Skip to content
Audio

lapwing

оценка произношения для изучающих английский

2 min readupdated

GersonHMG/lapwing — ONNX-модель для оценки произношения английского у изучающих язык, построенная на nvidia/parakeet-tdt-0.6b-v2. В ней 628M параметров: энкодер заморожен, добавлена phone-голова. Один проход по записи 16 кГц моно даёт phone posteriors и транскрипт.

Ключевые особенности

  • Phone posteriors log_probs [T, 40] на 20-мс кадрах — для goodness-of-pronunciation (GOP) scoring.
  • Один проход энкодера выдаёт и phone posteriors, и encoder outputs для штатных TDT decoder/joiner; транскрипт нужен только для проверки, что прочитано нужное предложение.
  • Runtime 685 MB (uint8 encoder + int8 decoder/joiner); 0.24 с на 5-секундный клип при 1 CPU thread; фиксированного лимита входа нет, проверяли до 20 с.
  • Гейт отсекает попытку, если CTC log-likelihood целевых phones на кадр L(c)/T ниже −0.489; порог подобран так, чтобы принимать 2.3% попыток с неверным предложением.
  • Лицензия CC BY-NC 4.0 — некоммерческая; phone-голова обучена на L2-ARCTIC. GOP-пороги цветов не калиброваны.

Пример

Сценарий из карточки: ученик читает предложение вслух. Модель возвращает phone posteriors для GOP-оценки и транскрипт для проверки, что прочитано именно нужное предложение; при низком CTC log-likelihood попытка отсекается гейтом.

ПодробнееPhone error rate: 16.6% на L2-ARCTIC B и 15.4% в целом; F1 детекции ошибок произношения 0.541; GOP phone AUC 0.854.
Запуск: onnxruntime ≥1.20, CPU; 685 MB, 0.24 с на 5-секундный клип при 1 потоке.
Ограничения: только английский и чтение вслух; слабее на спонтанной речи, других акцентах, шуме и дальнем микрофоне.
Лицензия CC BY-NC 4.0 — некоммерческая; phone-голова обучена на L2-ARCTIC.
Discuss Telegram 0

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 2.3 GB · ready files3.1 GB
  • BF16weights 1.2 GB1.8 GB
  • FP8 / INT8recommendedweights 0.6 GB1.1 GB
  • 4-bitweights 0.3 GB0.9 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F32 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

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.

    # GersonHMG/lapwing
    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 GersonHMG/lapwing --local-dir ~/models/lapwing
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to lapwing
  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 parakeet-tdt-0.6b-v2 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 6All on Hugging Face
Merges 1All on Hugging Face

About the model

GersonHMG/lapwing
Author
GersonHMG
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments