| title: Dual Inv Edit | |
| emoji: 🚀 | |
| colorFrom: purple | |
| colorTo: blue | |
| sdk: gradio | |
| app_file: "" | |
| pinned: false | |
| # Dual Inv Edit | |
| This repository contains the trained weights for the Dual Inversion Editing project. | |
| ## Model Description | |
| - **transformer_fwd**: Forward transformer LoRA weights (pytorch_lora_weights.safetensors) | |
| - **transformer_bwd**: Backward transformer LoRA weights (pytorch_lora_weights.safetensors) | |
| ## Training Configuration | |
| - Learning rate: 1e-5 | |
| - Batch size: 80 | |
| - CFG scale: 5.0 | |
| - Cycle forward: 1.0 | |
| - Cycle backward: 1.0 | |
| - Lambda regularization: 0.0 | |
| - Eta: 0.9 | |
| - K: 8 | |
| ## Usage | |
| ```python | |
| from huggingface_hub import hf_hub_download | |
| # Download forward LoRA weights | |
| lora_fwd_path = hf_hub_download(repo_id="Thrcle/Dual_Inv_Edit", filename="transformer_fwd/pytorch_lora_weights.safetensors") | |
| # Download backward LoRA weights | |
| lora_bwd_path = hf_hub_download(repo_id="Thrcle/Dual_Inv_Edit", filename="transformer_bwd/pytorch_lora_weights.safetensors") | |
| ``` | |