K2MAR/wiring-schema-lora

Wiring Schema LoRA

Fine-tuned Stable Diffusion v1.5 for generating electrical wiring schemas and circuit diagrams.

Model Details

• Base: Stable Diffusion v1.5
• Method: LoRA (Low-Rank Adaptation)
• Rank: 16 | Alpha: 32
• Training: 30 epochs on 50 wiring schemas
• Loss: 0.0998 → 0.0897 (10.1% improvement)

Quick Start

import torch
from diffusers import StableDiffusionPipeline
from peft import PeftModel

pipeline = StableDiffusionPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    torch_dtype=torch.float16
)
pipeline.unet = PeftModel.from_pretrained(pipeline.unet, "USERNAME/wiring-schema-lora")
pipeline = pipeline.to("cuda")
image = pipeline("electrical wiring diagram").images[0]
image.save("output.png")

API Usage

curl -X POST "https://api-inference.huggingface.co/models/USERNAME/wiring-schema-lora" \
  -H "Authorization: Bearer YOUR_HF_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"inputs": "electrical wiring diagram"}'

Training

• Learning Rate: 5e-5
• Batch Size: 1
• Weight Decay: 0.01
• Data Augmentation: On

License

OpenRAIL License