Update langchain_llm package

Author: xusenlin

libs/langchain_llm/README.md

@@ -153,4 +153,11 @@ chat_llm = ChatHuggingFace(
 
 
 + `flash_attn`: Enable FlashAttention-2.
-+ 
+
+## Merge Lora model
+
+```python
+from langchain_llm import apply_lora
+
+apply_lora("base_model_path", "lora_path", "target_model_path")
+```

libs/langchain_llm/langchain_llm/_huggingface.py

@@ -594,7 +594,7 @@ def _stream(
                     token=chunk.text, verbose=self.verbose, log_probs=logprobs
                 )
 
-    def _to_chat_prompt(self, messages: List[BaseMessage]) -> Union[List[int], Dict[str, Any]]:
+    def _to_chat_prompt(self, messages: List[BaseMessage]) -> List[Dict]:
         """Convert a list of messages into a prompt format expected by wrapped LLM."""
         if not messages:
             raise ValueError("at least one HumanMessage must be provided")

libs/langchain_llm/langchain_llm/adapters/__init__.py

@@ -1 +1,2 @@
 from .template import get_prompt_adapter
+from .model import load_model_and_tokenizer

libs/langchain_llm/langchain_llm/utils.py

@@ -3,7 +3,7 @@
 import torch
 from loguru import logger
 from peft import PeftModel
-from transformers import AutoTokenizer, AutoModelForCausalLM, PreTrainedModel
+from transformers import AutoTokenizer, AutoModelForCausalLM
 
 
 def apply_lora(
@@ -12,7 +12,7 @@ def apply_lora(
     target_model_path: str,
     max_shard_size: Optional[str] = "2GB",
     safe_serialization: Optional[bool] = True,
-) -> PreTrainedModel:
+):
 
     logger.info(f"Loading the base model from {base_model_path}")
     base = AutoModelForCausalLM.from_pretrained(