a

app.py

@@ -9,13 +9,18 @@ from transformers import AutoModel, AutoTokenizer , AutoModelForCausalLM
 import torch
 
 # Load model and tokenizer globally to avoid reloading for every request
-model_path = "llama_lora_model_1"
+model_path = "Heit39/llama_lora_model_1"
 
 # Load tokenizer
 tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=True, legacy=False)
 
-# Load model
-model = AutoModelForCausalLM.from_pretrained("Heit39/llama_lora_model_1")
+# Load the base model (e.g., LLaMA)
+base_model = AutoModelForCausalLM.from_pretrained("unsloth/Llama-3.2-3B-Instruct")
+
+# Load LoRA adapter
+from peft import PeftModel
+model = PeftModel.from_pretrained(base_model, model_path)
+
 
 # Define the response function
 def respond(

requirements.txt

@@ -1,4 +1,5 @@
 huggingface_hub==0.25.2
 
 transformers
-accelerate
+accelerate
+peft