PathoGen Model Weights

This repository contains the model weights for PathoGen, a diffusion-based histopathology inpainting model.

Model Components

🧠 UNet Weights

• File: diffusion_pytorch_model.bin (3.3GB)
• Description: Main diffusion model weights for the UNet architecture
• Usage: Loaded automatically by the PathoGen pipeline

🎯 Attention Weights

• File: attention.pt (190MB)
• Description: Fine-tuned attention module weights for histopathology-specific features
• Usage: Custom attention processors for enhanced feature transfer

Usage

This model is designed to be used with the PathoGen Space:

• Space: https://huggingface.co/spaces/mkoohim/PathoGen
• Code: Available in the Space repository

Loading the Model

from diffusers import UNet2DConditionModel
import torch

# Load UNet
unet = UNet2DConditionModel.from_pretrained("./", subfolder="unet")

# Load attention weights
attention_weights = torch.load("attention.pt", map_location="cpu")

Model Details

• Architecture: Diffusion-based inpainting model
• Input Size: 512x512 pixels
• Inference Steps: 10 (optimized for speed)
• Framework: PyTorch + Diffusers
• Application: Histopathology image feature transfer

License

MIT License - See LICENSE file for details.