refactor: use tolist instead of list comprehension calling .item() (#40646)

McPatate · web-flow · commit 459c1fa47a96 · 2025-09-03T19:25:29.000+02:00
diff --git a/src/transformers/commands/serving.py b/src/transformers/commands/serving.py
@@ -834,7 +834,7 @@ def continuous_batching_chat_completion(self, req: dict) -> Generator[str, None,
 
         def stream_chat_completion(_inputs):
             try:
-                decode_stream = DecodeStream([id.item() for id in _inputs], False)
+                decode_stream = DecodeStream(_inputs.tolist(), False)
                 request_id = self.running_continuous_batching_manager.add_request(
                     _inputs, request_id=req.get("request_id"), max_new_tokens=generation_config.max_new_tokens
                 )

Original file line number	Diff line number	Diff line change
`@@ -834,7 +834,7 @@ def continuous_batching_chat_completion(self, req: dict) -> Generator[str, None,`
`834`	`834`
`835`	`835`	`def stream_chat_completion(_inputs):`
`836`	`836`	`try:`
`837`		`- decode_stream = DecodeStream([id.item() for id in _inputs], False)`
	`837`	`+ decode_stream = DecodeStream(_inputs.tolist(), False)`
`838`	`838`	`request_id = self.running_continuous_batching_manager.add_request(`
`839`	`839`	`_inputs, request_id=req.get("request_id"), max_new_tokens=generation_config.max_new_tokens`
`840`	`840`	`)`