fix: fix local loading for .bin models (#1419)

server/text_generation_server/cli.py

@@ -198,7 +198,7 @@ def download_weights(
             if not extension == ".safetensors" or not auto_convert:
                 raise e
 
-    else:
+    elif (Path(model_id) / "adapter_config.json").exists():
         # Try to load as a local PEFT model
         try:
             utils.download_and_unload_peft(

server/text_generation_server/utils/peft.py

@@ -10,8 +10,7 @@ from peft import AutoPeftModelForCausalLM, AutoPeftModelForSeq2SeqLM
 def download_and_unload_peft(model_id, revision, trust_remote_code):
     torch_dtype = torch.float16
 
-    logger.info("Peft model detected.")
-    logger.info("Loading the model it might take a while without feedback")
+    logger.info("Trying to load a Peft model. It might take a while without feedback")
     try:
         model = AutoPeftModelForCausalLM.from_pretrained(
             model_id,
@@ -28,7 +27,7 @@ def download_and_unload_peft(model_id, revision, trust_remote_code):
             trust_remote_code=trust_remote_code,
             low_cpu_mem_usage=True,
         )
-    logger.info(f"Loaded.")
+    logger.info("Peft model detected.")
     logger.info(f"Merging the lora weights.")
 
     base_model_id = model.peft_config["default"].base_model_name_or_path