Update inference.py

inference.py

@@ -18,14 +18,13 @@ def load_model_and_tokenizer():
         MODEL_REPO,
         trust_remote_code=True,
     )
-
     model = AutoModelForCausalLM.from_pretrained(
-        MODEL_REPO,
-        trust_remote_code=True,
-        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
-        device_map="auto",
-        torch_dtype=torch.bfloat16
-    )
+    MODEL_REPO,
+    trust_remote_code=True,
+    torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
+    device_map="auto"
+)
+
 
     return tokenizer, model