PaddleSpeech r1.4.2

[luotao1]

S2T

• Add WavLM ASR-en, WavLM fine-tuning for ASR on LibriSpeech. Adding WavLM implementation #3242 by @jiamingkong
• Add HuBERT ASR-en, HuBERT fine-tuning for ASR on LibriSpeech. [ASR] Support Hubert, fintuned on the librispeech dataset #3088 by @Zth9730
• Add Squeezeformer model. [ASR]add squeezeformer model #2755 by @yeyupiaoling
• Add AMP for U2 conformer. [ASR] add amp for U2 conformer #3167 by @zxcd
• Mv dataset into paddlespeech.dataset. [s2t] mv dataset into paddlespeech.dataset #3183 [s2t] move s2t data preprocess into paddlespeech.dataset #3189 by @zh794390558
• Fix example/aishell local/train.sh if condition bug. fix example/aishell local/train.sh if condition bug, test=asr #3146 by @lemondy
• Fix cli args to config. [s2t] fix cli args to config #3194 by @zh794390558
• Fix scaler save, load, unscale_ blow, grad_clip. by @zxcd

T2S

• Add SVS(Singing Voice Synthesis) examples with Opencpop dataset, including DiffSinger、PWGAN ([TTS] add opencpop PWGAN example #3031 by @lym0302) and HiFiGAN ([TTS] add opencpop HIFIGAN example #3038 by @lym0302), the effect is continuously optimized.
• Add SVS frontend. [TTS] add svs frontend #3062 by @lym0302
• Add TTS iSTFTNet ([TTS]【Hackathon + No.190】 + 模型复现：iSTFTNet #3006 by @longRookie), TTS JETS ([TTS] [黑客松]Add JETS #3109 by @ljhzxc)
• Starganv2: by @yt605155624
  • Clean starganv2 vc model code and add docstring. [TTS]clean starganv2 vc model code and add docstring #2987
  • Add starganv2 vc trainer. [TTS]add starganv2 vc trainer #3143 [TTS]StarGANv2 VC fix some trainer bugs, add add reset_parameters #3182
  • Add StarGANv2VC preprocess. [TTS]add StarGANv2VC preprocess #3163
  • Fix losses of StarGAN v2 VC. [TTS]Fix losses of StarGAN v2 VC  #3184
• Support for LITE: by @yt605155624
  • Fix elementwise_floordiv's fill_constant. [TTS]fix elementwise_floordiv's fill_constant #3075
  • Fix VITS lite infer. [TTS]Fix VITS lite infer #3098
  • Fix vits reduce_sum's input/output dtype. [TTS]fix vits reduce_sum's input/output dtype #3028
  • Fix dtype diff of last expand_v2 op of VITS. [TTS]fix dtype diff of last expand_v2 op of VITS #3041
  • Fix input dtype of elementwise_mul op from bool to int64. [TTS]fix input dtype of elementwise_mul op from bool to int64 #3054
• Add XPU support for SpeedySpeech and FastSpeech2. Add XPU support for SpeedySpeech #3502 Add XPU support for FastSpeech2 #3514 by @USTCKAY
• Fix some preprocess bugs. [TTS]fix some preprocess bugs #3155 by @yt605155624
• Fix bug of merge_yi function. 【Fix Speech Issue No.8】issue 3652 merge_yi function has a bug #3786 by @mattheliu

Server

• Add code-switch conformer_talcs support. fix: 🐛 修复服务端 python ASREngine 无法使用conformer_talcs模型 #3230 by @Gsonovb
• Add subtitle file (.srt format) generation example. add function for generating srt file #3123 by @twoDogy
• Fix: add file read encoding. 增加文件编码读取 #3606 by @Coloryr
  Install & Benchmark
• Update paddle2onnx to newest install version. [TTS]fix paddle2onnx's install version, install the newest paddle2onnx in … #3084 by @yt605155624
• Update to py3.8, fix librosa==0.8.1 numpy==1.23.5 for paddleaudio. by @zh794390558
• Fix transformation import error. 【Fix Speech Issue No.5】issue 3444 transformation import error #3779 by @kk-2000
• Adapt view behavior change, fix KeyError. 【Fix Speech Issue No.3】adapt view behavior change, fix KeyError. #3794 by @zxcd
• Fix profiler, fix gpu_mem unit, add max_mem_reserved for benchmark. 【benchmark】fix new profiler #3323 【benchmark】fix gpu_mem unit #3634 【benchmark】add max_mem_reserved for benchmark  #3604 by @mmglove

Docs

• Fix some typos. [Typo] Fix some typos using Search-for-Typos #3178 by @Yulv-git
• Update svs_music_score.md. [TTS] update svs_music_score.md #3085 [TTS] get music score for svs #3070 by @lym0302
• Update quick_start.md. [Typo] Update quick_start.md #3175 [Typo] Update quick_start_cn.md #3176 by @46319943
• Add cli test readme. 【test】add cli test readme #3784 by @zxcd
• Update bug-report-tts.md. Update bug-report-tts.md #3120 by @yt605155624

Others

• Fix 0-d tensor, with the upgrade of paddlepaddle==2.5, the problem of modifying 0-d tensor has been solved. [ASR] fix asr 0-d tensor. #3214 by zxcd [s2t] old grad clip has 0d tensor problem, fix it #3334 by @zh794390558
• Add dtype param for arange API. add dtype param for arange API. #3302 by @zxcd
• Fix develop bug function:view to reshape. fix develop bug function:view to reshape #3633 by @luyao-cv
• Fix progress bar unit. [BUG] Fix progress bar unit. #3177 by @46319943
• Rm unused dep. [TTS] rm unused dep #3097 by @lym0302

Acknowledgements

Special thanks to @jiamingkong @Zth9730 @yeyupiaoling @zxcd @zh794390558 @lemondy @lym0302 @longRookie @ljhzxc @yt605155624 @USTCKAY @mattheliu @Gsonovb @twoDogy @Coloryr @kk-2000 @mmglove @Yulv-git @46319943 @luyao-cv

New Contributors

• @jiamingkong made their first contribution in Adding WavLM implementation #3242
• @yeyupiaoling made their first contribution in [ASR]add squeezeformer model #2755
• @lemondy made their first contribution in fix example/aishell local/train.sh if condition bug, test=asr #3146
• @longRookie made their first contribution in [TTS]【Hackathon + No.190】 + 模型复现：iSTFTNet #3006
• @ljhzxc made their first contribution in [TTS] [黑客松]Add JETS #3109
• @USTCKAY made their first contribution in Add XPU support for SpeedySpeech #3502
• @mattheliu made their first contribution in 【Fix Speech Issue No.8】issue 3652 merge_yi function has a bug #3786
• @Gsonovb made their first contribution in fix: 🐛 修复服务端 python ASREngine 无法使用conformer_talcs模型 #3230
• @twoDogy made their first contribution in add function for generating srt file #3123
• @Coloryr made their first contribution in 增加文件编码读取 #3606
• @kk-2000 made their first contribution in 【Fix Speech Issue No.5】issue 3444 transformation import error #3779
• @Yulv-git made their first contribution in [Typo] Fix some typos using Search-for-Typos #3178
• @46319943 made their first contribution in [Typo] Update quick_start.md #3175
• @luyao-cv made their first contribution in fix develop bug function:view to reshape #3633

---

This discussion was created from the release PaddleSpeech r1.4.2.