Skip to content
Images

PokemonFoilMask-LoRA-Qwen-Image-2.1

маска фольги для карт без ручной покраски

2 min read

LoRA для Qwen-Image-2.1 (7B DiT, image editing), которая по карте предсказывает grayscale-карту фольги: где голограмма блестит и насколько сильно. Также делает embossed etch-текстуру для full art и secret rare. Подходит для игр/приложений, 3D-вьюера карт, печатных макетов и сайтов коллекций.

Ключевые особенности

  • Одна LoRA для holo, reverse holo, full art, stamped и etch-масок.
  • На вход — карта 800×1024, на выходе — grayscale-маска 800×1024, выровненная пиксель в пиксель.
  • При фиксированном seed одна и та же карта даёт одинаковую маску.
  • В репозитории есть WebGL-шейдер с наклоном за курсором и Python-рендерер для анимаций.
  • Обучение: 15 945 пар card-to-mask из 14 340 карт; 820 карт полностью отложены для бенчмарка.

Пример

from PIL import Image
from diffusers import QwenImage21Pipeline
import torch
pipe = QwenImage21Pipeline.from_pretrained('Qwen/Qwen-Image-2.1', torch_dtype=torch.bfloat16)
pipe.load_lora_weights('newHashMap/PokemonFoilMask-LoRA-Qwen-Image-2.1', weight_name='pytorch_lora_weights.safetensors')
pipe.enable_model_cpu_offload()
card = Image.open('card.png').convert('RGB').resize((800, 1024))
mask = pipe(prompt='Draw the holo foil mask of this Pokemon card. Holo: only the artwork window shines. Inside the artwork window, paint the foil in grayscale, white where it shines. Paint everything outside the artwork window solid black: the border, name bar, attacks and text.', image=card, width=800, height=1024, output_resolution=905, num_inference_steps=28, generator=torch.Generator('cuda').manual_seed(0)).images[0].convert('L')
ПодробнееБенчмарк на 820 held-out картах (929 foil-версий): overall 41.1; holo 55.9; reverse holo 24.4; full art 49.5.
База + LoRA — около 33 GB в bf16; с enable_model_cpu_offload() хватает 24 GB GPU.
Ограничения: reverse holo заметно слабее, обучение только на английских сетах. Лицензия Qwen Research — non-commercial research/evaluation.
Примеры и видео
Discuss Telegram 10 71

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 14 GB · ready files16 GB
  • FP8 / INT8recommendedweights 6.9 GB9.1 GB
  • 4-bitweights 4.0 GB5.9 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 3080 · BF16 — from RTX 4090

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 Qwen-Image-2.1.

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.

    # newHashMap/PokemonFoilMask-LoRA-Qwen-Image-2.1
    # ===== 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 Qwen/Qwen-Image-2.1 --local-dir ~/models/Qwen-Image-2.1
    hf download newHashMap/PokemonFoilMask-LoRA-Qwen-Image-2.1 --local-dir ~/models/PokemonFoilMask-LoRA-Qwen-Image-2.1
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Qwen-Image-2.1
  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

Qwen/Qwen-Image-2.1

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

New posts about versions of Qwen-Image-2.1 (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 4All on Hugging Face

About the model

newHashMap/PokemonFoilMask-LoRA-Qwen-Image-2.1
Author
newHashMap
Category
Images
Released
Posted

On NeyroLearn

Hugging Face

7110

Ask AI about this model

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

Comments