[TTS]Add TTS Paddle-Lite x86 inference (#2667)

* Add export2lite, test=tts

* add tts paddlelite x86 inference, test=tts

* update released_model.md, test=tts

* add paddlelite in setup.py

* update
pull/2670/head
TianYuan 2 years ago committed by GitHub
parent 1c3d2cb89e
commit 0b4cf2211d
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23

@ -40,37 +40,35 @@ Language Model | Training Data | Token-based | Size | Descriptions
## Text-to-Speech Models ## Text-to-Speech Models
### Acoustic Models ### Acoustic Models
Model Type | Dataset| Example Link | Pretrained Models|Static/ONNX Models|Size (static) Model Type | Dataset| Example Link | Pretrained Models|Static / ONNX / Paddle-Lite Models|Size (static)
:-------------:| :------------:| :-----: | :-----:| :-----:| :-----: :-------------:| :------------:| :-----: | :-----:| :-----:| :-----:
Tacotron2|LJSpeech|[tacotron2-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/tts0)|[tacotron2_ljspeech_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/tacotron2/tacotron2_ljspeech_ckpt_0.2.0.zip)||| Tacotron2|LJSpeech|[tacotron2-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/tts0)|[tacotron2_ljspeech_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/tacotron2/tacotron2_ljspeech_ckpt_0.2.0.zip)|||
Tacotron2|CSMSC|[tacotron2-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/tts0)|[tacotron2_csmsc_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/tacotron2/tacotron2_csmsc_ckpt_0.2.0.zip)|[tacotron2_csmsc_static_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/tacotron2/tacotron2_csmsc_static_0.2.0.zip)|103MB| Tacotron2|CSMSC|[tacotron2-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/tts0)|[tacotron2_csmsc_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/tacotron2/tacotron2_csmsc_ckpt_0.2.0.zip)|[tacotron2_csmsc_static_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/tacotron2/tacotron2_csmsc_static_0.2.0.zip)|103MB|
TransformerTTS| LJSpeech| [transformer-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/tts1)|[transformer_tts_ljspeech_ckpt_0.4.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/transformer_tts/transformer_tts_ljspeech_ckpt_0.4.zip)||| TransformerTTS| LJSpeech| [transformer-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/tts1)|[transformer_tts_ljspeech_ckpt_0.4.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/transformer_tts/transformer_tts_ljspeech_ckpt_0.4.zip)|||
SpeedySpeech| CSMSC | [speedyspeech-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/tts2)|[speedyspeech_csmsc_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_ckpt_0.2.0.zip)|[speedyspeech_csmsc_static_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_static_0.2.0.zip) </br> [speedyspeech_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_onnx_0.2.0.zip)|13MB| SpeedySpeech| CSMSC | [speedyspeech-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/tts2)|[speedyspeech_csmsc_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_ckpt_0.2.0.zip)|[speedyspeech_csmsc_static_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_static_0.2.0.zip) </br> [speedyspeech_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_onnx_0.2.0.zip) </br> [speedyspeech_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_pdlite_1.3.0.zip)|13MB|
FastSpeech2| CSMSC |[fastspeech2-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/tts3)|[fastspeech2_nosil_baker_ckpt_0.4.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_baker_ckpt_0.4.zip)|[fastspeech2_csmsc_static_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_csmsc_static_0.2.0.zip) </br> [fastspeech2_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_csmsc_onnx_0.2.0.zip)|157MB| FastSpeech2| CSMSC |[fastspeech2-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/tts3)|[fastspeech2_nosil_baker_ckpt_0.4.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_baker_ckpt_0.4.zip)|[fastspeech2_csmsc_static_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_csmsc_static_0.2.0.zip) </br> [fastspeech2_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_csmsc_onnx_0.2.0.zip) </br> [fastspeech2_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_csmsc_pdlite_1.3.0.zip)|157MB|
FastSpeech2-Conformer| CSMSC |[fastspeech2-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/tts3)|[fastspeech2_conformer_baker_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_conformer_baker_ckpt_0.5.zip)||| FastSpeech2-Conformer| CSMSC |[fastspeech2-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/tts3)|[fastspeech2_conformer_baker_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_conformer_baker_ckpt_0.5.zip)|||
FastSpeech2-CNNDecoder| CSMSC| [fastspeech2-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/tts3)| [fastspeech2_cnndecoder_csmsc_ckpt_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_ckpt_1.0.0.zip) | [fastspeech2_cnndecoder_csmsc_static_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_static_1.0.0.zip) </br>[fastspeech2_cnndecoder_csmsc_streaming_static_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_streaming_static_1.0.0.zip) </br>[fastspeech2_cnndecoder_csmsc_onnx_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_onnx_1.0.0.zip) </br>[fastspeech2_cnndecoder_csmsc_streaming_onnx_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_streaming_onnx_1.0.0.zip) | 84MB| FastSpeech2-CNNDecoder| CSMSC| [fastspeech2-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/tts3)| [fastspeech2_cnndecoder_csmsc_ckpt_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_ckpt_1.0.0.zip) | [fastspeech2_cnndecoder_csmsc_static_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_static_1.0.0.zip) </br>[fastspeech2_cnndecoder_csmsc_streaming_static_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_streaming_static_1.0.0.zip) </br>[fastspeech2_cnndecoder_csmsc_onnx_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_onnx_1.0.0.zip) </br>[fastspeech2_cnndecoder_csmsc_streaming_onnx_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_streaming_onnx_1.0.0.zip) </br> [fastspeech2_cnndecoder_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_pdlite_1.3.0.zip) </br> [fastspeech2_cnndecoder_csmsc_streaming_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_streaming_pdlite_1.3.0.zip)| 84MB|
FastSpeech2| AISHELL-3 |[fastspeech2-aishell3](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/aishell3/tts3)|[fastspeech2_aishell3_ckpt_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_ckpt_1.1.0.zip)|[fastspeech2_aishell3_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_static_1.1.0.zip) </br> [fastspeech2_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_onnx_1.1.0.zip)|147MB| FastSpeech2| AISHELL-3 |[fastspeech2-aishell3](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/aishell3/tts3)|[fastspeech2_aishell3_ckpt_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_ckpt_1.1.0.zip)|[fastspeech2_aishell3_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_static_1.1.0.zip) </br> [fastspeech2_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_onnx_1.1.0.zip) </br> [fastspeech2_aishell3_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_pdlite_1.3.0.zip) |147MB|
FastSpeech2| LJSpeech |[fastspeech2-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/tts3)|[fastspeech2_nosil_ljspeech_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_ljspeech_ckpt_0.5.zip)|[fastspeech2_ljspeech_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_ljspeech_static_1.1.0.zip) </br> [fastspeech2_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_ljspeech_onnx_1.1.0.zip)|145MB| FastSpeech2| LJSpeech |[fastspeech2-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/tts3)|[fastspeech2_nosil_ljspeech_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_nosil_ljspeech_ckpt_0.5.zip)|[fastspeech2_ljspeech_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_ljspeech_static_1.1.0.zip) </br> [fastspeech2_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_ljspeech_onnx_1.1.0.zip) </br> [fastspeech2_ljspeech_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_ljspeech_pdlite_1.3.0.zip)|145MB|
FastSpeech2| VCTK |[fastspeech2-vctk](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/vctk/tts3)|[fastspeech2_vctk_ckpt_1.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_vctk_ckpt_1.2.0.zip)|[fastspeech2_vctk_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_vctk_static_1.1.0.zip) </br> [fastspeech2_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_vctk_onnx_1.1.0.zip) | 145MB| FastSpeech2| VCTK |[fastspeech2-vctk](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/vctk/tts3)|[fastspeech2_vctk_ckpt_1.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_vctk_ckpt_1.2.0.zip)|[fastspeech2_vctk_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_vctk_static_1.1.0.zip) </br> [fastspeech2_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_vctk_onnx_1.1.0.zip) </br> [fastspeech2_vctk_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_vctk_pdlite_1.3.0.zip)| 145MB|
FastSpeech2| ZH_EN |[fastspeech2-zh_en](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/zh_en_tts/tts3)|[fastspeech2_mix_ckpt_1.2.0.zip](https://paddlespeech.bj.bcebos.com/t2s/chinse_english_mixed/models/fastspeech2_mix_ckpt_1.2.0.zip)|[fastspeech2_mix_static_0.2.0.zip](https://paddlespeech.bj.bcebos.com/t2s/chinse_english_mixed/models/fastspeech2_mix_static_0.2.0.zip) </br> [fastspeech2_mix_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/t2s/chinse_english_mixed/models/fastspeech2_mix_onnx_0.2.0.zip) | 145MB| FastSpeech2| ZH_EN |[fastspeech2-zh_en](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/zh_en_tts/tts3)|[fastspeech2_mix_ckpt_1.2.0.zip](https://paddlespeech.bj.bcebos.com/t2s/chinse_english_mixed/models/fastspeech2_mix_ckpt_1.2.0.zip)|[fastspeech2_mix_static_0.2.0.zip](https://paddlespeech.bj.bcebos.com/t2s/chinse_english_mixed/models/fastspeech2_mix_static_0.2.0.zip) </br> [fastspeech2_mix_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/t2s/chinse_english_mixed/models/fastspeech2_mix_onnx_0.2.0.zip) | 145MB|
FastSpeech2| Male ||[fastspeech2_male_ckpt_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_male_ckpt_1.3.0.zip)| | | FastSpeech2| Male ||[fastspeech2_male_ckpt_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_male_ckpt_1.3.0.zip)| | |
### Vocoders ### Vocoders
Model Type | Dataset| Example Link | Pretrained Models| Static/ONNX Models|Size (static) Model Type | Dataset| Example Link | Pretrained Models| Static / ONNX / Paddle-Lite Models|Size (static)
:-----:| :-----:| :-----: | :-----:| :-----:| :-----: :-----:| :-----:| :-----: | :-----:| :-----:| :-----:
WaveFlow| LJSpeech |[waveflow-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/voc0)|[waveflow_ljspeech_ckpt_0.3.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/waveflow/waveflow_ljspeech_ckpt_0.3.zip)||| WaveFlow| LJSpeech |[waveflow-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/voc0)|[waveflow_ljspeech_ckpt_0.3.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/waveflow/waveflow_ljspeech_ckpt_0.3.zip)|||
Parallel WaveGAN| CSMSC |[PWGAN-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc1)|[pwg_baker_ckpt_0.4.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_baker_ckpt_0.4.zip)|[pwg_baker_static_0.4.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_baker_static_0.4.zip) </br> [pwgan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_csmsc_onnx_0.2.0.zip)|4.8MB| Parallel WaveGAN| CSMSC |[PWGAN-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc1)|[pwg_baker_ckpt_0.4.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_baker_ckpt_0.4.zip)|[pwg_baker_static_0.4.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_baker_static_0.4.zip) </br> [pwgan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_csmsc_onnx_0.2.0.zip) </br> [pwgan_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_csmsc_pdlite_1.3.0.zip)|4.8MB|
Parallel WaveGAN| LJSpeech |[PWGAN-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/voc1)|[pwg_ljspeech_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_ljspeech_ckpt_0.5.zip)|[pwgan_ljspeech_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_ljspeech_static_1.1.0.zip) </br> [pwgan_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_ljspeech_onnx_1.1.0.zip)|4.8MB| Parallel WaveGAN| LJSpeech |[PWGAN-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/voc1)|[pwg_ljspeech_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_ljspeech_ckpt_0.5.zip)|[pwgan_ljspeech_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_ljspeech_static_1.1.0.zip) </br> [pwgan_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_ljspeech_onnx_1.1.0.zip) </br> [pwgan_ljspeech_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_ljspeech_pdlite_1.3.0.zip)|4.8MB|
Parallel WaveGAN| AISHELL-3 |[PWGAN-aishell3](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/aishell3/voc1)|[pwg_aishell3_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_aishell3_ckpt_0.5.zip)| [pwgan_aishell3_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_aishell3_static_1.1.0.zip) </br> [pwgan_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_aishell3_onnx_1.1.0.zip)|4.8MB| Parallel WaveGAN| AISHELL-3 |[PWGAN-aishell3](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/aishell3/voc1)|[pwg_aishell3_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_aishell3_ckpt_0.5.zip)| [pwgan_aishell3_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_aishell3_static_1.1.0.zip) </br> [pwgan_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_aishell3_onnx_1.1.0.zip) </br> [pwgan_aishell3_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_aishell3_pdlite_1.3.0.zip)|4.8MB|
Parallel WaveGAN| VCTK |[PWGAN-vctk](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/vctk/voc1)|[pwg_vctk_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_vctk_ckpt_0.5.zip)|[pwgan_vctk_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_vctk_static_1.1.0.zip) </br> [pwgan_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_vctk_onnx_1.1.0.zip)|4.8MB| Parallel WaveGAN| VCTK |[PWGAN-vctk](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/vctk/voc1)|[pwg_vctk_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_vctk_ckpt_0.5.zip)|[pwgan_vctk_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_vctk_static_1.1.0.zip) </br> [pwgan_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_vctk_onnx_1.1.0.zip) </br> [pwgan_vctk_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_vctk_pdlite_1.3.0.zip)|4.8MB|
|Multi Band MelGAN | CSMSC |[MB MelGAN-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc3) | [mb_melgan_csmsc_ckpt_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_ckpt_0.1.1.zip) <br>[mb_melgan_baker_finetune_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_baker_finetune_ckpt_0.5.zip)|[mb_melgan_csmsc_static_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_static_0.1.1.zip) </br> [mb_melgan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_onnx_0.2.0.zip)|7.6MB| |Multi Band MelGAN | CSMSC |[MB MelGAN-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc3) | [mb_melgan_csmsc_ckpt_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_ckpt_0.1.1.zip) <br>[mb_melgan_baker_finetune_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_baker_finetune_ckpt_0.5.zip)|[mb_melgan_csmsc_static_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_static_0.1.1.zip) </br> [mb_melgan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_onnx_0.2.0.zip) </br> [mb_melgan_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_pdlite_1.3.0.zip)|7.6MB|
Style MelGAN | CSMSC |[Style MelGAN-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc4)|[style_melgan_csmsc_ckpt_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/style_melgan/style_melgan_csmsc_ckpt_0.1.1.zip)| | | Style MelGAN | CSMSC |[Style MelGAN-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc4)|[style_melgan_csmsc_ckpt_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/style_melgan/style_melgan_csmsc_ckpt_0.1.1.zip)| | |
HiFiGAN | CSMSC |[HiFiGAN-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc5)|[hifigan_csmsc_ckpt_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_ckpt_0.1.1.zip)|[hifigan_csmsc_static_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_static_0.1.1.zip) </br> [hifigan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_onnx_0.2.0.zip)|46MB| HiFiGAN | CSMSC |[HiFiGAN-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc5)|[hifigan_csmsc_ckpt_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_ckpt_0.1.1.zip)|[hifigan_csmsc_static_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_static_0.1.1.zip) </br> [hifigan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_onnx_0.2.0.zip) </br> [hifigan_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_pdlite_1.3.0.zip)|46MB|
HiFiGAN | LJSpeech |[HiFiGAN-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/voc5)|[hifigan_ljspeech_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_ckpt_0.2.0.zip)|[hifigan_ljspeech_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_static_1.1.0.zip) </br> [hifigan_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_onnx_1.1.0.zip) |49MB| HiFiGAN | LJSpeech |[HiFiGAN-ljspeech](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/ljspeech/voc5)|[hifigan_ljspeech_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_ckpt_0.2.0.zip)|[hifigan_ljspeech_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_static_1.1.0.zip) </br> [hifigan_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_onnx_1.1.0.zip) </br> [hifigan_ljspeech_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_pdlite_1.3.0.zip) |49MB|
HiFiGAN | AISHELL-3 |[HiFiGAN-aishell3](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/aishell3/voc5)|[hifigan_aishell3_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_ckpt_0.2.0.zip)|[hifigan_aishell3_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_static_1.1.0.zip) </br> [hifigan_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_onnx_1.1.0.zip)|46MB| HiFiGAN | AISHELL-3 |[HiFiGAN-aishell3](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/aishell3/voc5)|[hifigan_aishell3_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_ckpt_0.2.0.zip)|[hifigan_aishell3_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_static_1.1.0.zip) </br> [hifigan_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_onnx_1.1.0.zip) </br> [hifigan_aishell3_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_pdlite_1.3.0.zip)|46MB|
HiFiGAN | VCTK |[HiFiGAN-vctk](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/vctk/voc5)|[hifigan_vctk_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_ckpt_0.2.0.zip)|[hifigan_vctk_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_static_1.1.0.zip) </br> [hifigan_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_onnx_1.1.0.zip)|46MB| HiFiGAN | VCTK |[HiFiGAN-vctk](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/vctk/voc5)|[hifigan_vctk_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_ckpt_0.2.0.zip)|[hifigan_vctk_static_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_static_1.1.0.zip) </br> [hifigan_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_onnx_1.1.0.zip) </br> [hifigan_vctk_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_pdlite_1.3.0.zip)|46MB|
WaveRNN | CSMSC |[WaveRNN-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc6)|[wavernn_csmsc_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/wavernn/wavernn_csmsc_ckpt_0.2.0.zip)|[wavernn_csmsc_static_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/wavernn/wavernn_csmsc_static_0.2.0.zip)|18MB| WaveRNN | CSMSC |[WaveRNN-csmsc](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc6)|[wavernn_csmsc_ckpt_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/wavernn/wavernn_csmsc_ckpt_0.2.0.zip)|[wavernn_csmsc_static_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/wavernn/wavernn_csmsc_static_0.2.0.zip)|18MB|
Parallel WaveGAN| Male ||[pwg_male_ckpt_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_male_ckpt_1.3.0.zip)||| Parallel WaveGAN| Male ||[pwg_male_ckpt_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_male_ckpt_1.3.0.zip)|||

@ -226,6 +226,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [fastspeech2_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_onnx_1.1.0.zip) - [fastspeech2_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_onnx_1.1.0.zip)
The Paddle-Lite model can be downloaded here:
- [fastspeech2_aishell3_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_aishell3_pdlite_1.3.0.zip)
FastSpeech2 checkpoint contains files listed below. FastSpeech2 checkpoint contains files listed below.
```text ```text

@ -0,0 +1,32 @@
#!/bin/bash
train_output_path=$1
stage=0
stop_stage=0
# pwgan
if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=fastspeech2_aishell3 \
--voc=pwgan_aishell3 \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt \
--speaker_dict=dump/speaker_id_map.txt \
--spk_id=0
fi
# hifigan
if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=fastspeech2_aishell3 \
--voc=hifigan_aishell3 \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt \
--speaker_dict=dump/speaker_id_map.txt \
--spk_id=0
fi

@ -60,11 +60,11 @@ if [ ${stage} -le 6 ] && [ ${stop_stage} -ge 6 ]; then
fi fi
if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then
# This model is not supported, because 3 ops are not supported on 'arm'. These unsupported ops are: 'round, set_value, share_data'. ./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_aishell3 x86
# This model is not supported, because 4 ops are not supported on 'x86'. These unsupported ops are: 'matmul_v2, round, set_value, share_data'. ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_aishell3 x86
# ./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_aishell3 x86 # ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_aishell3 x86
# x86 ok, arm Segmentation fault fi
# ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_aishell3 x86
# x86 ok, arm ok if [ ${stage} -le 8 ] && [ ${stop_stage} -ge 8 ]; then
./local/export2lite.sh ${train_output_path} inference pdlite hifigan_aishell3 x86 CUDA_VISIBLE_DEVICES=${gpus} ./local/lite_predict.sh ${train_output_path} || exit -1
fi fi

@ -139,6 +139,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [pwgan_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_aishell3_onnx_1.1.0.zip) - [pwgan_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_aishell3_onnx_1.1.0.zip)
The Paddle-Lite model can be downloaded here:
- [pwgan_aishell3_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_aishell3_pdlite_1.3.0.zip)
Model | Step | eval/generator_loss | eval/log_stft_magnitude_loss:| eval/spectral_convergence_loss Model | Step | eval/generator_loss | eval/log_stft_magnitude_loss:| eval/spectral_convergence_loss
:-------------:| :------------:| :-----: | :-----: | :--------: :-------------:| :------------:| :-----: | :-----: | :--------:
default| 1(gpu) x 400000|1.968762|0.759008|0.218524 default| 1(gpu) x 400000|1.968762|0.759008|0.218524

@ -122,6 +122,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [hifigan_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_onnx_1.1.0.zip) - [hifigan_aishell3_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_onnx_1.1.0.zip)
The Paddle-Lite model can be downloaded here:
- [hifigan_aishell3_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_aishell3_pdlite_1.3.0.zip)
Model | Step | eval/generator_loss | eval/mel_loss| eval/feature_matching_loss Model | Step | eval/generator_loss | eval/mel_loss| eval/feature_matching_loss
:-------------:| :------------:| :-----: | :-----: | :--------: :-------------:| :------------:| :-----: | :-----: | :--------:
default| 1(gpu) x 2500000|24.060|0.1068|7.499 default| 1(gpu) x 2500000|24.060|0.1068|7.499

@ -230,6 +230,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [speedyspeech_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_onnx_0.2.0.zip) - [speedyspeech_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_onnx_0.2.0.zip)
The Paddle-Lite model can be downloaded here:
- [speedyspeech_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/speedyspeech/speedyspeech_csmsc_pdlite_1.3.0.zip)
Model | Step | eval/loss | eval/l1_loss | eval/duration_loss | eval/ssim_loss Model | Step | eval/loss | eval/l1_loss | eval/duration_loss | eval/ssim_loss
:-------------:| :------------:| :-----: | :-----: | :--------:|:--------: :-------------:| :------------:| :-----: | :-----: | :--------:|:--------:

@ -0,0 +1,43 @@
#!/bin/bash
train_output_path=$1
stage=0
stop_stage=0
# pwgan
if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=speedyspeech_csmsc \
--voc=pwgan_csmsc \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt \
--tones_dict=dump/tone_id_map.txt
fi
# for more GAN Vocoders
# multi band melgan
if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=speedyspeech_csmsc \
--voc=mb_melgan_csmsc \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt \
--tones_dict=dump/tone_id_map.txt
fi
# hifigan
if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=speedyspeech_csmsc \
--voc=hifigan_csmsc \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt \
--tones_dict=dump/tone_id_map.txt
fi

@ -63,13 +63,12 @@ fi
# must run after stage 3 (which stage generated static models) # must run after stage 3 (which stage generated static models)
if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then
# This model is not supported, because 3 ops are not supported on 'arm'. These unsupported ops are: 'round, set_value, share_data'.
# This model is not supported, because 4 ops are not supported on 'x86'. These unsupported ops are: 'matmul_v2, round, set_value, share_data'.
./local/export2lite.sh ${train_output_path} inference pdlite speedyspeech_csmsc x86 ./local/export2lite.sh ${train_output_path} inference pdlite speedyspeech_csmsc x86
# x86 ok, arm Segmentation fault ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_csmsc x86
# ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_csmsc x86
# x86 ok, arm Segmentation fault
# ./local/export2lite.sh ${train_output_path} inference pdlite mb_melgan_csmsc x86 # ./local/export2lite.sh ${train_output_path} inference pdlite mb_melgan_csmsc x86
# x86 ok, arm ok
# ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_csmsc x86 # ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_csmsc x86
fi fi
if [ ${stage} -le 8 ] && [ ${stop_stage} -ge 8 ]; then
CUDA_VISIBLE_DEVICES=${gpus} ./local/lite_predict.sh ${train_output_path} || exit -1
fi

@ -238,6 +238,12 @@ The ONNX model can be downloaded here:
- [fastspeech2_cnndecoder_csmsc_onnx_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_onnx_1.0.0.zip) - [fastspeech2_cnndecoder_csmsc_onnx_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_onnx_1.0.0.zip)
- [fastspeech2_cnndecoder_csmsc_streaming_onnx_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_streaming_onnx_1.0.0.zip) - [fastspeech2_cnndecoder_csmsc_streaming_onnx_1.0.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_streaming_onnx_1.0.0.zip)
The Paddle-Lite model can be downloaded here:
> please compile develop version of Paddle-Lite to export and run TTS models, cause TTS models are supported by https://github.com/PaddlePaddle/Paddle-Lite/pull/9587 and https://github.com/PaddlePaddle/Paddle-Lite/pull/9706
- [fastspeech2_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_csmsc_pdlite_1.3.0.zip)
- [fastspeech2_cnndecoder_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_pdlite_1.3.0.zip)
- [fastspeech2_cnndecoder_csmsc_streaming_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_cnndecoder_csmsc_streaming_pdlite_1.3.0.zip)
Model | Step | eval/loss | eval/l1_loss | eval/duration_loss | eval/pitch_loss| eval/energy_loss Model | Step | eval/loss | eval/l1_loss | eval/duration_loss | eval/pitch_loss| eval/energy_loss
:-------------:| :------------:| :-----: | :-----: | :--------: |:--------:|:---------: :-------------:| :------------:| :-----: | :-----: | :--------: |:--------:|:---------:
default| 2(gpu) x 76000|1.0991|0.59132|0.035815|0.31915|0.15287| default| 2(gpu) x 76000|1.0991|0.59132|0.035815|0.31915|0.15287|

@ -7,12 +7,12 @@ valid_targets=$5
model_name=${model%_*} model_name=${model%_*}
echo model_name: ${model_name} echo model_name: ${model_name}
suffix=${valid_targets%,*}
mkdir -p ${train_output_path}/${output_dir} mkdir -p ${train_output_path}/${output_dir}
paddle_lite_opt \ paddle_lite_opt \
--model_file ${train_output_path}/${model_dir}/${model}.pdmodel \ --model_file ${train_output_path}/${model_dir}/${model}.pdmodel \
--param_file ${train_output_path}/${model_dir}/${model}.pdiparams \ --param_file ${train_output_path}/${model_dir}/${model}.pdiparams \
--optimize_out ${train_output_path}/${output_dir}/${model}_${valid_targets} \ --optimize_out ${train_output_path}/${output_dir}/${model}_${suffix} \
--valid_targets ${valid_targets} --valid_targets ${valid_targets}

@ -0,0 +1,40 @@
#!/bin/bash
train_output_path=$1
stage=0
stop_stage=0
# pwgan
if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=fastspeech2_csmsc \
--voc=pwgan_csmsc \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt
fi
# for more GAN Vocoders
# multi band melgan
if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=fastspeech2_csmsc \
--voc=mb_melgan_csmsc \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt
fi
# hifigan
if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=fastspeech2_csmsc \
--voc=hifigan_csmsc \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt
fi

@ -0,0 +1,47 @@
#!/bin/bash
train_output_path=$1
stage=0
stop_stage=0
# pwgan
if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
python3 ${BIN_DIR}/../lite_predict_streaming.py \
--inference_dir=${train_output_path}/pdlite_streaming \
--am=fastspeech2_csmsc \
--am_stat=dump/train/speech_stats.npy \
--voc=pwgan_csmsc \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out_streaming \
--phones_dict=dump/phone_id_map.txt \
--am_streaming=True
fi
# for more GAN Vocoders
# multi band melgan
if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
python3 ${BIN_DIR}/../lite_predict_streaming.py \
--inference_dir=${train_output_path}/pdlite_streaming \
--am=fastspeech2_csmsc \
--am_stat=dump/train/speech_stats.npy \
--voc=mb_melgan_csmsc \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out_streaming \
--phones_dict=dump/phone_id_map.txt \
--am_streaming=True
fi
# hifigan
if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
python3 ${BIN_DIR}/../lite_predict_streaming.py \
--inference_dir=${train_output_path}/pdlite_streaming \
--am=fastspeech2_csmsc \
--am_stat=dump/train/speech_stats.npy \
--voc=hifigan_csmsc \
--text=${BIN_DIR}/../sentences.txt \
--output_dir=${train_output_path}/lite_infer_out_streaming \
--phones_dict=dump/phone_id_map.txt \
--am_streaming=True
fi

@ -64,13 +64,15 @@ fi
# must run after stage 3 (which stage generated static models) # must run after stage 3 (which stage generated static models)
if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then
# This model is not supported, because 3 ops are not supported on 'arm'. These unsupported ops are: 'round, set_value, share_data'. # NOTE by yuantian 2022.11.21: please compile develop version of Paddle-Lite to export and run TTS models,
# This model is not supported, because 4 ops are not supported on 'x86'. These unsupported ops are: 'matmul_v2, round, set_value, share_data'. # cause TTS models are supported by https://github.com/PaddlePaddle/Paddle-Lite/pull/9587
# and https://github.com/PaddlePaddle/Paddle-Lite/pull/9706
./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_csmsc x86 ./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_csmsc x86
# x86 ok, arm Segmentation fault ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_csmsc x86
# ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_csmsc x86
# x86 ok, arm Segmentation fault
# ./local/export2lite.sh ${train_output_path} inference pdlite mb_melgan_csmsc x86 # ./local/export2lite.sh ${train_output_path} inference pdlite mb_melgan_csmsc x86
# x86 ok, arm ok
# ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_csmsc x86 # ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_csmsc x86
fi fi
if [ ${stage} -le 8 ] && [ ${stop_stage} -ge 8 ]; then
CUDA_VISIBLE_DEVICES=${gpus} ./local/lite_predict.sh ${train_output_path} || exit -1
fi

@ -98,32 +98,27 @@ fi
# must run after stage 3 (which stage generated static models) # must run after stage 3 (which stage generated static models)
if [ ${stage} -le 11 ] && [ ${stop_stage} -ge 11 ]; then if [ ${stage} -le 11 ] && [ ${stop_stage} -ge 11 ]; then
# This model is not supported, because 3 ops are not supported on 'arm'. These unsupported ops are: 'round, set_value, share_data'.
# This model is not supported, because 4 ops are not supported on 'x86'. These unsupported ops are: 'matmul_v2, round, set_value, share_data'.
./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_csmsc x86 ./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_csmsc x86
# x86 ok, arm Segmentation fault ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_csmsc x86
# ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_csmsc x86
# x86 ok, arm Segmentation fault
# ./local/export2lite.sh ${train_output_path} inference pdlite mb_melgan_csmsc x86 # ./local/export2lite.sh ${train_output_path} inference pdlite mb_melgan_csmsc x86
# x86 ok, arm ok
# ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_csmsc x86 # ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_csmsc x86
fi fi
# must run after stage 5 (which stage generated static models)
if [ ${stage} -le 12 ] && [ ${stop_stage} -ge 12 ]; then if [ ${stage} -le 12 ] && [ ${stop_stage} -ge 12 ]; then
CUDA_VISIBLE_DEVICES=${gpus} ./local/lite_predict.sh ${train_output_path} || exit -1
fi
# must run after stage 5 (which stage generated static models)
if [ ${stage} -le 13 ] && [ ${stop_stage} -ge 13 ]; then
# streaming acoustic model # streaming acoustic model
# This model is not supported, because 3 ops are not supported on 'arm'. These unsupported ops are: 'round, set_value, share_data'.
# This model is not supported, because 4 ops are not supported on 'x86'. These unsupported ops are: 'matmul_v2, round, set_value, share_data'.
# ./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_csmsc x86
./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming fastspeech2_csmsc_am_encoder_infer x86 ./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming fastspeech2_csmsc_am_encoder_infer x86
# x86 ok, arm Segmentation fault
./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming fastspeech2_csmsc_am_decoder x86 ./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming fastspeech2_csmsc_am_decoder x86
# x86 ok, arm Segmentation fault
./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming fastspeech2_csmsc_am_postnet x86 ./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming fastspeech2_csmsc_am_postnet x86
# x86 ok, arm Segmentation fault ./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming pwgan_csmsc x86
# ./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming pwgan_csmsc x86
# x86 ok, arm Segmentation fault
# ./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming mb_melgan_csmsc x86 # ./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming mb_melgan_csmsc x86
# x86 ok, arm ok
# ./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming hifigan_csmsc x86 # ./local/export2lite.sh ${train_output_path} inference_streaming pdlite_streaming hifigan_csmsc x86
fi fi
if [ ${stage} -le 14 ] && [ ${stop_stage} -ge 14 ]; then
CUDA_VISIBLE_DEVICES=${gpus} ./local/lite_predict_streaming.sh ${train_output_path} || exit -1
fi

@ -136,6 +136,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [pwgan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_csmsc_onnx_0.2.0.zip) - [pwgan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_csmsc_onnx_0.2.0.zip)
The Paddle-Lite model can be downloaded here:
- [pwgan_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_csmsc_pdlite_1.3.0.zip)
Model | Step | eval/generator_loss | eval/log_stft_magnitude_loss| eval/spectral_convergence_loss Model | Step | eval/generator_loss | eval/log_stft_magnitude_loss| eval/spectral_convergence_loss
:-------------:| :------------:| :-----: | :-----: | :--------: :-------------:| :------------:| :-----: | :-----: | :--------:
default| 1(gpu) x 400000|1.948763|0.670098|0.248882 default| 1(gpu) x 400000|1.948763|0.670098|0.248882

@ -164,6 +164,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [mb_melgan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_onnx_0.2.0.zip) - [mb_melgan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_onnx_0.2.0.zip)
The Paddle-Lite model can be downloaded here:
- [mb_melgan_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/mb_melgan/mb_melgan_csmsc_pdlite_1.3.0.zip)
Model | Step | eval/generator_loss | eval/log_stft_magnitude_loss|eval/spectral_convergence_loss |eval/sub_log_stft_magnitude_loss|eval/sub_spectral_convergence_loss Model | Step | eval/generator_loss | eval/log_stft_magnitude_loss|eval/spectral_convergence_loss |eval/sub_log_stft_magnitude_loss|eval/sub_spectral_convergence_loss
:-------------:| :------------:| :-----: | :-----: | :--------:| :--------:| :--------: :-------------:| :------------:| :-----: | :-----: | :--------:| :--------:| :--------:
default| 1(gpu) x 1000000| 2.4851|0.71778 |0.2761 |0.66334 |0.2777| default| 1(gpu) x 1000000| 2.4851|0.71778 |0.2761 |0.66334 |0.2777|

@ -121,6 +121,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [hifigan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_onnx_0.2.0.zip) - [hifigan_csmsc_onnx_0.2.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_onnx_0.2.0.zip)
The Paddle-Lite model can be downloaded here:
- [hifigan_csmsc_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_pdlite_1.3.0.zip)
Model | Step | eval/generator_loss | eval/mel_loss| eval/feature_matching_loss Model | Step | eval/generator_loss | eval/mel_loss| eval/feature_matching_loss
:-------------:| :------------:| :-----: | :-----: | :--------: :-------------:| :------------:| :-----: | :-----: | :--------:
default| 1(gpu) x 2500000|24.927|0.1262|7.554 default| 1(gpu) x 2500000|24.927|0.1262|7.554

@ -221,6 +221,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [fastspeech2_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_ljspeech_onnx_1.1.0.zip) - [fastspeech2_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_ljspeech_onnx_1.1.0.zip)
The Paddle-Lite model can be downloaded here:
- [fastspeech2_ljspeech_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_ljspeech_pdlite_1.3.0.zip)
Model | Step | eval/loss | eval/l1_loss | eval/duration_loss | eval/pitch_loss| eval/energy_loss Model | Step | eval/loss | eval/l1_loss | eval/duration_loss | eval/pitch_loss| eval/energy_loss
:-------------:| :------------:| :-----: | :-----: | :--------: |:--------:|:---------: :-------------:| :------------:| :-----: | :-----: | :--------: |:--------:|:---------:

@ -0,0 +1,30 @@
#!/bin/bash
train_output_path=$1
stage=0
stop_stage=0
# pwgan
if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=fastspeech2_ljspeech \
--voc=pwgan_ljspeech \
--text=${BIN_DIR}/../sentences_en.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt \
--lang=en
fi
# hifigan
if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=fastspeech2_ljspeech \
--voc=hifigan_ljspeech \
--text=${BIN_DIR}/../sentences_en.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt \
--lang=en
fi

@ -62,11 +62,11 @@ fi
# must run after stage 3 (which stage generated static models) # must run after stage 3 (which stage generated static models)
if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then
# This model is not supported, because 3 ops are not supported on 'arm'. These unsupported ops are: 'round, set_value, share_data'.
# This model is not supported, because 4 ops are not supported on 'x86'. These unsupported ops are: 'matmul_v2, round, set_value, share_data'.
./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_ljspeech x86 ./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_ljspeech x86
# x86 ok, arm Segmentation fault ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_ljspeech x86
# ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_ljspeech x86
# x86 ok, arm ok
# ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_ljspeech x86 # ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_ljspeech x86
fi fi
if [ ${stage} -le 8 ] && [ ${stop_stage} -ge 8 ]; then
CUDA_VISIBLE_DEVICES=${gpus} ./local/lite_predict.sh ${train_output_path} || exit -1
fi

@ -136,6 +136,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [pwgan_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_ljspeech_onnx_1.1.0.zip) - [pwgan_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_ljspeech_onnx_1.1.0.zip)
The Paddle-Lite model can be downloaded here:
- [pwgan_ljspeech_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_ljspeech_pdlite_1.3.0.zip)
Parallel WaveGAN checkpoint contains files listed below. Parallel WaveGAN checkpoint contains files listed below.

@ -121,6 +121,8 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [hifigan_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_onnx_1.1.0.zip) - [hifigan_ljspeech_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_onnx_1.1.0.zip)
The Paddle-Lite model can be downloaded here:
- [hifigan_ljspeech_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_ljspeech_pdlite_1.3.0.zip)
Model | Step | eval/generator_loss | eval/mel_loss| eval/feature_matching_loss Model | Step | eval/generator_loss | eval/mel_loss| eval/feature_matching_loss
:-------------:| :------------:| :-----: | :-----: | :--------: :-------------:| :------------:| :-----: | :-----: | :--------:

@ -224,6 +224,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [fastspeech2_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_vctk_onnx_1.1.0.zip) - [fastspeech2_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_vctk_onnx_1.1.0.zip)
The Paddle-Lite model can be downloaded here:
- [fastspeech2_vctk_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/fastspeech2/fastspeech2_vctk_pdlite_1.3.0.zip)
FastSpeech2 checkpoint contains files listed below. FastSpeech2 checkpoint contains files listed below.
```text ```text
fastspeech2_vctk_ckpt_1.2.0 fastspeech2_vctk_ckpt_1.2.0

@ -0,0 +1,34 @@
#!/bin/bash
train_output_path=$1
stage=0
stop_stage=0
# pwgan
if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=fastspeech2_vctk \
--voc=pwgan_vctk \
--text=${BIN_DIR}/../sentences_en.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt \
--speaker_dict=dump/speaker_id_map.txt \
--spk_id=0 \
--lang=en
fi
# hifigan
if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
python3 ${BIN_DIR}/../lite_predict.py \
--inference_dir=${train_output_path}/pdlite \
--am=fastspeech2_vctk \
--voc=hifigan_vctk \
--text=${BIN_DIR}/../sentences_en.txt \
--output_dir=${train_output_path}/lite_infer_out \
--phones_dict=dump/phone_id_map.txt \
--speaker_dict=dump/speaker_id_map.txt \
--spk_id=0 \
--lang=en
fi

@ -61,11 +61,11 @@ fi
# must run after stage 3 (which stage generated static models) # must run after stage 3 (which stage generated static models)
if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then if [ ${stage} -le 7 ] && [ ${stop_stage} -ge 7 ]; then
# This model is not supported, because 3 ops are not supported on 'arm'. These unsupported ops are: 'round, set_value, share_data'.
# This model is not supported, because 4 ops are not supported on 'x86'. These unsupported ops are: 'matmul_v2, round, set_value, share_data'.
./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_vctk x86 ./local/export2lite.sh ${train_output_path} inference pdlite fastspeech2_vctk x86
# x86 ok, arm Segmentation fault ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_vctk x86
# ./local/export2lite.sh ${train_output_path} inference pdlite pwgan_vctk x86
# x86 ok, arm ok
# ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_vctk x86 # ./local/export2lite.sh ${train_output_path} inference pdlite hifigan_vctk x86
fi fi
if [ ${stage} -le 8 ] && [ ${stop_stage} -ge 8 ]; then
CUDA_VISIBLE_DEVICES=${gpus} ./local/lite_predict.sh ${train_output_path} || exit -1
fi

@ -141,6 +141,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [pwgan_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_vctk_onnx_1.1.0.zip) - [pwgan_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_vctk_onnx_1.1.0.zip)
The Paddle-Lite model can be downloaded here:
- [pwgan_vctk_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwgan_vctk_pdlite_1.3.0.zip)
Parallel WaveGAN checkpoint contains files listed below. Parallel WaveGAN checkpoint contains files listed below.

@ -127,6 +127,9 @@ The static model can be downloaded here:
The ONNX model can be downloaded here: The ONNX model can be downloaded here:
- [hifigan_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_onnx_1.1.0.zip) - [hifigan_vctk_onnx_1.1.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_onnx_1.1.0.zip)
The Paddle-Lite model can be downloaded here:
- [hifigan_vctk_pdlite_1.3.0.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_vctk_pdlite_1.3.0.zip)
Model | Step | eval/generator_loss | eval/mel_loss| eval/feature_matching_loss Model | Step | eval/generator_loss | eval/mel_loss| eval/feature_matching_loss
:-------------:| :------------:| :-----: | :-----: | :--------: :-------------:| :------------:| :-----: | :-----: | :--------:

@ -145,7 +145,7 @@ def main():
# warmup # warmup
for utt_id, sentence in sentences[:3]: for utt_id, sentence in sentences[:3]:
with timer() as t: with timer() as t:
am_output_data = get_am_output( mel = get_am_output(
input=sentence, input=sentence,
am_predictor=am_predictor, am_predictor=am_predictor,
am=args.am, am=args.am,
@ -154,12 +154,11 @@ def main():
merge_sentences=merge_sentences, merge_sentences=merge_sentences,
speaker_dict=args.speaker_dict, speaker_dict=args.speaker_dict,
spk_id=args.spk_id, ) spk_id=args.spk_id, )
wav = get_voc_output( wav = get_voc_output(voc_predictor=voc_predictor, input=mel)
voc_predictor=voc_predictor, input=am_output_data)
speed = wav.size / t.elapse speed = wav.size / t.elapse
rtf = fs / speed rtf = fs / speed
print( print(
f"{utt_id}, mel: {am_output_data.shape}, wave: {wav.shape}, time: {t.elapse}s, Hz: {speed}, RTF: {rtf}." f"{utt_id}, mel: {mel.shape}, wave: {wav.shape}, time: {t.elapse}s, Hz: {speed}, RTF: {rtf}."
) )
print("warm up done!") print("warm up done!")
@ -168,7 +167,7 @@ def main():
T = 0 T = 0
for utt_id, sentence in sentences: for utt_id, sentence in sentences:
with timer() as t: with timer() as t:
am_output_data = get_am_output( mel = get_am_output(
input=sentence, input=sentence,
am_predictor=am_predictor, am_predictor=am_predictor,
am=args.am, am=args.am,
@ -177,8 +176,7 @@ def main():
merge_sentences=merge_sentences, merge_sentences=merge_sentences,
speaker_dict=args.speaker_dict, speaker_dict=args.speaker_dict,
spk_id=args.spk_id, ) spk_id=args.spk_id, )
wav = get_voc_output( wav = get_voc_output(voc_predictor=voc_predictor, input=mel)
voc_predictor=voc_predictor, input=am_output_data)
N += wav.size N += wav.size
T += t.elapse T += t.elapse
@ -187,7 +185,7 @@ def main():
sf.write(output_dir / (utt_id + ".wav"), wav, samplerate=fs) sf.write(output_dir / (utt_id + ".wav"), wav, samplerate=fs)
print( print(
f"{utt_id}, mel: {am_output_data.shape}, wave: {wav.shape}, time: {t.elapse}s, Hz: {speed}, RTF: {rtf}." f"{utt_id}, mel: {mel.shape}, wave: {wav.shape}, time: {t.elapse}s, Hz: {speed}, RTF: {rtf}."
) )
print(f"{utt_id} done!") print(f"{utt_id} done!")

@ -0,0 +1,168 @@
# Copyright (c) 2022 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 argparse
from pathlib import Path
import soundfile as sf
from timer import timer
from paddlespeech.t2s.exps.syn_utils import get_frontend
from paddlespeech.t2s.exps.syn_utils import get_lite_am_output
from paddlespeech.t2s.exps.syn_utils import get_lite_predictor
from paddlespeech.t2s.exps.syn_utils import get_lite_voc_output
from paddlespeech.t2s.exps.syn_utils import get_sentences
def parse_args():
parser = argparse.ArgumentParser(
description="Paddle Infernce with acoustic model & vocoder.")
# acoustic model
parser.add_argument(
'--am',
type=str,
default='fastspeech2_csmsc',
choices=[
'speedyspeech_csmsc',
'fastspeech2_csmsc',
'fastspeech2_aishell3',
'fastspeech2_ljspeech',
'fastspeech2_vctk',
'fastspeech2_mix',
],
help='Choose acoustic model type of tts task.')
parser.add_argument(
"--phones_dict", type=str, default=None, help="phone vocabulary file.")
parser.add_argument(
"--tones_dict", type=str, default=None, help="tone vocabulary file.")
parser.add_argument(
"--speaker_dict", type=str, default=None, help="speaker id map file.")
parser.add_argument(
'--spk_id',
type=int,
default=0,
help='spk id for multi speaker acoustic model')
# voc
parser.add_argument(
'--voc',
type=str,
default='pwgan_csmsc',
choices=[
'pwgan_csmsc',
'pwgan_aishell3',
'pwgan_ljspeech',
'pwgan_vctk',
'mb_melgan_csmsc',
'hifigan_csmsc',
'hifigan_aishell3',
'hifigan_ljspeech',
'hifigan_vctk',
],
help='Choose vocoder type of tts task.')
# other
parser.add_argument(
'--lang',
type=str,
default='zh',
help='Choose model language. zh or en or mix')
parser.add_argument(
"--text",
type=str,
help="text to synthesize, a 'utt_id sentence' pair per line")
parser.add_argument(
"--inference_dir", type=str, help="dir to save inference models")
parser.add_argument("--output_dir", type=str, help="output dir")
args, _ = parser.parse_known_args()
return args
# only inference for models trained with csmsc now
def main():
args = parse_args()
# frontend
frontend = get_frontend(
lang=args.lang,
phones_dict=args.phones_dict,
tones_dict=args.tones_dict)
# am_predictor
am_predictor = get_lite_predictor(
model_dir=args.inference_dir, model_file=args.am + "_x86.nb")
# model: {model_name}_{dataset}
am_dataset = args.am[args.am.rindex('_') + 1:]
# voc_predictor
voc_predictor = get_lite_predictor(
model_dir=args.inference_dir, model_file=args.voc + "_x86.nb")
output_dir = Path(args.output_dir)
output_dir.mkdir(parents=True, exist_ok=True)
sentences = get_sentences(text_file=args.text, lang=args.lang)
merge_sentences = True
fs = 24000 if am_dataset != 'ljspeech' else 22050
# warmup
for utt_id, sentence in sentences[:3]:
with timer() as t:
mel = get_lite_am_output(
input=sentence,
am_predictor=am_predictor,
am=args.am,
frontend=frontend,
lang=args.lang,
merge_sentences=merge_sentences,
speaker_dict=args.speaker_dict,
spk_id=args.spk_id, )
wav = get_lite_voc_output(voc_predictor=voc_predictor, input=mel)
speed = wav.size / t.elapse
rtf = fs / speed
print(
f"{utt_id}, mel: {mel.shape}, wave: {wav.shape}, time: {t.elapse}s, Hz: {speed}, RTF: {rtf}."
)
print("warm up done!")
N = 0
T = 0
for utt_id, sentence in sentences:
with timer() as t:
mel = get_lite_am_output(
input=sentence,
am_predictor=am_predictor,
am=args.am,
frontend=frontend,
lang=args.lang,
merge_sentences=merge_sentences,
speaker_dict=args.speaker_dict,
spk_id=args.spk_id, )
wav = get_lite_voc_output(voc_predictor=voc_predictor, input=mel)
N += wav.size
T += t.elapse
speed = wav.size / t.elapse
rtf = fs / speed
sf.write(output_dir / (utt_id + ".wav"), wav, samplerate=fs)
print(
f"{utt_id}, mel: {mel.shape}, wave: {wav.shape}, time: {t.elapse}s, Hz: {speed}, RTF: {rtf}."
)
print(f"{utt_id} done!")
print(f"generation speed: {N / T}Hz, RTF: {fs / (N / T) }")
if __name__ == "__main__":
main()

@ -0,0 +1,230 @@
# Copyright (c) 2022 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 argparse
from pathlib import Path
import numpy as np
import soundfile as sf
from timer import timer
from paddlespeech.t2s.exps.syn_utils import denorm
from paddlespeech.t2s.exps.syn_utils import get_chunks
from paddlespeech.t2s.exps.syn_utils import get_frontend
from paddlespeech.t2s.exps.syn_utils import get_lite_am_sublayer_output
from paddlespeech.t2s.exps.syn_utils import get_lite_predictor
from paddlespeech.t2s.exps.syn_utils import get_lite_streaming_am_output
from paddlespeech.t2s.exps.syn_utils import get_lite_voc_output
from paddlespeech.t2s.exps.syn_utils import get_sentences
from paddlespeech.t2s.exps.syn_utils import run_frontend
from paddlespeech.t2s.utils import str2bool
def parse_args():
parser = argparse.ArgumentParser(
description="Paddle Infernce with acoustic model & vocoder.")
# acoustic model
parser.add_argument(
'--am',
type=str,
default='fastspeech2_csmsc',
choices=['fastspeech2_csmsc'],
help='Choose acoustic model type of tts task.')
parser.add_argument(
"--am_stat",
type=str,
default=None,
help="mean and standard deviation used to normalize spectrogram when training acoustic model."
)
parser.add_argument(
"--phones_dict", type=str, default=None, help="phone vocabulary file.")
parser.add_argument(
"--tones_dict", type=str, default=None, help="tone vocabulary file.")
parser.add_argument(
"--speaker_dict", type=str, default=None, help="speaker id map file.")
parser.add_argument(
'--spk_id',
type=int,
default=0,
help='spk id for multi speaker acoustic model')
# voc
parser.add_argument(
'--voc',
type=str,
default='pwgan_csmsc',
choices=['pwgan_csmsc', 'mb_melgan_csmsc', 'hifigan_csmsc'],
help='Choose vocoder type of tts task.')
# other
parser.add_argument(
'--lang',
type=str,
default='zh',
help='Choose model language. zh or en')
parser.add_argument(
"--text",
type=str,
help="text to synthesize, a 'utt_id sentence' pair per line")
parser.add_argument(
"--inference_dir", type=str, help="dir to save inference models")
parser.add_argument("--output_dir", type=str, help="output dir")
# inference
# streaming related
parser.add_argument(
"--am_streaming",
type=str2bool,
default=False,
help="whether use streaming acoustic model")
parser.add_argument(
"--block_size", type=int, default=42, help="block size of am streaming")
parser.add_argument(
"--pad_size", type=int, default=12, help="pad size of am streaming")
args, _ = parser.parse_known_args()
return args
# only inference for models trained with csmsc now
def main():
args = parse_args()
# frontend
frontend = get_frontend(
lang=args.lang,
phones_dict=args.phones_dict,
tones_dict=args.tones_dict)
# am_predictor
am_encoder_infer_predictor = get_lite_predictor(
model_dir=args.inference_dir,
model_file=args.am + "_am_encoder_infer" + "_x86.nb")
am_decoder_predictor = get_lite_predictor(
model_dir=args.inference_dir,
model_file=args.am + "_am_decoder" + "_x86.nb")
am_postnet_predictor = get_lite_predictor(
model_dir=args.inference_dir,
model_file=args.am + "_am_postnet" + "_x86.nb")
am_mu, am_std = np.load(args.am_stat)
# model: {model_name}_{dataset}
am_dataset = args.am[args.am.rindex('_') + 1:]
# voc_predictor
voc_predictor = get_lite_predictor(
model_dir=args.inference_dir, model_file=args.voc + "_x86.nb")
output_dir = Path(args.output_dir)
output_dir.mkdir(parents=True, exist_ok=True)
sentences = get_sentences(text_file=args.text, lang=args.lang)
merge_sentences = True
fs = 24000 if am_dataset != 'ljspeech' else 22050
# warmup
for utt_id, sentence in sentences[:3]:
with timer() as t:
normalized_mel = get_lite_streaming_am_output(
input=sentence,
am_encoder_infer_predictor=am_encoder_infer_predictor,
am_decoder_predictor=am_decoder_predictor,
am_postnet_predictor=am_postnet_predictor,
frontend=frontend,
lang=args.lang,
merge_sentences=merge_sentences, )
mel = denorm(normalized_mel, am_mu, am_std)
wav = get_lite_voc_output(voc_predictor=voc_predictor, input=mel)
speed = wav.size / t.elapse
rtf = fs / speed
print(
f"{utt_id}, mel: {mel.shape}, wave: {wav.shape}, time: {t.elapse}s, Hz: {speed}, RTF: {rtf}."
)
print("warm up done!")
N = 0
T = 0
block_size = args.block_size
pad_size = args.pad_size
get_tone_ids = False
for utt_id, sentence in sentences:
with timer() as t:
# frontend
frontend_dict = run_frontend(
frontend=frontend,
text=sentence,
merge_sentences=merge_sentences,
get_tone_ids=get_tone_ids,
lang=args.lang)
phone_ids = frontend_dict['phone_ids']
phones = phone_ids[0].numpy()
# acoustic model
orig_hs = get_lite_am_sublayer_output(
am_encoder_infer_predictor, input=phones)
if args.am_streaming:
hss = get_chunks(orig_hs, block_size, pad_size)
chunk_num = len(hss)
mel_list = []
for i, hs in enumerate(hss):
am_decoder_output = get_lite_am_sublayer_output(
am_decoder_predictor, input=hs)
am_postnet_output = get_lite_am_sublayer_output(
am_postnet_predictor,
input=np.transpose(am_decoder_output, (0, 2, 1)))
am_output_data = am_decoder_output + np.transpose(
am_postnet_output, (0, 2, 1))
normalized_mel = am_output_data[0]
sub_mel = denorm(normalized_mel, am_mu, am_std)
# clip output part of pad
if i == 0:
sub_mel = sub_mel[:-pad_size]
elif i == chunk_num - 1:
# 最后一块的右侧一定没有 pad 够
sub_mel = sub_mel[pad_size:]
else:
# 倒数几块的右侧也可能没有 pad 够
sub_mel = sub_mel[pad_size:(block_size + pad_size) -
sub_mel.shape[0]]
mel_list.append(sub_mel)
mel = np.concatenate(mel_list, axis=0)
else:
am_decoder_output = get_lite_am_sublayer_output(
am_decoder_predictor, input=orig_hs)
am_postnet_output = get_lite_am_sublayer_output(
am_postnet_predictor,
input=np.transpose(am_decoder_output, (0, 2, 1)))
am_output_data = am_decoder_output + np.transpose(
am_postnet_output, (0, 2, 1))
normalized_mel = am_output_data[0]
mel = denorm(normalized_mel, am_mu, am_std)
# vocoder
wav = get_lite_voc_output(voc_predictor=voc_predictor, input=mel)
N += wav.size
T += t.elapse
speed = wav.size / t.elapse
rtf = fs / speed
sf.write(output_dir / (utt_id + ".wav"), wav, samplerate=24000)
print(
f"{utt_id}, mel: {mel.shape}, wave: {wav.shape}, time: {t.elapse}s, Hz: {speed}, RTF: {rtf}."
)
print(f"{utt_id} done!")
print(f"generation speed: {N / T}Hz, RTF: {fs / (N / T) }")
if __name__ == "__main__":
main()

@ -26,6 +26,8 @@ import paddle
from paddle import inference from paddle import inference
from paddle import jit from paddle import jit
from paddle.static import InputSpec from paddle.static import InputSpec
from paddlelite.lite import create_paddle_predictor
from paddlelite.lite import MobileConfig
from yacs.config import CfgNode from yacs.config import CfgNode
from paddlespeech.t2s.datasets.data_table import DataTable from paddlespeech.t2s.datasets.data_table import DataTable
@ -510,3 +512,105 @@ def get_sess(model_path: Optional[os.PathLike],
sess = ort.InferenceSession( sess = ort.InferenceSession(
model_path, providers=providers, sess_options=sess_options) model_path, providers=providers, sess_options=sess_options)
return sess return sess
# Paddle-Lite
def get_lite_predictor(model_dir: Optional[os.PathLike]=None,
model_file: Optional[os.PathLike]=None,
cpu_threads: int=1):
config = MobileConfig()
config.set_model_from_file(str(Path(model_dir) / model_file))
predictor = create_paddle_predictor(config)
return predictor
def get_lite_am_output(
input: str,
am_predictor,
am: str,
frontend: object,
lang: str='zh',
merge_sentences: bool=True,
speaker_dict: Optional[os.PathLike]=None,
spk_id: int=0, ):
am_name = am[:am.rindex('_')]
am_dataset = am[am.rindex('_') + 1:]
get_spk_id = False
get_tone_ids = False
if am_name == 'speedyspeech':
get_tone_ids = True
if am_dataset in {"aishell3", "vctk", "mix"} and speaker_dict:
get_spk_id = True
spk_id = np.array([spk_id])
frontend_dict = run_frontend(
frontend=frontend,
text=input,
merge_sentences=merge_sentences,
get_tone_ids=get_tone_ids,
lang=lang)
if get_tone_ids:
tone_ids = frontend_dict['tone_ids']
tones = tone_ids[0].numpy()
tones_handle = am_predictor.get_input(1)
tones_handle.from_numpy(tones)
if get_spk_id:
spk_id_handle = am_predictor.get_input(1)
spk_id_handle.from_numpy(spk_id)
phone_ids = frontend_dict['phone_ids']
phones = phone_ids[0].numpy()
phones_handle = am_predictor.get_input(0)
phones_handle.from_numpy(phones)
am_predictor.run()
am_output_handle = am_predictor.get_output(0)
am_output_data = am_output_handle.numpy()
return am_output_data
def get_lite_voc_output(voc_predictor, input):
mel_handle = voc_predictor.get_input(0)
mel_handle.from_numpy(input)
voc_predictor.run()
voc_output_handle = voc_predictor.get_output(0)
wav = voc_output_handle.numpy()
return wav
def get_lite_am_sublayer_output(am_sublayer_predictor, input):
input_handle = am_sublayer_predictor.get_input(0)
input_handle.from_numpy(input)
am_sublayer_predictor.run()
am_sublayer_handle = am_sublayer_predictor.get_output(0)
am_sublayer_output = am_sublayer_handle.numpy()
return am_sublayer_output
def get_lite_streaming_am_output(input: str,
am_encoder_infer_predictor,
am_decoder_predictor,
am_postnet_predictor,
frontend,
lang: str='zh',
merge_sentences: bool=True):
get_tone_ids = False
frontend_dict = run_frontend(
frontend=frontend,
text=input,
merge_sentences=merge_sentences,
get_tone_ids=get_tone_ids,
lang=lang)
phone_ids = frontend_dict['phone_ids']
phones = phone_ids[0].numpy()
am_encoder_infer_output = get_lite_am_sublayer_output(
am_encoder_infer_predictor, input=phones)
am_decoder_output = get_lite_am_sublayer_output(
am_decoder_predictor, input=am_encoder_infer_output)
am_postnet_output = get_lite_am_sublayer_output(
am_postnet_predictor, input=np.transpose(am_decoder_output, (0, 2, 1)))
am_output_data = am_decoder_output + np.transpose(am_postnet_output,
(0, 2, 1))
normalized_mel = am_output_data[0]
return normalized_mel

@ -75,6 +75,7 @@ base = [
"braceexpand", "braceexpand",
"pyyaml", "pyyaml",
"pybind11", "pybind11",
"paddlelite",
"paddleslim==2.3.4", "paddleslim==2.3.4",
] ]

Loading…
Cancel
Save