Fix: Add peft to requirements.txt for LoRA adapter support

PEFT (Parameter-Efficient Fine-Tuning) is required for loading LoRA
adapters with pipe.load_lora_weights(). Without it, LoRA loading fails
with: 'PEFT backend is required for this method.'

requirements.txt

@@ -16,6 +16,7 @@ ftfy>=6.1.0
 Pillow>=10.0.0
 safetensors>=0.4.0
 huggingface-hub>=0.19.0
+peft>=0.7.0  # Required for LoRA adapter loading
 
 # Audio Dependencies (Optional - for TTS and music generation)
 scipy>=1.11.0