biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[misc] hide best_of from engine (#9261)

Browse Source 
Co-authored-by: Brendan Wong <bjwpokemon@gmail.com>
This commit is contained in:
youkaichao
2024-10-10 21:30:44 -07:00
committed by GitHub
parent 94bf9ae4e9
commit cbc2ef5529
14 changed files with 46 additions and 73 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/engine/metrics.py
View File

@@ -134,12 +134,6 @@ class Metrics:
labelnames=labelnames,
buckets=build_1_2_5_buckets(max_model_len),
)
self.histogram_best_of_request = self._histogram_cls(
name="vllm:request_params_best_of",
documentation="Histogram of the best_of request parameter.",
labelnames=labelnames,
buckets=[1, 2, 5, 10, 20],
)
self.histogram_n_request = self._histogram_cls(
name="vllm:request_params_n",
documentation="Histogram of the n request parameter.",
@@ -473,8 +467,6 @@ class PrometheusStatLogger(StatLoggerBase):
self.metrics.histogram_num_generation_tokens_request,
stats.num_generation_tokens_requests)
self._log_histogram(self.metrics.histogram_n_request, stats.n_requests)
self._log_histogram(self.metrics.histogram_best_of_request,
stats.best_of_requests)
def _log_prometheus_interval(self, prompt_throughput: float,
generation_throughput: float) -> None: