Skip to content

v1.0.1 beta: Aspiration Detection, Reduced Model Size
Compare
Choose a tag to compare
@qiuqiao qiuqiao released this 12 Dec 18:35
· 78 commits to main since this release
v1.0.1
1fac4aa

Version 1.0.1 is compatible with the models from version 1.0.

New Features

  • Aspiration Detection: Introduced the ap_detector module, which automatically detects aspiration noises. The module is customizable, similar to the g2p module.
  • Reduced Disk Space Usage for Model Weights: After training completion, a model without optimization parameters is saved, significantly reducing the model's size.

Improved Features

  • Optimized Exception Prompts: Enhanced the prompts when the program throws exceptions, making it easier for users to troubleshoot errors.
  • Post-Processing: Added post-processing steps during inference to ensure that the final annotations are more in line with annotation standards.

Bug Fixes

  • Fixed many bugs.

v1.0.1 beta：AP检测，减小模型体积

v1.0.1版本兼容v1.0版本的模型。

新功能

  • AP检测： 加入ap_detector模块，自动检测呼吸音，模块可自定义，类似于g2p模块
  • 减小模型权重体积： 在训练完成后，保存一个无优化器参数的模型，大大减小了模型的体积

改进功能

  • 优化异常提示： 优化了程序抛出异常时的提示，便于用户排查错误
  • 后处理： 推理时加入后处理步骤，使得最终的标注更加符合标注规范

bug修复

  • 修复了许多bug
Assets 4
Loading