whisper model

#concept

whisper is trained on ___ hours of audio data >> 680k### References

1. https://github.com/openai/whisper/blob/main/model-card.md

Notes §

#todo

• skim
  • colab https://colab.research.google.com/github/openai/whisper/blob/master/notebooks/LibriSpeech.ipynb
    • know step by step how it works.
  • model card https://github.com/openai/whisper/blob/main/model-card.md
• 1-pager explaining
• brainstorm a proposal why it’s relevant to Viva
  • https://web.stanford.edu/class/cs224n/project/project-proposal-instructions-2022.pdf
• https://github.com/openai/whisper/blob/main/model-card.md

---

• how does a vanilla transformer LLM (2017) work?
  • is a transformer an encoder and decoder?
• How was the 680k audio data scraped?

Vanilla transformer 2017

About whisper