danielrosehill's picture
Deploy from ASR-Training
5f3b21c verified
|
Raw History Blame Contribute Delete
2.41 kB
metadata
title: Parakeet Training Data Map
emoji: 🗺️
colorFrom: blue
colorTo: yellow
sdk: static
pinned: false
license: cc-by-4.0
short_description: Master dataset, modular subsets, fine-tune blends
datasets:
  - danielrosehill/Read-Aloud-Voice-Sentences
  - danielrosehill/English-Hebrew-Codeswitching-Sentences
  - danielrosehill/Tech-Sentences
  - danielrosehill/AI-Interactions-Sentences
  - danielrosehill/Linux-Admin-Sentences
  - danielrosehill/Smart-Home-Sentences
  - danielrosehill/Personal-Vocabulary-Sentences
  - danielrosehill/Homophone-Context-Sentences
  - Hebrish/Yiddish-Sentences
  - Hebrish/Jewish-Religious-Terminology-Sentences
  - danielrosehill/General-English-Sentences
  - Hebrish/Noisy-Condition-Sentences
  - Hebrish/Disfluent-Read-Sentences
  - Hebrish/Self-Correction-Sentences
  - Hebrish/Spoken-Aside-Sentences
  - Hebrish/Hard-ASR-Sentences
  - Hebrish/Brand-Name-Sentences
  - Hebrish/Acronym-Sentences
  - Hebrish/Irish-English-Sentences
  - Hebrish/Repeated-Word-Sentences
  - Hebrish/Singing-Sentences
  - Hebrish/Weird-Voice-Sentences
  - Hebrish/Currency-Sentences
  - Hebrish/Expanded-Acronym-Sentences
  - Hebrish/AI-Tech-Vocabulary-Sentence-Bank
  - Hebrish/AI-Interactions-Sentence-Bank
  - Hebrish/Linux-Admin-Sentence-Bank
  - Hebrish/Smart-Home-Sentence-Bank
  - Hebrish/Personal-Vocabulary-Sentence-Bank
  - Hebrish/Homophone-Context-Sentence-Bank

Parakeet Training Data Map

How the training data for the English-Hebrew code-switching fine-tune of nvidia/parakeet-tdt-0.6b-v2 is organised: one master dataset (danielrosehill/Read-Aloud-Voice-Sentences) where every take lands, modular datasets copied out of it (domain subsets and one task dataset per speech phenomenon), the text-only sentence banks, and which datasets each fine-tune blends.

A single static page, generated by scripts/build_datasets_space.py in the project repo from the same registries the sync and training code use, then deployed with scripts/deploy_space.py datasets-map. Do not edit it here; regenerate it.

All datasets: training-data collection. Project page: Hebrish/English-Hebrew-Codeswitching-Parakeet.