Update app.py

app.py

@@ -315,7 +315,7 @@ class Phi3MiniEducationalLLM(Runnable):
                     model_path,
                     quantization_config=quant_config,
                     device_map="auto",
-                    dtype=torch.float16,
+                    torch_dtype=torch.float16,
                     trust_remote_code=True,
                     low_cpu_mem_usage=True,
                     token=hf_token
@@ -343,7 +343,7 @@ class Phi3MiniEducationalLLM(Runnable):
         """Optimized model loading for Phi-3-mini."""
         self.model = AutoModelForCausalLM.from_pretrained(
             model_path,
-            dtype=torch.float16,  # Use float16 to save memory
+            torch_dtype=torch.float16,  # Use float16 to save memory
             device_map="auto",          # Let transformers decide placement
             trust_remote_code=True,
             low_cpu_mem_usage=True,
@@ -407,7 +407,7 @@ class Phi3MiniEducationalLLM(Runnable):
                     repetition_penalty=1.1,
                     pad_token_id=self.tokenizer.eos_token_id,
                     early_stopping=True,
-                    use_cache=True
+                    use_cache=False         # Disable cache to avoid compatibility issues
                 )
 
             # Decode only new tokens