Skip to content
Other

Hy-Nomas

интерактивная генерация миров по тексту и камере

1 min read

Tencent представила Hy-Nomas — видеомодель для свободной многошаговой генерации миров. Она строит видео по сегментам: можно добавлять объекты, менять стиль и сцены, описывать действия и управлять камерой. В основе — посегментный авторегрессионный DiT на 54 dual-stream MMDiT-блока (hidden size 2048) с текстовым энкодером Qwen2.5-VL-7B, камерными модулями, декодером PixelUMM и постоянным KV cache.

Ключевые особенности

  • Свободное взаимодействие: инструкции по субъектам, действиям, внешнему виду, стилю и окружению.
  • Управление камерой: текстовые команды вместе с траекториями движения и точкой обзора.
  • Непрерывная генерация: контекст предыдущих сегментов переносится через KV cache.
  • Дистиллированный инференс: 8 шагов денойзинга DiT и отдельное декодирование PixelUMM.
  • Гибкое развёртывание: стандартная sequence-parallel конфигурация или одна RTX 4090.
  • Hy-Nomas-Bench: 4 бенчмарка и 280 тестов для интерактивной управляемости и консистентности.

Пример

Быстрый запуск через рекомендованный Docker-образ:

git clone https://github.com/Tencent-Hunyuan/Hy-Nomas.git
cd Hy-Nomas
export HY_NOMAS_IMAGE="hynomasoss/hynomasoss:cu130"
docker pull "$HY_NOMAS_IMAGE"
docker run --rm -it --gpus all --ipc=host --network=host \
  -v "$PWD:/workspace/Hy-Nomas" -w /workspace/Hy-Nomas \
  "$HY_NOMAS_IMAGE" bash
ПодробнееСтандартный inference: 8 шагов, 720×1280 при 24 FPS.
Конфигурация для одной RTX 4090: INT8 DiT, INT8 KV cache, FP8 декодер и CPU offload.
На одной GPU история ограничена 18 латентными кадрами; стандартная конфигурация хранит всю историю.
INT8-загрузчик не поддерживает отдельный LoRA-адаптер при инференсе.
Discuss Telegram 132 63K

Will it run?

Only with RAM offload

Even 4-bit is larger than your VRAM. Some layers can stay in system RAM (llama.cpp, ComfyUI), but it will be several times slower.

  • BF16weights 81 GB · ready files90 GB
  • FP8 / INT8recommendedweights 41 GB45 GB
  • 4-bitweights 23 GB26 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: Won't fit any home GPU — renting is cheaper.

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.

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

Quantised versions 3All on Hugging Face
Fine-tunes 2All on Hugging Face

About the model

tencent/Hy-Nomas
Author
tencent
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

63K132

Ask AI about this model

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

Comments

Other

NanoKV

NanoKV
Other

PrimeSim

PrimeSim