Update app.py

app.py

@@ -18,20 +18,31 @@ def load_model_org():
     )
     return model, tokenizer
 
+
 def load_model():
-    tokenizer = AutoTokenizer.from_pretrained(
-        MODEL_NAME,
-        trust_remote_code=True,
-        model_max_length=128256  # Match your model's vocabulary size
+    model, tokenizer = FastLanguageModel.from_pretrained(
+        model_name="satishpednekar/sbxcertqueryhelper",  # Use the path where you saved the model
+        max_seq_length=4096,  # Use the same as during training
+        dtype=torch.float16,
+        load_in_4bit=False,
+        token="ff"
     )
     
-    model = AutoModelForCausalLM.from_pretrained(
-        MODEL_NAME,
-        torch_dtype=torch.float16,
-        device_map="auto",
-        trust_remote_code=True,
-        load_in_8bit=False
+    # Configure PEFT settings exactly as during training
+    model = FastLanguageModel.get_peft_model(
+        model,
+        r=16,
+        target_modules=["q_proj", "k_proj", "v_proj", "o_proj",
+                       "gate_proj", "up_proj", "down_proj"],
+        lora_alpha=16,
+        lora_dropout=0,
+        bias="none",
+        use_gradient_checkpointing="unsloth",
+        random_state=3407,
+        use_rslora=False,
+        loftq_config=None
     )
+    
     return model, tokenizer