You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Qwen-Image-Edit-2511 LoRA (v4, 1000 steps)

Style LoRA for Qwen/Qwen-Image-Edit-2511: selfie → pop-character profile picture (flat grey background).

This is checkpoint-1-1000 from the v4 run (2,011 QA-valid touching_bottom pairs). Rank 16, lora_alpha 16, targets to_k / to_q / to_v / to_out.0.

The later 4500-step overshoot checkpoint is not in this repo (that file was not retained). Quality on this recipe was already largely flat after ~2000 steps; 1000 is an earlier, still-usable snapshot.

Files

pytorch_lora_weights.safetensors — standard PEFT/diffusers filename
checkpoint-1-1000_pytorch_lora_weights.safetensors — same weights, original training name
avatar_lora_v4_qa2011_touchingbottom.yaml — training config

Load

# PEFT / diffusers-style: point pretrained_weight / load_lora_adapter at
# pytorch_lora_weights.safetensors in this repo.

Downloads last month: -

Model tree for ValocheRoomates/QwenLora1000FineTune

Base model

Qwen/Qwen-Image-Edit-2511

Adapter

(169)

this model