[Metrics] Deprecate TPOT in favor of ITL (#24110)

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
This commit is contained in:
Mark McLoughlin
2025-09-02 18:10:10 +00:00
committed by GitHub
parent 9480ae24e3
commit 2417798471
7 changed files with 64 additions and 26 deletions
@@ -402,7 +402,7 @@
},
"disableTextWrap": false,
"editorMode": "builder",
"expr": "histogram_quantile(0.99, sum by(le) (rate(vllm:time_per_output_token_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
"expr": "histogram_quantile(0.99, sum by(le) (rate(vllm:inter_token_latency_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
"fullMetaSearch": false,
"includeNullMetadata": false,
"instant": false,
@@ -418,7 +418,7 @@
},
"disableTextWrap": false,
"editorMode": "builder",
"expr": "histogram_quantile(0.95, sum by(le) (rate(vllm:time_per_output_token_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
"expr": "histogram_quantile(0.95, sum by(le) (rate(vllm:inter_token_latency_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
"fullMetaSearch": false,
"hide": false,
"includeNullMetadata": false,
@@ -435,7 +435,7 @@
},
"disableTextWrap": false,
"editorMode": "builder",
"expr": "histogram_quantile(0.9, sum by(le) (rate(vllm:time_per_output_token_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
"expr": "histogram_quantile(0.9, sum by(le) (rate(vllm:inter_token_latency_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
"fullMetaSearch": false,
"hide": false,
"includeNullMetadata": false,
@@ -452,7 +452,7 @@
},
"disableTextWrap": false,
"editorMode": "builder",
"expr": "histogram_quantile(0.5, sum by(le) (rate(vllm:time_per_output_token_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
"expr": "histogram_quantile(0.5, sum by(le) (rate(vllm:inter_token_latency_seconds_bucket{model_name=\"$model_name\"}[$__rate_interval])))",
"fullMetaSearch": false,
"hide": false,
"includeNullMetadata": false,
@@ -468,7 +468,7 @@
"uid": "${DS_PROMETHEUS}"
},
"editorMode": "code",
"expr": "rate(vllm:time_per_output_token_seconds_sum{model_name=\"$model_name\"}[$__rate_interval])\n/\nrate(vllm:time_per_output_token_seconds_count{model_name=\"$model_name\"}[$__rate_interval])",
"expr": "rate(vllm:inter_token_latency_seconds_sum{model_name=\"$model_name\"}[$__rate_interval])\n/\nrate(vllm:inter_token_latency_seconds_count{model_name=\"$model_name\"}[$__rate_interval])",
"hide": false,
"instant": false,
"legendFormat": "Mean",
@@ -476,7 +476,7 @@
"refId": "E"
}
],
"title": "Time Per Output Token Latency",
"title": "Inter Token Latency",
"type": "timeseries"
},
{