added torch type in load_model

app.py

@@ -93,19 +93,26 @@ import torch
 # Load the model and tokenizer
 def load_model():
     lora_model_name = "sreyanghosh/lora_model"  # Replace with your LoRA model path
+
+    # Try loading without 4-bit quantization
     model = AutoPeftModelForCausalLM.from_pretrained(
         lora_model_name,
-        load_in_4bit=False,
-    )
+        torch_dtype=torch.float32,  # Ensure no low-bit quantization
+        device_map="auto" if torch.cuda.is_available() else None,  # Use standard device mapping
+        load_in_4bit=False,  # Redundant, but safe to explicitly specify
+        )
+
     tokenizer = AutoTokenizer.from_pretrained(lora_model_name)
+
     if tokenizer.pad_token_id is None:
         tokenizer.pad_token_id = tokenizer.eos_token_id
+
     model.eval()
     device = "cuda" if torch.cuda.is_available() else "cpu"
     model = model.to(device)
+
     return tokenizer, model
 
-tokenizer, model = load_model()
 
 # Define the respond function
 def respond(