Skip to content
Audio

Whisper-small Kinyarwanda

ASR-база для стандартной речи

1 min read

Файнтюн openai/whisper-small (0,24 млрд параметров) для распознавания стандартной речи на киньяруанда. Обучен на Common Voice 27.0 и AfriVoice Kinyarwanda v0.2; это stage-1 база для будущих CDLI-моделей по нарушенной речи, сама на такой речи не обучалась.

Ключевые особенности

  • ASR для киньяруанда на базе Whisper-small, библиотека Transformers.
  • Обучение: 30 000 шагов, effective batch 64, lr 1e-5; данные Common Voice 27.0 и AfriVoice v0.2.
  • Декодирование по конвенции CDLI: language="sw", greedy num_beams=1.
  • Устойчивость к шуму: при white noise 20 dB WER 0.337 на CV и 0.105 на AfriVoice.
  • Не обучена на нарушенной речи — для таких данных нужен следующий этап CDLI.

Пример

Сценарий из карточки: подайте аудио на киньяруанда и декодируйте с language="sw" и num_beams=1. Модель вернёт транскрипцию; для нарушенной речи нужен следующий этап CDLI.

ПодробнееWER на тестах: Common Voice 27 — 0.288, AfriVoice — 0.089, CDLI impaired rw — 0.474 (у cdli-std 0.697, zero-shot 0.983).
Размер: Whisper-small, 0,24 млрд параметров; VRAM в карточке не указан.
Обучение: 30 000 шагов, effective batch 64, lr 1e-5; данные CV 27.0 + AfriVoice v0.2.
Ограничение: не обучалась на нарушенной речи; английский тест CDLI — WER 0.658.
Discuss Telegram 0 24

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # emmilly/kinyarwanda-standard-asr-whisper-small_rw_stage1-cvafv_s42
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download emmilly/kinyarwanda-standard-asr-whisper-small_rw_stage1-cvafv_s42 --local-dir ~/models/kinyarwanda-standard-asr-whisper-small_rw_stage1-cvafv_s42
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of kinyarwanda-standard-asr-whisper-small_rw_stage1-cvafv_s42 (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

emmilly/kinyarwanda-standard-asr-whisper-small_rw_stage1-cvafv_s42
Author
emmilly
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

240

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments