fastspeech models #4337

supotato6 · 2025-07-08T07:23:32Z

No description provided.

paddle-bot · 2025-07-08T07:23:39Z

Thanks for your contribution!

zxcd · 2025-07-08T09:06:50Z

paddlex/configs/modules/text_to_speech_acoustic/fastspeech2_csmsc.yaml

+Predict:
+  batch_size: 1
+  model_dir: "fastspeech2csmsc"
+  input: "今天天气真不错"


input 应该是phone?

zxcd · 2025-07-08T09:07:10Z

paddlex/configs/modules/text_to_speech_vocoder/pwgan_csmsc.yaml

+Predict:
+  batch_size: 1
+  model_dir: "pwgan_csmsc"
+  input: "今天天气真不错"


input应该是npy或者tensor？

zxcd · 2025-07-16T03:22:46Z

paddlex/modules/text_to_speech_acoustic/__init__.py

+# limitations under the License.
+
+from .dataset_checker import TextToSpeechAcousticDatasetChecker
+# from .trainer import TextToSpeechTrainer


这个地方也记得改一下

zxcd · 2025-07-16T03:25:56Z

docs/module_usage/tutorials/speech_modules/text_to_speech_acoustic_en.md

@@ -0,0 +1,208 @@
+---


文件名不太对，应该是.en.md

zxcd · 2025-07-16T03:26:10Z

docs/module_usage/tutorials/speech_modules/text_to_speech_vocoder_en.md

@@ -0,0 +1,209 @@
+---


zxcd · 2025-07-16T03:27:09Z

docs/module_usage/tutorials/speech_modules/text_to_speech_vocoder_en.md

+
+## III. Quick Integration
+Before quick integration, first install the PaddleX wheel package. For wheel installation methods, please refer to [PaddleX Local Installation Tutorial](../../../installation/installation.md). After installing the wheel package, inference for the multilingual speech synthesis acoustic module can be completed with just a few lines of code. You can freely switch models within this module, or integrate model inference from the multilingual speech synthesis module into your project.
+<!-- Before running the following code, please download the [sample audio](https://paddlespeech.bj.bcebos.com/PaddleAudio/zh.wav){target="_blank"} to your local machine. -->


这个时候你的输入是个npy，sample应该不再使用audio？

zxcd · 2025-07-21T03:37:41Z

paddlex/inference/models/text_to_speech_acoustic/processors.py

@@ -0,0 +1,14 @@
+# copyright (c) 2025 PaddlePaddle Authors. All Rights Reserve.


useless file?

zxcd · 2025-07-21T03:38:03Z

paddlex/inference/models/text_to_speech_vocoder/processors.py

@@ -0,0 +1,14 @@
+# copyright (c) 2025 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");


useless file?

zxcd · 2025-07-21T03:39:25Z

paddlex/modules/text_to_speech_vocoder/__init__.py

@@ -0,0 +1,18 @@
+# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.


2024 -> 2025

…o develop_fastspeech_item

zxcd · 2025-08-20T04:02:19Z

paddlex/modules/__init__.py

    TextRecExportor,
    TextRecTrainer,
 )
+from .text_to_speech_vocoder import TextToSpeechVocoderDatasetChecker


whether should import others like trainer, exporter etc

zxcd · 2025-08-20T04:03:24Z

paddlex/modules/text_to_speech_acoustic/exportor.py

+    entities = MODELS
+
+    def __init__(self, config):
+            """


data format error

zxcd · 2025-08-20T04:07:21Z

paddlex/modules/text_to_speech_acoustic/evaluator.py

+
+
+class TextToSpeechAcousticEvaluator(BaseEvaluator):
+    """Instance Fastspeech2Model Model Evaluator"""


Fastspeech2

TingquanGao

一些问题，有空再看看吧。

TingquanGao · 2025-09-09T06:25:19Z

paddlex/configs/modules/text_to_speech_acoustic/fastspeech2_csmsc.yaml

+  use_trt: False
+  use_mkldnn: False
+  cpu_threads: 1
+  precision: "fp32"
+  output: "output"
+  model_name: "fastspeech2_csmsc"
+  speaker_dict: None
+  lang: zh
+  speaker_id: 0


除了output，其他参数只是推理相关的吧，是不是应该放到Predict中
另外为什么还有model_name？

TingquanGao · 2025-09-09T06:26:09Z

paddlex/configs/modules/text_to_speech_vocoder/pwgan_csmsc.yaml

+  use_trt: False
+  use_mkldnn: False
+  cpu_threads: 1
+  precision: "fp32"
+  output: "output"
+  model_name: "pwgan_csmsc"
+  speaker_dict: None
+  lang: zh
+  speaker_id: 0


TingquanGao · 2025-09-09T06:37:40Z

paddlex/inference/models/text_to_pinyin/processors.py

+                # # tone sandhi
+                # sub_finals = self.tone_modifier.modified_tone(word, pos,
+                #                                                 sub_finals)
+                # er hua                                
+                # if with_erhua:
+                #     sub_initials, sub_finals = self._merge_erhua(
+                #         sub_initials, sub_finals, word, pos)


如果是没用的代码，就删掉吧。

TingquanGao · 2025-09-09T06:37:41Z

paddlex/inference/models/text_to_pinyin/processors.py

+                # 多音字消歧
+                # word_pinyins = self.corrector.correct_pronunciation(
+                #     word, word_pinyins)


如果是没用的代码，就删掉吧。

TingquanGao · 2025-09-09T06:37:49Z

paddlex/inference/models/text_to_pinyin/processors.py

+            # fix wordseg bad case for sandhi
+            # seg_cut = self.tone_modifier.pre_merge_for_modify(seg_cut)
+            # 为了多音词获得更好的效果，这里采用整句预测


如果是没用的代码，就删掉吧。

TingquanGao · 2025-09-09T06:38:04Z

paddlex/inference/models/text_to_pinyin/processors.py

+                    )
+                elif char in self.char_bopomofo_dict:
+                    partial_result[i] = pypinyin_result[i][0]
+                    # partial_result[i] =  self.style_convert_func(self.char_bopomofo_dict[char][0])


如果是没用的代码，就删掉吧。

TingquanGao · 2025-09-09T06:40:23Z

paddlex/inference/utils/io/writers.py

+        self.sample_rate = sample_rate
+    def write(self, out_path, obj):


code style好像不对，过一下pre-commit

TingquanGao · 2025-09-09T06:41:55Z

paddlex/inference/utils/io/writers.py

+        self.sample_rate = sample_rate
+    def _write_obj(self, out_path, obj):


fastspeech models

024ab46

paddle-bot bot added the contributor External developers label Jul 8, 2025

add model config

87daa60

zxcd reviewed Jul 8, 2025

View reviewed changes

supotato6 added 2 commits July 16, 2025 02:32

config and readme

869557e

change filename

b0888d8

zxcd reviewed Jul 16, 2025

View reviewed changes

supotato6 added 7 commits July 21, 2025 02:41

model file link

2425bc7

add module file

fd10598

fix readme link

d1ccc00

filename

4b5ed89

filename1

b34303c

module methods

10db1c2

module methods init

98972c3

zxcd reviewed Jul 21, 2025

View reviewed changes

supotato6 and others added 5 commits August 6, 2025 09:35

fix file

7c1cfa8

Merge branch 'develop' of https://github.com/PaddlePaddle/PaddleX int…

1de42bf

…o develop_fastspeech_item

fix readme

bd74e33

rm useless file

709c9f8

fix information

4bdd7bd

zxcd reviewed Aug 20, 2025

View reviewed changes

supotato6 and others added 3 commits August 20, 2025 15:37

fix comments

d1c0367

Merge branch 'develop' into develop_fastspeech_item

176e2e6

fix bug

01dcb8c

TingquanGao reviewed Sep 9, 2025

View reviewed changes

supotato6 added 2 commits September 10, 2025 11:44

fix comments

d4b4847

delete code

9ff4230

		@@ -0,0 +1,14 @@
		# copyright (c) 2025 PaddlePaddle Authors. All Rights Reserve.

		@@ -0,0 +1,18 @@
		# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.



		class TextToSpeechAcousticEvaluator(BaseEvaluator):
		"""Instance Fastspeech2Model Model Evaluator"""

		self.sample_rate = sample_rate
		def write(self, out_path, obj):

		self.sample_rate = sample_rate
		def _write_obj(self, out_path, obj):

fastspeech models #4337

Are you sure you want to change the base?

fastspeech models #4337

Uh oh!

Conversation

supotato6 commented Jul 8, 2025

Uh oh!

paddle-bot bot commented Jul 8, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

TingquanGao left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!