A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Misc] Separate prompt logging to debug (#26713)
Signed-off-by: Aleksei Tsvetkov <aitsvet@ya.ru>
Aleksei Tsvetkov committed
Oct 13, 2025 at 19:04 UTC
5c7fe25491825b95936c011a43337c7d4fb7e472
1 file changed
+10
-6
vllm/entrypoints/logger.py
+10
-6
@@ -34,16 +34,20 @@ class RequestLogger:
34
if prompt_token_ids is not None:
35
prompt_token_ids = prompt_token_ids[:max_log_len]
36
37
- logger.info(
38
- "Received request %s: prompt: %r, "
39
- "params: %s, prompt_token_ids: %s, "
40
- "prompt_embeds shape: %s, "
41
- "lora_request: %s.",
37
+ logger.debug(
38
+ "Request %s details: prompt: %r, "
39
+ "prompt_token_ids: %s, "
40
+ "prompt_embeds shape: %s.",
41
request_id,
42
prompt,
44
- params,
43
prompt_token_ids,
44
prompt_embeds.shape if prompt_embeds is not None else None,
45
+ )
46
+
47
+ logger.info(
48
+ "Received request %s: params: %s, lora_request: %s.",
49
+ request_id,
50
+ params,
51
lora_request,
52
)
53