Beta

Explore
Marketplace
Neural Labs
Chat
Wallet
Docs

Beta

Explore Marketplace Neural Labs Wallet Docs

Build

Manage

My assets Manage wallet Organization

Community

Discord X (Twitter)Telegram

Docs Privacy Policy Terms and Conditions

Proud member of

Proud member of

Docs Cookie Policy

Privacy Policy Terms and Conditions

© 2026 AlphaNeural AI. All rights reserved.

hair-swap-sd-lora-stage2 – AI Model by siik | AlphaNeural AI

You can deploy this model and start earning money today!

siik/
hair-swap-sd-lora-stage2

like

0

stable-diffusion

stable-diffusion-inpainting

hair-style-transfer

runwayml/stable-diffusion-inpainting

Views

No views yet

siik/hair-swap-sd-lora-stage2

Recovered SD Inpainting LoRA bundle for the hair swap generation pipeline.

What This Is

Trained target: Stable Diffusion Inpainting UNet LoRA
Frozen during training: ControlNet, IP-Adapter, VAE, text encoder
Base model: runwayml/stable-diffusion-inpainting
ControlNet during training: lllyasviel/control_v11p_sd15_canny
IP-Adapter weight during training: ip-adapter-plus-face_sd15.bin

Training Summary

Resolution: 512
Final global step: 1000
Best step: 400
Best validation loss: 0.16283671130319513
LoRA rank: 16
LoRA alpha: 16
LoRA dropout: 0.05
Learning rate: 7e-05

Files

pytorch_lora_weights.safetensors: LoRA weights
training_summary.json: recovered training metadata
reports/: recovered markdown notes and test summaries
samples/: sample outputs, masks, and debug artifacts

Diffusers Usage

python

from diffusers import StableDiffusionControlNetInpaintPipeline

repo_id = "siik/hair-swap-sd-lora-stage2"
pipe = StableDiffusionControlNetInpaintPipeline.from_pretrained(
    "runwayml/stable-diffusion-inpainting",
    safety_checker=None,
)
pipe.load_lora_weights(repo_id, weight_name="pytorch_lora_weights.safetensors", adapter_name="hairgen_runtime")
pipe.set_adapters("hairgen_runtime", 1.0)

Notes

This repo contains the LoRA only, not the full SD checkpoint.
Downstream use should follow the license and policy constraints of the base model.