Skip to content
Audio

StaleyNET

текст в музыку на Rust и edge-GPU

2 min read

StaleyNET — это набор чекпоинтов text-prompted генератора музыки, обученного с нуля на Rust на NVIDIA edge-GPU. В проекте реализованы Transformer, residual vector quantizer, Griffin-Lim вокодер и GPU-ядра; опубликовано 9 запусков, включая lyrics-conditioned и pq-large.

Ключевые особенности

  • 9 чекпоинтов: от базовой text-to-music модели до vocal, lyrics и pq-large; 7,099,904–7,165,440 параметров у малых runs и 42,858,496 у pq-large.
  • Архитектура: width 256, 8 heads, 6 layers, context 768; 4 residual codebooks по 1024 записи, 128 mel bands at 24 kHz. pq-large — width 512, 8 layers, context 1152 и 16 codebooks.
  • Lyrics-модели добавляют 256-byte lyrics field; есть отдельные запуски lyrics-orin, lyrics-thor и lyrics-thor-long.
  • Обучение на Song Describer Dataset и JamendoLyrics MultiLang; лицензия чекпоинтов CC BY-SA 4.0 с оговорками upstream-датасетов.
  • Код, paper, book и crate staleynet-audio 0.1.2 доступны в companion-репозитории; релиз v0.1.2 обновляет GPU-ядра и добавляет новые runs.
  • Это research artifacts, а не продакшн-система: Griffin-Lim даёт phasey-звук, а разборчивость лирики ограничена.

Пример

hf download arpanpathak/staleynet-checkpoints --repo-type model --local-dir staleynet-checkpoints
staleynet sing --cache data/cache --run staleynet-checkpoints/checkpoints/thor-base/best --prompt "slow acoustic ballad, warm female voice" --out song.wav
ПодробнееВалидационный loss pq-large — 5.135 nats per token на шаге 7,000 из 20,000; он усреднён по 16 codebooks другого токенизатора и напрямую не сравним с 4-codebook runs.
pq-large всё ещё обучается на Jetson AGX Thor (128 GB), файлы — снапшот и будут заменяться.
Ограничения: phasey-аудио от Griffin-Lim, малые edge-модели, ограниченная разборчивость лирики, отсутствие гарантий.
Discuss Telegram 1

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 1.2 GB · ready files1.9 GB
  • FP8 / INT8recommendedweights 0.6 GB1.2 GB
  • 4-bitweights 0.4 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 · 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.

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

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

About the model

arpanpathak/staleynet-checkpoints
Author
arpanpathak
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

01

Ask AI about this model

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

Comments

Other

BIGR

BIGR