Add zero-gpu support, enhance model export with quantization/gpu acceleration helpers, optimize inference pipeline with vae fixes, modernize pipeline loading with unified decorators, implement gpu decorator infrastructure.

cleaned up code

Remove Python cache files and update .gitignore

Add cancel button, validate cache files, detect cache hits, update peft dependency, fix race conditions, enhance feedback, improve interrupt handling.

updated

Update Gradio to 6.9.0, add cached file checks in LoRA loading, fix unloading errors, optimize download cancellation, improve device/dtype handling, update optimum dependency.

code