1 Release 1 Tag

RSS Feed

• 1.0 496efa87bc

  Compare

  Agnlash 1.0: Automatic speech recognition for Uyghur Stable

  alikia2x released this 2024-10-21 16:18:12 +00:00 | 0 commits to master since this release

  Based on uyghur-asr-ctc, we fine-tuned the CommonVoice dataset for 6 epochs and achieved some improvement:
  Loss decreased by 0.03, WER decreased by 1.5%, and CER decreased by 0.3%.

  Downloads

  • Source Code (ZIP)
  • Source Code (TAR.GZ)
  • UModel_last.pth
    101 MiB
  • UModel.log
    4.4 KiB