Skip to content
Other

yue2_1980s_two_tone_ska

LoRA для two-tone ска

2 min readupdated

Это NAR-адаптер (acoustic-rendering) для YuE2 от jamespetts: он смещает рендерер к британскому two-tone ска начала 1980-х — offbeat guitar skank, organ shuffle, walking bass, brass stabs, сухие барабаны и slapback echo. Меняет только рендеринг: тембр, текстуру и продакшн-ощущение, но не сочинение, мелодию или аранжировку — это AR-ветка.

Ключевые особенности

  • Триггер twotone нужно ставить первым в style prompt; без него дельты не управляются conditioning.
  • Два формата весов: engine «unfused» для audio.cpp и ComfyUI fused для LoraLoaderModelOnly.
  • Меняет только NAR-рендеринг; композиция, гармония и ритмический план остаются за AR-веткой и промптом.
  • Обучение: 19 дублей / ~54 мин собственного AI-ска, rank/alpha 16, 2000 шагов, 27:02 на RTX 2080 Ti 11 GB.
  • На слух все четыре ступени теста звучали одинаково: при scale 1.0 вклад адаптера — на уровне текстуры. Для усиления предлагают nar_lora_scale 1.5–2.0, но это не тестировалось.

Пример

generate.ps1 -LyricsFile lyrics.txt -Style "twotone, English, early 1980s British two-tone ska, ..." `
  -Out song.wav -Cot full -MaxTokens 4800 -Seed 256198995 `
  -SemanticTemp 1.1 -AbcTemp 0.9 -NarSteps 12 `
  -NarLora models\lora\yue2_1980s_two_tone_ska_unfused.safetensors -NarScale 1.0
ПодробнееОценка: расхождение с финалом 17.4% энергии сигнала (corr 0.985) на шаге 600 → 9.4% на 1200 → 3.0% на 1800; на слух все четыре варианта одинаковы.
Слияние весов при старте сессии ~55 с (196 тензоров), это не зависание.
Обучение — 27:02 на RTX 2080 Ti 11 GB; веса CC0 1.0, база YuE2 — CC BY-NC 4.0 (только некоммерческое использование).
Ограничения: корпус — собственный AI-выход, не канонический звук 2-Tone; leak-тест не проводился; адаптер только NAR.
Discuss Telegram 1 125

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 8.7 GB · ready files10 GB
  • FP8 / INT8recommendedweights 4.4 GB5.3 GB
  • 4-bitweights 2.5 GB3.3 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 2080 Ti

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

This is a LoRA: counted together with its base model YuE2.

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.

    # jamespetts/yue2-1980s-two-tone-ska-lora
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    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 Comfy-Org/YuE2 --local-dir ~/models/YuE2
    hf download jamespetts/yue2-1980s-two-tone-ska-lora --local-dir ~/models/yue2-1980s-two-tone-ska-lora
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to YuE2
  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

Requires the base model

Comfy-Org/YuE2

This model is a LoRA / adapter. Download the base model first, then load this one on top.

  1. YuE2-3Bfamily root
  2. YuE2
  3. yue2-1980s-two-tone-ska-lora

New posts about versions of YuE2-3B (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

jamespetts/yue2-1980s-two-tone-ska-lora
Author
jamespetts
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

1251

Ask AI about this model

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

Comments