Skip to content
Other

ROSETok

устойчивый речевой токенизатор

2 min read

ROSETok — речевой токенизатор от wxzyd123, представленный как работа NeurIPS 2026. Он сочетает шумное обучение и семантическую дистилляцию, чтобы сохранять реконструкцию речи и семантические представления при шуме. В основе — DAC-энкодер и декодер с residual vector quantization; вход — mono 16 kHz, обучение на английском и китайском.

Ключевые особенности

  • ROSETok-Base обучен на ~96.7k часов английской и китайской речи из Emilia с Gaussian-noise augmentation и семантической дистилляцией.
  • ROSETok-AAF — attack-aware fine-tuning на 2 000 фразах LibriSpeech, perturbed with T-SemAttack.
  • Semantic teacher — Whisper large-v3; модель сохраняет и реконструкцию, и семантические представления.
  • 9 residual codebooks по 1024 entries, codebook dimension 8; hop size 320 samples, или 50 frames per second; feature dimension 1280.
  • На выходе — reconstructed audio, discrete codes, encoded features, quantized features и другие поля.

Пример

git clone https://github.com/wxzyd123/T-SemAttack.git
cd T-SemAttack
conda create -n rosetok python=3.10 -y
conda activate rosetok
python -m pip install -r requirements.txt
hf download wxzyd123/ROSETok rosetok_base.pth rosetok_aaf.pth --local-dir checkpoints/rosetok
ПодробнееЧекпоинты ~2.86 GB: внутри также semantic teacher, для инференса используются только encoder.*, quantizer.*, decoder.*.
Обучение: ~96.7k часов английской и китайской речи из Emilia; AAF — 2 000 фраз LibriSpeech с T-SemAttack.
Ограничение: устойчивость проверена только на шумах, атаках, датасетах и моделях из статьи; другие языки и условия требуют отдельной оценки.
Лицензия MIT.
Discuss Telegram 2

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 2.7 GB · ready files3.4 GB
  • FP8 / INT8recommendedweights 1.3 GB2.0 GB
  • 4-bitweights 0.7 GB1.3 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 · BF16 — 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.

    # wxzyd123/ROSETok
    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 wxzyd123/ROSETok --local-dir ~/models/ROSETok
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to ROSETok
  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 ROSETok (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

wxzyd123/ROSETok
Author
wxzyd123
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

02

Ask AI about this model

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

Comments

Other

USFM

USFM