Update app/main.py

app/main.py

@@ -13,8 +13,9 @@ tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(
     model_id,
     trust_remote_code=True,
-    attn_implementation="eager",  # ✅ This resolves the window_size error
-    device_map="auto"
+    attn_implementation="eager",  # Solves the window_size error
+    device_map="auto",            # Requires accelerate
+    torch_dtype="auto",           # Use the best dtype automatically
 )
 llm_pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)