WoRA integration into PEFT#2872

Open

sambhavnoobcoder wants to merge 12 commits intohuggingface:mainfrom

sambhavnoobcoder:feature/wora-integration

Commits on Oct 26, 2025

Add WoRA core implementation with weighted direction and learnable alpha/beta parameters
sambhavnoobcoder
committed
Update LoraLayer to support WoRA with alpha/beta parameter storage
sambhavnoobcoder
committed
Update Linear, Embedding, and Conv layer classes to support WoRA variants
sambhavnoobcoder
committed
Add use_wora parameter propagation in LoraModel layer creation
sambhavnoobcoder
committed
Fix parameter ordering in update_layer methods for Python syntax compliance
sambhavnoobcoder
committed
Fix WoRA parameter registration and config validation for complete gradient flow
sambhavnoobcoder
committed
Enable gradient flow through WoRA alpha and beta parameters
sambhavnoobcoder
committed
Set requires_grad=True for WoRA alpha and beta parameters
sambhavnoobcoder
committed
Move wora_alpha and wora_beta to adapter_layer_names for trainability
sambhavnoobcoder
committed
Rename wora_alpha/wora_beta to lora_wora_alpha/lora_wora_beta for prefix matching
sambhavnoobcoder
committed
Add WoRA tests to test_lora_variants.py
sambhavnoobcoder
committed
Add WoRA tests and fix gradient flow issues
sambhavnoobcoder
committed