You need to agree to share your Hugging Face information to access this model
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
To get access, please fill in the short form at https://saidzoda.com/access. We approve requests after we receive the form.
Log in or Sign Up to review the conditions and access this model content.
sl-vireo
How to get access: please fill in the short form at saidzoda.com/access. We approve requests after we receive the form.
Whisper-turbo fine-tuned on Khazina real-audio Tajik data (checkpoint-28000, best of 17 scored checkpoints by WER).
Result
Negative result: WER 9.91% vs base Whisper-turbo-CA 9.85% on the frozen judge set. This checkpoint did NOT improve on the base model. Published as a negative result per lab policy (failures are first-class, recorded for the negative-results paper).
Sixth of six Khazina-derived training families tested; none of the six beat base.
Contents
Deployable inference weights only (model.safetensors + config/tokenizer/preprocessor files). Optimizer state excluded (training-resume only, not needed for inference).
- Downloads last month
- 3