Added a bunch of improvements for the fine-tune gui and fixed a bug when running on only cpu

[DrewThomasson]

Pull Request Details

• Fixed Whisper CPU bug:
  Now switches to int8 on CPU to avoid float16 issues. GPU usage remains float16.

• Updated README:
  Added usage guide and pre-made Docker images to pull for x86 CPU, NVIDIA GPU, and Apple Silicon.

• Apple Silicon Instructions:
  Included specific install steps with a dedicated apple_silicon_requirements.txt.

• New Export Model Button:
  Exports a folder with:

  • dataset.zip
  • model.pth
  • config.json
  • vocab.json

  Removes optimizer state for smaller model size. :)

  PS: Thank you so much for keeping Coqui-TTS running smoothly! 🙏