[Core] Support dynamically loading Lora adapter from HuggingFace (#6234)

Co-authored-by: Antoni Baum <antoni.baum@protonmail.com>

vllm/transformers_utils/tokenizer.py

@@ -137,14 +137,13 @@ def get_lora_tokenizer(lora_request: LoRARequest, *args,
     if lora_request is None:
         return None
     try:
-        tokenizer = get_tokenizer(lora_request.lora_local_path, *args,
-                                  **kwargs)
+        tokenizer = get_tokenizer(lora_request.lora_path, *args, **kwargs)
     except OSError as e:
         # No tokenizer was found in the LoRA folder,
         # use base model tokenizer
         logger.warning(
             "No tokenizer found in %s, using base model tokenizer instead. "
-            "(Exception: %s)", lora_request.lora_local_path, e)
+            "(Exception: %s)", lora_request.lora_path, e)
         tokenizer = None
     return tokenizer