Updating LoRA Weights VLLM

[BY571]

Hello!

When fine-tuning my unsloth model using fast_inference=True. How can I update my LoRA weights manually, or are those weights getting updated automatically for the vllm_engine of my model?

I found you have the function load_lora. Am I supposed to update the weights like this:

adapter_path = "current_saved_adapter"
lora_request = load_lora(self.policy, adapter_path)
self.policy.vllm_engine.vllm_lora_request = lora_request