Skip to content

Conversation

HighCWu
Copy link
Contributor

@HighCWu HighCWu commented Feb 9, 2023

PR types

Performance optimization

PR changes

Models

Describe

When I use pndm scheduler 25 steps to sample, I found that the generated results are very poor. I found that the default ddpm scheduler sampling comes with noise clip, but pndm scheduler does not. Add the clip_noise option to the diffusion module, users will be able to use faster sampling methods like pndm to generate much better sampling results than when there is no noise clip.

当我使用pndm scheduler 25步进行采样时,发现生成结果效果很差,我发现默认的ddpm scheduler采样自带noise clip,而pndm scheduler没有。往diffusion模块增加clip_noise选项,用户将可以采用像pndm这样更快的采样方式生成相比没有noise clip时好很多的采样结果

@mergify mergify bot added the T2S label Feb 9, 2023
@yt605155624 yt605155624 requested a review from lym0302 February 9, 2023 04:46
@yt605155624 yt605155624 added this to the r1.4.0 milestone Feb 9, 2023
@yt605155624 yt605155624 mentioned this pull request Feb 9, 2023
@yt605155624 yt605155624 changed the title [TTS] Add diffusion noise clip to optimize sample result [TTS]Add diffusion noise clip to optimize sample result Feb 9, 2023
@lym0302 lym0302 merged commit bcd8e30 into PaddlePaddle:develop Feb 9, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
Status: Done
Development

Successfully merging this pull request may close these issues.

3 participants