Skip to content
3D

Dirtflow v0

постройки Minecraft по тексту и окружению

1 min readupdated

Dirtflow v0 — двухстадийная flow-модель для генерации и редактирования построек Minecraft по тексту и блокам вокруг. Автор — IvanBanny; геометрическая стадия основана на microsoft/TRELLIS-text-xlarge, выход — правки в полном словаре block states Minecraft 1.21.11.

Ключевые особенности

  • Две стадии: геометрия (TRELLIS XL, 340k EMA) и материалы (400k LIVE).
  • Текст кодируется frozen CLIP ViT-L/14 с контекстом 77 токенов; учитываются наблюдаемые block states и окружение.
  • Задачи: build/replace, complete, repair и materials; граница редактирования фиксирована, блоки вне выделения восстанавливаются.
  • Выход: 29 669 легальных full block states; canvas 64×64×64, до 58 блоков на редактируемую ось с контекстом.
  • Кастомный inference engine и Minecraft mod: результат можно посмотреть до принятия.

Пример

dirtflow-engine
# скачать и проверить модель без GPU
dirtflow-engine --download-only
ПодробнееБандл: 9,3 ГБ; stage1.pt — 4,47 ГБ, stage2.pt — 4,32 ГБ, CLIP — 0,49 ГБ.
Нужен NVIDIA GPU с BF16; эталон — Python 3.12, PyTorch 2.12.0/CUDA 13.2, Transformers 5.13.0. Веса хранятся в FP32, денойзинг идёт в BF16.
Ограничения: материалы бывают хаотичными, геометрию вне кандидатов Stage 1 добавить нельзя, сущности и block-entity не генерируются; длинные промпты обрезаются до 77 токенов.
Оценка — небольшие панели, визуальный неослеплённый обзор AI-агентом; человеческие предпочтения не измерены.
Видео-пример
Discuss Telegram 1

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 4.6 GB · ready files5.6 GB
  • FP8 / INT8recommendedweights 2.3 GB3.0 GB
  • 4-bitweights 1.3 GB1.9 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.

    # IvanBanny/dirtflow-v0
    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 IvanBanny/dirtflow-v0 --local-dir ~/models/dirtflow-v0
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to dirtflow-v0
  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 TRELLIS-text-xlarge (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 4All on Hugging Face

About the model

IvanBanny/dirtflow-v0
Author
IvanBanny
Category
3D
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