xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-25 03:15:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[V0][Metrics] Deprecate some questionable request time metrics (#14135)

Browse Source 
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
This commit is contained in:
Mark McLoughlin 2025-03-04 15:11:33 +00:00 committed by GitHub
parent 5db6b2c961
commit c8525f06fc
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 17 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
vllm/engine/metrics.py
View File

@ -197,24 +197,35 @@ class Metrics:
"Histogram of time spent in DECODE phase for request.",
labelnames=labelnames,
buckets=request_latency_buckets)
# Deprecated in 0.8 - duplicates vllm:request_queue_time_seconds:
# TODO: in 0.9, only enable if show_hidden_metrics=True
self.histogram_time_in_queue_request = self._histogram_cls(
name="vllm:time_in_queue_requests",
documentation=
"Histogram of time the request spent in the queue in seconds.",
documentation=(
"Histogram of time the request spent in the queue in seconds. "
"DEPRECATED: use vllm:request_queue_time_seconds instead."),
labelnames=labelnames,
buckets=request_latency_buckets)
# Deprecated in 0.8 - use prefill/decode/inference time metrics
# TODO: in 0.9, only enable if show_hidden_metrics=True
self.histogram_model_forward_time_request = self._histogram_cls(
name="vllm:model_forward_time_milliseconds",
documentation=
"Histogram of time spent in the model forward pass in ms.",
documentation=(
"Histogram of time spent in the model forward pass in ms. "
"DEPRECATED: use prefill/decode/inference time metrics instead."
),
labelnames=labelnames,
buckets=build_1_2_3_5_8_buckets(3000))
self.histogram_model_execute_time_request = self._histogram_cls(
name="vllm:model_execute_time_milliseconds",
documentation=
"Histogram of time spent in the model execute function in ms.",
documentation=(
"Histogram of time spent in the model execute function in ms."
"DEPRECATED: use prefill/decode/inference time metrics instead."
),
labelnames=labelnames,
buckets=build_1_2_3_5_8_buckets(3000))
# Metadata
self.histogram_num_prompt_tokens_request = self._histogram_cls(
name="vllm:request_prompt_tokens",