float):
+                Dropout rate for positional encoding in conformer block of text encoder.
+            text_encoder_attention_dropout_rate (float):
+                Dropout rate for attention in conformer block of text encoder.
+            text_encoder_conformer_kernel_size (int):
+                Conformer conv kernel size. It will be used when only use_conformer_conv_in_text_encoder = True.
+            use_macaron_style_in_text_encoder (bool):
+                Whether to use macaron style FFN in conformer block of text encoder.
+            use_conformer_conv_in_text_encoder (bool):
+                Whether to use covolution in conformer block of text encoder.
+            decoder_kernel_size (int):
+                Decoder kernel size.
+            decoder_channels (int):
+                Number of decoder initial channels.
+            decoder_upsample_scales (List[int]):
+                List of upsampling scales in decoder.
+            decoder_upsample_kernel_sizes (List[int]):
+                List of kernel size for upsampling layers in decoder.
+            decoder_resblock_kernel_sizes (List[int]):
+                List of kernel size for resblocks in decoder.
+            decoder_resblock_dilations (List[List[int]]):
+                List of list of dilations for resblocks in decoder.
+            use_weight_norm_in_decoder (bool):
+                Whether to apply weight normalization in decoder.
+            posterior_encoder_kernel_size (int):
+                Posterior encoder kernel size.
+            posterior_encoder_layers (int):
+                Number of layers of posterior encoder.
+            posterior_encoder_stacks (int):
+                Number of stacks of posterior encoder.
+            posterior_encoder_base_dilation (int):
+                Base dilation of posterior encoder.
+            posterior_encoder_dropout_rate (float):
+                Dropout rate for posterior encoder.
+            use_weight_norm_in_posterior_encoder (bool): 
+                Whether to apply weight normalization in posterior encoder.
+            flow_flows (int):
+                Number of flows in flow.
+            flow_kernel_size (int):
+                Kernel size in flow.
+            flow_base_dilation (int):
+                Base dilation in flow.
+            flow_layers (int):
+                Number of layers in flow.
+            flow_dropout_rate (float):
+                Dropout rate in flow
+            use_weight_norm_in_flow (bool):
+                Whether to apply weight normalization in flow.
+            use_only_mean_in_flow (bool):
+                Whether to use only mean in flow.
+            stochastic_duration_predictor_kernel_size (int): 
+                Kernel size in stochastic duration predictor.
+            stochastic_duration_predictor_dropout_rate (float):
+                Dropout rate in stochastic duration predictor.
+            stochastic_duration_predictor_flows (int):
+                Number of flows in stochastic duration predictor.
+            stochastic_duration_predictor_dds_conv_layers (int):
+                Number of DDS conv layers in stochastic duration predictor.
        """
        super().__init__()
        self.segment_size = segment_size
@@ -272,27 +295,40 @@ class VITSGenerator(nn.Layer):
                     paddle.Tensor, paddle.Tensor, ], ]:
        """Calculate forward propagation.
        Args:
-            text (Tensor): Text index tensor (B, T_text).
-            text_lengths (Tensor): Text length tensor (B,).
-            feats (Tensor): Feature tensor (B, aux_channels, T_feats).
-            feats_lengths (Tensor): Feature length tensor (B,).
-            sids (Optional[Tensor]): Speaker index tensor (B,) or (B, 1).
-            spembs (Optional[Tensor]): Speaker embedding tensor (B, spk_embed_dim).
-            lids (Optional[Tensor]): Language index tensor (B,) or (B, 1).
+            text (Tensor):
+                Text index tensor (B, T_text).
+            text_lengths (Tensor):
+                Text length tensor (B,).
+            feats (Tensor):
+                Feature tensor (B, aux_channels, T_feats).
+            feats_lengths (Tensor):
+                Feature length tensor (B,).
+            sids (Optional[Tensor]):
+                Speaker index tensor (B,) or (B, 1).
+            spembs (Optional[Tensor]):
+                Speaker embedding tensor (B, spk_embed_dim).
+            lids (Optional[Tensor]):
+                Language index tensor (B,) or (B, 1).
        Returns:
-            Tensor: Waveform tensor (B, 1, segment_size * upsample_factor).
-            Tensor: Duration negative log-likelihood (NLL) tensor (B,).
-            Tensor: Monotonic attention weight tensor (B, 1, T_feats, T_text).
-            Tensor: Segments start index tensor (B,).
-            Tensor: Text mask tensor (B, 1, T_text).
-            Tensor: Feature mask tensor (B, 1, T_feats).
-            tuple[Tensor, Tensor, Tensor, Tensor, Tensor, Tensor]:
-                - Tensor: Posterior encoder hidden representation (B, H, T_feats).
-                - Tensor: Flow hidden representation (B, H, T_feats).
-                - Tensor: Expanded text encoder projected mean (B, H, T_feats).
-                - Tensor: Expanded text encoder projected scale (B, H, T_feats).
-                - Tensor: Posterior encoder projected mean (B, H, T_feats).
-                - Tensor: Posterior encoder projected scale (B, H, T_feats).
+            Tensor:
+                Waveform tensor (B, 1, segment_size * upsample_factor).
+            Tensor:
+                Duration negative log-likelihood (NLL) tensor (B,).
+            Tensor:
+                Monotonic attention weight tensor (B, 1, T_feats, T_text).
+            Tensor:
+                Segments start index tensor (B,).
+            Tensor:
+                Text mask tensor (B, 1, T_text).
+            Tensor: 
+                Feature mask tensor (B, 1, T_feats).
+                tuple[Tensor, Tensor, Tensor, Tensor, Tensor, Tensor]:
+                    - Tensor: Posterior encoder hidden representation (B, H, T_feats).
+                    - Tensor: Flow hidden representation (B, H, T_feats).
+                    - Tensor: Expanded text encoder projected mean (B, H, T_feats).
+                    - Tensor: Expanded text encoder projected scale (B, H, T_feats).
+                    - Tensor: Posterior encoder projected mean (B, H, T_feats).
+                    - Tensor: Posterior encoder projected scale (B, H, T_feats).
        """
        # forward text encoder
        x, m_p, logs_p, x_mask = self.text_encoder(text, text_lengths)
@@ -402,24 +438,40 @@ class VITSGenerator(nn.Layer):
    ) -> Tuple[paddle.Tensor, paddle.Tensor, paddle.Tensor]:
        """Run inference.
        Args:
-            text (Tensor): Input text index tensor (B, T_text,).
-            text_lengths (Tensor): Text length tensor (B,).
-            feats (Tensor): Feature tensor (B, aux_channels, T_feats,).
-            feats_lengths (Tensor): Feature length tensor (B,).
-            sids (Optional[Tensor]): Speaker index tensor (B,) or (B, 1).
-            spembs (Optional[Tensor]): Speaker embedding tensor (B, spk_embed_dim).
-            lids (Optional[Tensor]): Language index tensor (B,) or (B, 1).
-            dur (Optional[Tensor]): Ground-truth duration (B, T_text,). If provided,
+            text (Tensor):
+                Input text index tensor (B, T_text,).
+            text_lengths (Tensor):
+                Text length tensor (B,).
+            feats (Tensor):
+                Feature tensor (B, aux_channels, T_feats,).
+            feats_lengths (Tensor):
+                Feature length tensor (B,).
+            sids (Optional[Tensor]):
+                Speaker index tensor (B,) or (B, 1).
+            spembs (Optional[Tensor]):
+                Speaker embedding tensor (B, spk_embed_dim).
+            lids (Optional[Tensor]):
+                Language index tensor (B,) or (B, 1).
+            dur (Optional[Tensor]):
+                Ground-truth duration (B, T_text,). If provided,
                skip the prediction of durations (i.e., teacher forcing).
-            noise_scale (float): Noise scale parameter for flow.
-            noise_scale_dur (float): Noise scale parameter for duration predictor.
-            alpha (float): Alpha parameter to control the speed of generated speech.
-            max_len (Optional[int]): Maximum length of acoustic feature sequence.
-            use_teacher_forcing (bool): Whether to use teacher forcing.
+            noise_scale (float):
+                Noise scale parameter for flow.
+            noise_scale_dur (float):
+                Noise scale parameter for duration predictor.
+            alpha (float):
+                Alpha parameter to control the speed of generated speech.
+            max_len (Optional[int]):
+                Maximum length of acoustic feature sequence.
+            use_teacher_forcing (bool):
+                Whether to use teacher forcing.
        Returns:
-            Tensor: Generated waveform tensor (B, T_wav).
-            Tensor: Monotonic attention weight tensor (B, T_feats, T_text).
-            Tensor: Duration tensor (B, T_text).
+            Tensor: 
+                Generated waveform tensor (B, T_wav).
+            Tensor:
+                Monotonic attention weight tensor (B, T_feats, T_text).
+            Tensor:
+                Duration tensor (B, T_text).
        """
        # encoder
        x, m_p, logs_p, x_mask = self.text_encoder(text, text_lengths)
@@ -533,15 +585,23 @@ class VITSGenerator(nn.Layer):
            lids: Optional[paddle.Tensor]=None, ) -> paddle.Tensor:
        """Run voice conversion.
        Args:
-            feats (Tensor): Feature tensor (B, aux_channels, T_feats,).
-            feats_lengths (Tensor): Feature length tensor (B,).
-            sids_src (Optional[Tensor]): Speaker index tensor of source feature (B,) or (B, 1).
-            sids_tgt (Optional[Tensor]): Speaker index tensor of target feature (B,) or (B, 1).
-            spembs_src (Optional[Tensor]): Speaker embedding tensor of source feature (B, spk_embed_dim).
-            spembs_tgt (Optional[Tensor]): Speaker embedding tensor of target feature (B, spk_embed_dim).
-            lids (Optional[Tensor]): Language index tensor (B,) or (B, 1).
+            feats (Tensor):
+                Feature tensor (B, aux_channels, T_feats,).
+            feats_lengths (Tensor):
+                Feature length tensor (B,).
+            sids_src (Optional[Tensor]):
+                Speaker index tensor of source feature (B,) or (B, 1).
+            sids_tgt (Optional[Tensor]):
+                Speaker index tensor of target feature (B,) or (B, 1).
+            spembs_src (Optional[Tensor]):
+                Speaker embedding tensor of source feature (B, spk_embed_dim).
+            spembs_tgt (Optional[Tensor]):
+                Speaker embedding tensor of target feature (B, spk_embed_dim).
+            lids (Optional[Tensor]):
+                Language index tensor (B,) or (B, 1).
        Returns:
-            Tensor: Generated waveform tensor (B, T_wav).
+            Tensor:
+                Generated waveform tensor (B, T_wav).
        """
        # encoder
        g_src = None
@@ -602,10 +662,13 @@ class VITSGenerator(nn.Layer):
                       mask: paddle.Tensor) -> paddle.Tensor:
        """Generate path a.k.a. monotonic attention.
        Args:
-            dur (Tensor): Duration tensor (B, 1, T_text).
-            mask (Tensor): Attention mask tensor (B, 1, T_feats, T_text).
+            dur (Tensor):
+                Duration tensor (B, 1, T_text).
+            mask (Tensor):
+                Attention mask tensor (B, 1, T_feats, T_text).
        Returns:
-            Tensor: Path tensor (B, 1, T_feats, T_text).
+            Tensor:
+                Path tensor (B, 1, T_feats, T_text).
        """
        b, _, t_y, t_x = paddle.shape(mask)
        cum_dur = paddle.cumsum(dur, -1)

--- a/paddlespeech/t2s/models/vits/posterior_encoder.py
+++ b/paddlespeech/t2s/models/vits/posterior_encoder.py
@@ -52,17 +52,28 @@ class PosteriorEncoder(nn.Layer):
        """Initilialize PosteriorEncoder module.

        Args:
-            in_channels (int): Number of input channels.
-            out_channels (int): Number of output channels.
-            hidden_channels (int): Number of hidden channels.
-            kernel_size (int): Kernel size in WaveNet.
-            layers (int): Number of layers of WaveNet.
-            stacks (int): Number of repeat stacking of WaveNet.
-            base_dilation (int): Base dilation factor.
-            global_channels (int): Number of global conditioning channels.
-            dropout_rate (float): Dropout rate.
-            bias (bool): Whether to use bias parameters in conv.
-            use_weight_norm (bool): Whether to apply weight norm.
+            in_channels (int):
+                Number of input channels.
+            out_channels (int):
+                Number of output channels.
+            hidden_channels (int):
+                Number of hidden channels.
+            kernel_size (int):
+                Kernel size in WaveNet.
+            layers (int):
+                Number of layers of WaveNet.
+            stacks (int):
+                Number of repeat stacking of WaveNet.
+            base_dilation (int):
+                Base dilation factor.
+            global_channels (int):
+                Number of global conditioning channels.
+            dropout_rate (float):
+                Dropout rate.
+            bias (bool):
+                Whether to use bias parameters in conv.
+            use_weight_norm (bool):
+                Whether to apply weight norm.

        """
        super().__init__()
@@ -99,15 +110,22 @@ class PosteriorEncoder(nn.Layer):
        """Calculate forward propagation.

        Args:
-            x (Tensor): Input tensor (B, in_channels, T_feats).
-            x_lengths (Tensor): Length tensor (B,).
-            g (Optional[Tensor]): Global conditioning tensor (B, global_channels, 1).
+            x (Tensor):
+                Input tensor (B, in_channels, T_feats).
+            x_lengths (Tensor):
+                Length tensor (B,).
+            g (Optional[Tensor]):
+                Global conditioning tensor (B, global_channels, 1).

        Returns:
-            Tensor: Encoded hidden representation tensor (B, out_channels, T_feats).
-            Tensor: Projected mean tensor (B, out_channels, T_feats).
-            Tensor: Projected scale tensor (B, out_channels, T_feats).
-            Tensor: Mask tensor for input tensor (B, 1, T_feats).
+            Tensor:
+                Encoded hidden representation tensor (B, out_channels, T_feats).
+            Tensor:
+                Projected mean tensor (B, out_channels, T_feats).
+            Tensor:
+                Projected scale tensor (B, out_channels, T_feats).
+            Tensor:
+                Mask tensor for input tensor (B, 1, T_feats).

        """
        x_mask = make_non_pad_mask(x_lengths).unsqueeze(1)

--- a/paddlespeech/t2s/models/vits/residual_coupling.py
+++ b/paddlespeech/t2s/models/vits/residual_coupling.py
@@ -55,18 +55,30 @@ class ResidualAffineCouplingBlock(nn.Layer):
        """Initilize ResidualAffineCouplingBlock module.

        Args:
-            in_channels (int): Number of input channels.
-            hidden_channels (int): Number of hidden channels.
-            flows (int): Number of flows.
-            kernel_size (int): Kernel size for WaveNet.
-            base_dilation (int): Base dilation factor for WaveNet.
-            layers (int): Number of layers of WaveNet.
-            stacks (int): Number of stacks of WaveNet.
-            global_channels (int): Number of global channels.
-            dropout_rate (float): Dropout rate.
-            use_weight_norm (bool): Whether to use weight normalization in WaveNet.
-            bias (bool): Whether to use bias paramters in WaveNet.
-            use_only_mean (bool): Whether to estimate only mean.
+            in_channels (int):
+                Number of input channels.
+            hidden_channels (int):
+                Number of hidden channels.
+            flows (int):
+                Number of flows.
+            kernel_size (int):
+                Kernel size for WaveNet.
+            base_dilation (int):
+                Base dilation factor for WaveNet.
+            layers (int):
+                Number of layers of WaveNet.
+            stacks (int):
+                Number of stacks of WaveNet.
+            global_channels (int):
+                Number of global channels.
+            dropout_rate (float):
+                Dropout rate.
+            use_weight_norm (bool):
+                Whether to use weight normalization in WaveNet.
+            bias (bool):
+                Whether to use bias paramters in WaveNet.
+            use_only_mean (bool):
+                Whether to estimate only mean.

        """
        super().__init__()
@@ -97,10 +109,14 @@ class ResidualAffineCouplingBlock(nn.Layer):
        """Calculate forward propagation.

        Args:
-            x (Tensor): Input tensor (B, in_channels, T).
-            x_mask (Tensor): Length tensor (B, 1, T).
-            g (Optional[Tensor]): Global conditioning tensor (B, global_channels, 1).
-            inverse (bool): Whether to inverse the flow.
+            x (Tensor):
+                Input tensor (B, in_channels, T).
+            x_mask (Tensor):
+                Length tensor (B, 1, T).
+            g (Optional[Tensor]):
+                Global conditioning tensor (B, global_channels, 1).
+            inverse (bool):
+                Whether to inverse the flow.

        Returns:
            Tensor: Output tensor (B, in_channels, T).
@@ -134,17 +150,28 @@ class ResidualAffineCouplingLayer(nn.Layer):
        """Initialzie ResidualAffineCouplingLayer module.

        Args:
-            in_channels (int): Number of input channels.
-            hidden_channels (int): Number of hidden channels.
-            kernel_size (int): Kernel size for WaveNet.
-            base_dilation (int): Base dilation factor for WaveNet.
-            layers (int): Number of layers of WaveNet.
-            stacks (int): Number of stacks of WaveNet.
-            global_channels (int): Number of global channels.
-            dropout_rate (float): Dropout rate.
-            use_weight_norm (bool): Whether to use weight normalization in WaveNet.
-            bias (bool): Whether to use bias paramters in WaveNet.
-            use_only_mean (bool): Whether to estimate only mean.
+            in_channels (int):
+                Number of input channels.
+            hidden_channels (int):
+                Number of hidden channels.
+            kernel_size (int):
+                Kernel size for WaveNet.
+            base_dilation (int):
+                Base dilation factor for WaveNet.
+            layers (int):
+                Number of layers of WaveNet.
+            stacks (int):
+                Number of stacks of WaveNet.
+            global_channels (int):
+                Number of global channels.
+            dropout_rate (float):
+                Dropout rate.
+            use_weight_norm (bool):
+                Whether to use weight normalization in WaveNet.
+            bias (bool):
+                Whether to use bias paramters in WaveNet.
+            use_only_mean (bool):
+                Whether to estimate only mean.

        """
        assert in_channels % 2 == 0, "in_channels should be divisible by 2"
@@ -211,14 +238,20 @@ class ResidualAffineCouplingLayer(nn.Layer):
        """Calculate forward propagation.

        Args:
-            x (Tensor): Input tensor (B, in_channels, T).
-            x_lengths (Tensor): Length tensor (B,).
-            g (Optional[Tensor]): Global conditioning tensor (B, global_channels, 1).
-            inverse (bool): Whether to inverse the flow.
+            x (Tensor):
+                Input tensor (B, in_channels, T).
+            x_lengths (Tensor):
+                Length tensor (B,).
+            g (Optional[Tensor]):
+                Global conditioning tensor (B, global_channels, 1).
+            inverse (bool):
+                Whether to inverse the flow.

        Returns:
-            Tensor: Output tensor (B, in_channels, T).
-            Tensor: Log-determinant tensor for NLL (B,) if not inverse.
+            Tensor:
+                Output tensor (B, in_channels, T).
+            Tensor:
+                Log-determinant tensor for NLL (B,) if not inverse.

        """
        xa, xb = paddle.split(x, 2, axis=1)

--- a/paddlespeech/t2s/models/vits/text_encoder.py
+++ b/paddlespeech/t2s/models/vits/text_encoder.py
@@ -62,23 +62,40 @@ class TextEncoder(nn.Layer):
        """Initialize TextEncoder module.

        Args:
-            vocabs (int): Vocabulary size.
-            attention_dim (int): Attention dimension.
-            attention_heads (int): Number of attention heads.
-            linear_units (int): Number of linear units of positionwise layers.
-            blocks (int): Number of encoder blocks.
-            positionwise_layer_type (str): Positionwise layer type.
-            positionwise_conv_kernel_size (int): Positionwise layer's kernel size.
-            positional_encoding_layer_type (str): Positional encoding layer type.
-            self_attention_layer_type (str): Self-attention layer type.
-            activation_type (str): Activation function type.
-            normalize_before (bool): Whether to apply LayerNorm before attention.
-            use_macaron_style (bool): Whether to use macaron style components.
-            use_conformer_conv (bool): Whether to use conformer conv layers.
-            conformer_kernel_size (int): Conformer's conv kernel size.
-            dropout_rate (float): Dropout rate.
-            positional_dropout_rate (float): Dropout rate for positional encoding.
-            attention_dropout_rate (float): Dropout rate for attention.
+            vocabs (int):
+                Vocabulary size.
+            attention_dim (int):
+                Attention dimension.
+            attention_heads (int):
+                Number of attention heads.
+            linear_units (int):
+                Number of linear units of positionwise layers.
+            blocks (int):
+                Number of encoder blocks.
+            positionwise_layer_type (str):
+                Positionwise layer type.
+            positionwise_conv_kernel_size (int):
+                Positionwise layer's kernel size.
+            positional_encoding_layer_type (str):
+                Positional encoding layer type.
+            self_attention_layer_type (str):
+                Self-attention layer type.
+            activation_type (str):
+                Activation function type.
+            normalize_before (bool):
+                Whether to apply LayerNorm before attention.
+            use_macaron_style (bool):
+                Whether to use macaron style components.
+            use_conformer_conv (bool):
+                Whether to use conformer conv layers.
+            conformer_kernel_size (int):
+                Conformer's conv kernel size.
+            dropout_rate (float):
+                Dropout rate.
+            positional_dropout_rate (float):
+                Dropout rate for positional encoding.
+            attention_dropout_rate (float):
+                Dropout rate for attention.

        """
        super().__init__()
@@ -121,14 +138,20 @@ class TextEncoder(nn.Layer):
        """Calculate forward propagation.

        Args:
-            x (Tensor): Input index tensor (B, T_text).
-            x_lengths (Tensor): Length tensor (B,).
+            x (Tensor):
+                Input index tensor (B, T_text).
+            x_lengths (Tensor):
+                Length tensor (B,).

        Returns:
-            Tensor: Encoded hidden representation (B, attention_dim, T_text).
-            Tensor: Projected mean tensor (B, attention_dim, T_text).
-            Tensor: Projected scale tensor (B, attention_dim, T_text).
-            Tensor: Mask tensor for input tensor (B, 1, T_text).
+            Tensor:
+                Encoded hidden representation (B, attention_dim, T_text).
+            Tensor:
+                Projected mean tensor (B, attention_dim, T_text).
+            Tensor:
+                Projected scale tensor (B, attention_dim, T_text).
+            Tensor:
+                Mask tensor for input tensor (B, 1, T_text).

        """
        x = self.emb(x) * math.sqrt(self.attention_dim)

--- a/paddlespeech/t2s/models/vits/vits.py
+++ b/paddlespeech/t2s/models/vits/vits.py
@@ -156,17 +156,25 @@ class VITS(nn.Layer):
            init_type: str="xavier_uniform", ):
        """Initialize VITS module.
        Args:
-            idim (int): Input vocabrary size.
-            odim (int): Acoustic feature dimension. The actual output channels will
+            idim (int):
+                Input vocabrary size.
+            odim (int):
+                Acoustic feature dimension. The actual output channels will
                be 1 since VITS is the end-to-end text-to-wave model but for the
                compatibility odim is used to indicate the acoustic feature dimension.
-            sampling_rate (int): Sampling rate, not used for the training but it will
+            sampling_rate (int):
+                Sampling rate, not used for the training but it will
                be referred in saving waveform during the inference.
-            generator_type (str): Generator type.
-            generator_params (Dict[str, Any]): Parameter dict for generator.
-            discriminator_type (str): Discriminator type.
-            discriminator_params (Dict[str, Any]): Parameter dict for discriminator.
-            cache_generator_outputs (bool): Whether to cache generator outputs.
+            generator_type (str):
+                Generator type.
+            generator_params (Dict[str, Any]):
+                Parameter dict for generator.
+            discriminator_type (str):
+                Discriminator type.
+            discriminator_params (Dict[str, Any]):
+                Parameter dict for discriminator.
+            cache_generator_outputs (bool):
+                Whether to cache generator outputs.
        """
        assert check_argument_types()
        super().__init__()
@@ -218,14 +226,22 @@ class VITS(nn.Layer):
            forward_generator: bool=True, ) -> Dict[str, Any]:
        """Perform generator forward.
        Args:
-            text (Tensor): Text index tensor (B, T_text).
-            text_lengths (Tensor): Text length tensor (B,).
-            feats (Tensor): Feature tensor (B, T_feats, aux_channels).
-            feats_lengths (Tensor): Feature length tensor (B,).
-            sids (Optional[Tensor]): Speaker index tensor (B,) or (B, 1).
-            spembs (Optional[Tensor]): Speaker embedding tensor (B, spk_embed_dim).
-            lids (Optional[Tensor]): Language index tensor (B,) or (B, 1).
-            forward_generator (bool): Whether to forward generator.
+            text (Tensor):
+                Text index tensor (B, T_text).
+            text_lengths (Tensor):
+                Text length tensor (B,).
+            feats (Tensor):
+                Feature tensor (B, T_feats, aux_channels).
+            feats_lengths (Tensor):
+                Feature length tensor (B,).
+            sids (Optional[Tensor]):
+                Speaker index tensor (B,) or (B, 1).
+            spembs (Optional[Tensor]):
+                Speaker embedding tensor (B, spk_embed_dim).
+            lids (Optional[Tensor]):
+                Language index tensor (B,) or (B, 1).
+            forward_generator (bool):
+                    Whether to forward generator.
        Returns:
        
        """
@@ -259,13 +275,20 @@ class VITS(nn.Layer):
            lids: Optional[paddle.Tensor]=None, ) -> Dict[str, Any]:
        """Perform generator forward.
        Args:
-            text (Tensor): Text index tensor (B, T_text).
-            text_lengths (Tensor): Text length tensor (B,).
-            feats (Tensor): Feature tensor (B, T_feats, aux_channels).
-            feats_lengths (Tensor): Feature length tensor (B,).
-            sids (Optional[Tensor]): Speaker index tensor (B,) or (B, 1).
-            spembs (Optional[Tensor]): Speaker embedding tensor (B, spk_embed_dim).
-            lids (Optional[Tensor]): Language index tensor (B,) or (B, 1).
+            text (Tensor):
+                Text index tensor (B, T_text).
+            text_lengths (Tensor):
+                Text length tensor (B,).
+            feats (Tensor):
+                Feature tensor (B, T_feats, aux_channels).
+            feats_lengths (Tensor):
+                Feature length tensor (B,).
+            sids (Optional[Tensor]):
+                Speaker index tensor (B,) or (B, 1).
+            spembs (Optional[Tensor]):
+                Speaker embedding tensor (B, spk_embed_dim).
+            lids (Optional[Tensor]):
+                Language index tensor (B,) or (B, 1).
        Returns:
            
        """
@@ -304,13 +327,20 @@ class VITS(nn.Layer):
            lids: Optional[paddle.Tensor]=None, ) -> Dict[str, Any]:
        """Perform discriminator forward.
        Args:
-            text (Tensor): Text index tensor (B, T_text).
-            text_lengths (Tensor): Text length tensor (B,).
-            feats (Tensor): Feature tensor (B, T_feats, aux_channels).
-            feats_lengths (Tensor): Feature length tensor (B,).
-            sids (Optional[Tensor]): Speaker index tensor (B,) or (B, 1).
-            spembs (Optional[Tensor]): Speaker embedding tensor (B, spk_embed_dim).
-            lids (Optional[Tensor]): Language index tensor (B,) or (B, 1).
+            text (Tensor):
+                Text index tensor (B, T_text).
+            text_lengths (Tensor):
+                Text length tensor (B,).
+            feats (Tensor):
+                Feature tensor (B, T_feats, aux_channels).
+            feats_lengths (Tensor):
+                Feature length tensor (B,).
+            sids (Optional[Tensor]):
+                Speaker index tensor (B,) or (B, 1).
+            spembs (Optional[Tensor]):
+                Speaker embedding tensor (B, spk_embed_dim).
+            lids (Optional[Tensor]):
+                Language index tensor (B,) or (B, 1).
        Returns:

        """
@@ -353,22 +383,36 @@ class VITS(nn.Layer):
            use_teacher_forcing: bool=False, ) -> Dict[str, paddle.Tensor]:
        """Run inference.
        Args:
-            text (Tensor): Input text index tensor (T_text,).
-            feats (Tensor): Feature tensor (T_feats, aux_channels).
-            sids (Tensor): Speaker index tensor (1,).
-            spembs (Optional[Tensor]): Speaker embedding tensor (spk_embed_dim,).
-            lids (Tensor): Language index tensor (1,).
-            durations (Tensor): Ground-truth duration tensor (T_text,).
-            noise_scale (float): Noise scale value for flow.
-            noise_scale_dur (float): Noise scale value for duration predictor.
-            alpha (float): Alpha parameter to control the speed of generated speech.
-            max_len (Optional[int]): Maximum length.
-            use_teacher_forcing (bool): Whether to use teacher forcing.
+            text (Tensor):
+                Input text index tensor (T_text,).
+            feats (Tensor):
+                Feature tensor (T_feats, aux_channels).
+            sids (Tensor):
+                Speaker index tensor (1,).
+            spembs (Optional[Tensor]):
+                Speaker embedding tensor (spk_embed_dim,).
+            lids (Tensor):
+                Language index tensor (1,).
+            durations (Tensor):
+                Ground-truth duration tensor (T_text,).
+            noise_scale (float):
+                Noise scale value for flow.
+            noise_scale_dur (float):
+                Noise scale value for duration predictor.
+            alpha (float):
+                Alpha parameter to control the speed of generated speech.
+            max_len (Optional[int]):
+                Maximum length.
+            use_teacher_forcing (bool):
+                Whether to use teacher forcing.
        Returns:
            Dict[str, Tensor]:
-                * wav (Tensor): Generated waveform tensor (T_wav,).
-                * att_w (Tensor): Monotonic attention weight tensor (T_feats, T_text).
-                * duration (Tensor): Predicted duration tensor (T_text,).
+                * wav (Tensor):
+                    Generated waveform tensor (T_wav,).
+                * att_w (Tensor):
+                    Monotonic attention weight tensor (T_feats, T_text).
+                * duration (Tensor):
+                    Predicted duration tensor (T_text,).
        """
        # setup
        text = text[None]
@@ -417,15 +461,22 @@ class VITS(nn.Layer):
            lids: Optional[paddle.Tensor]=None, ) -> paddle.Tensor:
        """Run voice conversion.
        Args:
-            feats (Tensor): Feature tensor (T_feats, aux_channels).
-            sids_src (Optional[Tensor]): Speaker index tensor of source feature (1,).
-            sids_tgt (Optional[Tensor]): Speaker index tensor of target feature (1,).
-            spembs_src (Optional[Tensor]): Speaker embedding tensor of source feature (spk_embed_dim,).
-            spembs_tgt (Optional[Tensor]): Speaker embedding tensor of target feature (spk_embed_dim,).
-            lids (Optional[Tensor]): Language index tensor (1,).
+            feats (Tensor):
+                Feature tensor (T_feats, aux_channels).
+            sids_src (Optional[Tensor]):
+                Speaker index tensor of source feature (1,).
+            sids_tgt (Optional[Tensor]):
+                Speaker index tensor of target feature (1,).
+            spembs_src (Optional[Tensor]):
+                Speaker embedding tensor of source feature (spk_embed_dim,).
+            spembs_tgt (Optional[Tensor]):
+                Speaker embedding tensor of target feature (spk_embed_dim,).
+            lids (Optional[Tensor]):
+                Language index tensor (1,).
        Returns:
            Dict[str, Tensor]:
-                * wav (Tensor): Generated waveform tensor (T_wav,).
+                * wav (Tensor):
+                    Generated waveform tensor (T_wav,).
        """
        assert feats is not None
        feats = feats[None].transpose([0, 2, 1])

--- a/paddlespeech/t2s/models/vits/wavenet/residual_block.py
+++ b/paddlespeech/t2s/models/vits/wavenet/residual_block.py
@@ -39,14 +39,22 @@ class ResidualBlock(nn.Layer):
        """Initialize ResidualBlock module.

        Args:
-            kernel_size (int): Kernel size of dilation convolution layer.
-            residual_channels (int): Number of channels for residual connection.
-            skip_channels (int): Number of channels for skip connection.
-            aux_channels (int): Number of local conditioning channels.
-            dropout (float): Dropout probability.
-            dilation (int): Dilation factor.
-            bias (bool): Whether to add bias parameter in convolution layers.
-            scale_residual (bool): Whether to scale the residual outputs.
+            kernel_size (int):
+                Kernel size of dilation convolution layer.
+            residual_channels (int):
+                Number of channels for residual connection.
+            skip_channels (int):
+                Number of channels for skip connection.
+            aux_channels (int):
+                Number of local conditioning channels.
+            dropout (float):
+                Dropout probability.
+            dilation (int):
+                Dilation factor.
+            bias (bool):
+                Whether to add bias parameter in convolution layers.
+            scale_residual (bool):
+                Whether to scale the residual outputs.

        """
        super().__init__()

--- a/paddlespeech/t2s/models/vits/wavenet/wavenet.py
+++ b/paddlespeech/t2s/models/vits/wavenet/wavenet.py
@@ -47,25 +47,42 @@ class WaveNet(nn.Layer):
        """Initialize WaveNet module.

        Args:
-            in_channels (int): Number of input channels.
-            out_channels (int): Number of output channels.
-            kernel_size (int): Kernel size of dilated convolution.
-            layers (int): Number of residual block layers.
-            stacks (int): Number of stacks i.e., dilation cycles.
-            base_dilation (int): Base dilation factor.
-            residual_channels (int): Number of channels in residual conv.
-            gate_channels (int):  Number of channels in gated conv.
-            skip_channels (int): Number of channels in skip conv.
-            aux_channels (int): Number of channels for local conditioning feature.
-            global_channels (int): Number of channels for global conditioning feature.
-            dropout_rate (float): Dropout rate. 0.0 means no dropout applied.
-            bias (bool): Whether to use bias parameter in conv layer.
-            use_weight_norm (bool): Whether to use weight norm. If set to true, it will
-                be applied to all of the conv layers.
-            use_first_conv (bool): Whether to use the first conv layers.
-            use_last_conv (bool): Whether to use the last conv layers.
-            scale_residual (bool): Whether to scale the residual outputs.
-            scale_skip_connect (bool): Whether to scale the skip connection outputs.
+            in_channels (int):
+                Number of input channels.
+            out_channels (int):
+                Number of output channels.
+            kernel_size (int):
+                Kernel size of dilated convolution.
+            layers (int):
+                Number of residual block layers.
+            stacks (int):
+                Number of stacks i.e., dilation cycles.
+            base_dilation (int):
+                Base dilation factor.
+            residual_channels (int):
+                Number of channels in residual conv.
+            gate_channels (int):
+                Number of channels in gated conv.
+            skip_channels (int):
+                Number of channels in skip conv.
+            aux_channels (int):
+                Number of channels for local conditioning feature.
+            global_channels (int):
+                Number of channels for global conditioning feature.
+            dropout_rate (float):
+                Dropout rate. 0.0 means no dropout applied.
+            bias (bool):
+                Whether to use bias parameter in conv layer.
+            use_weight_norm (bool):
+                Whether to use weight norm. If set to true, it will be applied to all of the conv layers.
+            use_first_conv (bool):
+                Whether to use the first conv layers.
+            use_last_conv (bool):
+                Whether to use the last conv layers.
+            scale_residual (bool):
+                Whether to scale the residual outputs.
+            scale_skip_connect (bool):
+                Whether to scale the skip connection outputs.

        """
        super().__init__()
@@ -128,15 +145,18 @@ class WaveNet(nn.Layer):
        """Calculate forward propagation.

        Args:
-            x (Tensor): Input noise signal (B, 1, T) if use_first_conv else
-                (B, residual_channels, T).
-            x_mask (Optional[Tensor]): Mask tensor (B, 1, T).
-            c (Optional[Tensor]): Local conditioning features (B, aux_channels, T).
-            g (Optional[Tensor]): Global conditioning features (B, global_channels, 1).
+            x (Tensor):
+                Input noise signal (B, 1, T) if use_first_conv else (B, residual_channels, T).
+            x_mask (Optional[Tensor]):
+                Mask tensor (B, 1, T).
+            c (Optional[Tensor]):
+                Local conditioning features (B, aux_channels, T).
+            g (Optional[Tensor]):
+                Global conditioning features (B, global_channels, 1).

        Returns:
-            Tensor: Output tensor (B, out_channels, T) if use_last_conv else
-                (B, residual_channels, T).
+            Tensor:
+                Output tensor (B, out_channels, T) if use_last_conv else(B, residual_channels, T).

        """
        # encode to hidden representation

--- a/paddlespeech/t2s/models/wavernn/wavernn.py
+++ b/paddlespeech/t2s/models/wavernn/wavernn.py
@@ -69,9 +69,11 @@ class MelResNet(nn.Layer):
    def forward(self, x):
        '''
        Args:
-            x (Tensor): Input tensor (B, in_dims, T).
+            x (Tensor):
+                Input tensor (B, in_dims, T).
        Returns:
-            Tensor: Output tensor (B, res_out_dims, T).
+            Tensor:
+                Output tensor (B, res_out_dims, T).
        '''

        x = self.conv_in(x)
@@ -119,10 +121,13 @@ class UpsampleNetwork(nn.Layer):
    def forward(self, m):
        '''
        Args:
-            c (Tensor): Input tensor (B, C_aux, T).
+            c (Tensor):
+                Input tensor (B, C_aux, T).
        Returns:
-            Tensor: Output tensor (B, (T - 2 * pad) *  prob(upsample_scales), C_aux).
-            Tensor: Output tensor (B, (T - 2 * pad) *  prob(upsample_scales), res_out_dims).
+            Tensor:
+                Output tensor (B, (T - 2 * pad) *  prob(upsample_scales), C_aux).
+            Tensor:
+                Output tensor (B, (T - 2 * pad) *  prob(upsample_scales), res_out_dims).
        '''
        # aux: [B, C_aux, T] 
        # -> [B, res_out_dims, T - 2 * aux_context_window]
@@ -302,7 +307,8 @@ class WaveRNN(nn.Layer):
                number of samples for crossfading between batches
            mu_law(bool)
        Returns: 
-            wav sequence: Output (T' * prod(upsample_scales), out_channels, C_out).
+            wav sequence:
+                Output (T' * prod(upsample_scales), out_channels, C_out).
        """

        self.eval()
@@ -423,7 +429,7 @@ class WaveRNN(nn.Layer):
            x(Tensor): 
                mel, [1, n_frames, 80]
            pad(int): 
-            side(str, optional):  (Default value = 'both')
+                side(str, optional):  (Default value = 'both')

        Returns:
            Tensor