Skip to content
Images

ZYI-256 Flash

343M модель для картинок и правок по тексту

2 min read

ZYI-256 Flash от caikybaldo999 — компактная text-to-image модель на 343M параметров, которая генерирует изображения 256×256 и умеет редактировать их по инструкции на английском. Архитектура — DiT с rectified flow, латентное пространство DC-AE, текстовый энкодер flan-T5-base.

Ключевые особенности

  • 343M параметров, DiT + rectified flow; скрытый размер 1024, глубина 20, 16 голов.
  • 256×256, DC-AE f32c32: всего 64 латентных токена; текст — flan-T5-base, 96 токенов.
  • Режим editing: исходник кодируется DC-AE, его 64 токена идут в одной последовательности с целевыми; добавлен один обучаемый source-embedding.
  • Обучение с нуля на 1 226 758 изображениях FLUX-Reason-6M; для editing — 190 988 пар из OmniEdit-Filtered-1.2M.
  • Лучше всего удаются пейзажи, интерьеры, еда, объекты и живописные стили.

Пример

from PIL import Image
pipe = zyi256.ZyiPipeline("caikybaldo999/zyi-256-flash")
pipe("a red fox sitting in a snowy forest", n=4, seed=0)
pipe.edit(Image.open("photo.jpg"), "make it winter, add snow")[0]

Английские промпты работают лучше: подписи в обучении на английском.

ПодробнееБаза: 9 544 шага при batch size 1024; файнтюн: ещё 10 233 шага, всего 42 981 шаг.
Железо: 1× NVIDIA RTX PRO 6000 (96 GB), около 5 ч на базу и 5,7 ч на файнтюн; editing-апдейт — 6,5 ч.
Ограничения: слабо с анатомией животных, лицами, руками и читаемым текстом; пространственные отношения соблюдаются не всегда.
Editing: +18 613 шагов, 60% editing / 40% text-to-image; низкое разрешение мешает мелким правкам.
Примеры
Discuss Telegram 3 358

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 1.3 GB · ready files2.9 GB
  • FP8 / INT8recommendedweights 0.6 GB2.2 GB
  • 4-bitweights 0.4 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.

    # caikybaldo999/zyi-256-flash
    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 caikybaldo999/zyi-256-flash --local-dir ~/models/zyi-256-flash
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to zyi-256-flash
  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.

  • SageAttention 3v2.2.0from October 28, 2025

    even faster on RTX 50 / Blackwell

    FP4 attention for Blackwell. Faster than SageAttention 2, but the authors suggest v2 when accuracy matters.

    How to enable: Build from the sageattention3_blackwell folder of the repo; in ComfyUI via KJNodes.

    Earlier generations (2)
    • SageAttention 2++ — Quantised (INT8/FP8) attention with no visible quality loss. The main speed-up for image and video generation.
    • SageAttention 1 — First version: INT8 for Q·K. Still works, but 2++ is faster and more accurate.
  • 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 (3)
    • 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.
    • xFormers — The older way to speed up attention, works almost everywhere. Sage or Flash are faster where available.
  • cache-dit (DBCache)v1.5.6from November 20, 2026new release

    1.5–3× at nearly the same quality

    Reuses block outputs between diffusion steps. Supports most DiT models.

    How to enable: pip install cache-dit; one line cache_dit.enable_cache(pipe) for diffusers.

    Earlier generations (2)
    • MagCache — Skips steps by magnitude of change. More accurate than TeaCache, no calibration.
    • TeaCache — The first popular diffusion step cache. Works, newer methods are more accurate.
  • Nunchaku (SVDQuant 4-bit)v1.2.1from January 25, 2026

    ≈3× faster and 3.5× less memory

    4-bit versions of popular image models with almost no quality loss (FP4 on RTX 50).

    How to enable: Ready weights for supported models; in ComfyUI use ComfyUI-nunchaku.

  • 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 zyi-256-flash (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 2All on Hugging Face

About the model

caikybaldo999/zyi-256-flash
Author
caikybaldo999
Category
Images
Released
Posted

On NeyroLearn

Hugging Face

3583

Ask AI about this model

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

Comments

Images

Aurora

Aurora