Skip to content
Images

data266-pokemon-lora

генерация существ в стиле покемонов для SD 1.5

2 min readupdated

Это LoRA (небольшая надстройка над базовой моделью), которая учит Stable Diffusion 1.5 рисовать вымышленных существ в узнаваемой стилистике покемонов. Адаптер создан автором dhruv1672 на базе 833 иллюстраций из датасета pokemon-blip-captions и применяется к слоям внимания UNet. В отличие от базовой модели, которая смешивает разные художественные манеры, эта версия уверенно переносит характерную рисовку и композицию с одним персонажем на белом фоне. Пригодится для генерации концептов фэнтези-существ, стилизованных иллюстраций или экспериментов с настройкой рангов LoRA.

Ключевые особенности

  • Версия v2.0 прошла серию тестов по разным рангам (от 4 до 128), лучшим выбран вариант lora_r16_a16 с EMA-весами.
  • Снижает метрику KID (отличие от целевого стиля) на тестовом наборе с 0.06483 у базовой модели до 0.02923.
  • В репозитории сохранены все промежуточные прогоны экспериментов — их можно загружать по отдельности через subfolder.
  • Поддерживает стандартный пайплайн diffusers и работает в разрешении 512×512 пикселей.

Что можно сделать

  • Придумывать новых существ в духе покемонов по текстовым описаниям вроде «сине-жёлтый покемон».
  • Сравнить работу адаптеров с разным рангом (например, компактный lora_r4_a4 против финального lora_r16_a16) в одном проекте через diffusers.
  • Использовать как отправную точку для генерации спрайтов и концепт-артов одиночных персонажей на чистом светлом фоне.

Пример

from diffusers import StableDiffusionPipeline
import torch
pipe = StableDiffusionPipeline.from_pretrained("stable-diffusion-v1-5/stable-diffusion-v1-5", torch_dtype=torch.float16).to("cuda")
pipe.load_lora_weights("dhruv1672/data266-pokemon-lora", revision="v2.0")
image = pipe("a drawing of a blue and yellow pokemon", num_inference_steps=30, guidance_scale=7.5).images[0]
ПодробнееТест (20 промптов, 30 шагов DDIM): KID — 0.02923 (база — 0.06483), CLIP score — 31.865, LPIPS diversity — 0.782.
Обучение: 4000 шагов на 685 картинках (512 px, AdamW, lr 0.0001, EMA 0.999, min-SNR 5.0); пиковое потребление при обучении на A100 составило ~6.93 GiB.
Ограничения: короткие машинные подписи хуже дают точный контроль над деталями, а сложные сцены с несколькими героями тяготеют к одному существу на белом фоне. Модель предназначена только для некоммерческого использования.
Discuss Telegram 0 1

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 3.2 GB · ready files5.1 GB
  • BF16weights 1.6 GB3.3 GB
  • FP8 / INT8recommendedweights 0.8 GB2.4 GB
  • 4-bitweights 0.5 GB2.0 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 · F32 — from RTX 2060

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 stable-diffusion-v1-5.

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.

    # dhruv1672/data266-pokemon-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 stable-diffusion-v1-5/stable-diffusion-v1-5 --local-dir ~/models/stable-diffusion-v1-5
    hf download dhruv1672/data266-pokemon-lora --local-dir ~/models/data266-pokemon-lora
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to stable-diffusion-v1-5
  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

Requires the base model

stable-diffusion-v1-5/stable-diffusion-v1-5

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

New posts about versions of stable-diffusion-v1-5 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

dhruv1672/data266-pokemon-lora
Author
dhruv1672
Category
Images
Released
Posted

On NeyroLearn

Hugging Face

10

Ask AI about this model

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

Comments