huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-09 19:34:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix buckets

Browse Source
This commit is contained in:
OlivierDehaene 2023-04-09 17:38:41 +02:00
parent 18e44a6aad
commit 23b55861bc
1 changed files with 7 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
router/src/server.rs
View File

@ -528,22 +528,22 @@ pub async fn run(
} }
// Input Length buckets // Input Length buckets
let input_length_matcher = Matcher::Full(String::from("tgi_request_input_length")); let input_length_matcher = Matcher::Full(String::from("tgi_request_input_length"));
let input_length_buckets: Vec<f64> = (1..100) let input_length_buckets: Vec<f64> = (0..100)
.map(|x| (max_input_length as f64 / 100.0) * x as f64) .map(|x| (max_input_length as f64 / 100.0) * (x + 1) as f64)
.collect(); .collect();
// Generated tokens buckets // Generated tokens buckets
let generated_tokens_matcher = Matcher::Full(String::from("tgi_request_generated_tokens")); let generated_tokens_matcher = Matcher::Full(String::from("tgi_request_generated_tokens"));
let generated_tokens_buckets: Vec<f64> = (1..100) let generated_tokens_buckets: Vec<f64> = (0..100)
.map(|x| (max_total_tokens as f64 / 100.0) * x as f64) .map(|x| (max_total_tokens as f64 / 100.0) * (x + 1) as f64)
.collect(); .collect();
// Input Length buckets // Input Length buckets
let max_new_tokens_matcher = Matcher::Full(String::from("tgi_request_max_new_tokens")); let max_new_tokens_matcher = Matcher::Full(String::from("tgi_request_max_new_tokens"));
let max_new_tokens_buckets: Vec<f64> = (1..100) let max_new_tokens_buckets: Vec<f64> = (0..100)
.map(|x| (max_total_tokens as f64 / 100.0) * x as f64) .map(|x| (max_total_tokens as f64 / 100.0) * (x + 1) as f64)
.collect(); .collect();
// Batch size buckets // Batch size buckets
let batch_size_matcher = Matcher::Full(String::from("tgi_batch_next_size")); let batch_size_matcher = Matcher::Full(String::from("tgi_batch_next_size"));
let batch_size_buckets: Vec<f64> = (1..max_batch_size).map(|x| x as f64).collect(); let batch_size_buckets: Vec<f64> = (0..max_batch_size).map(|x| (x + 1) as f64).collect();
// Prometheus handler // Prometheus handler
let builder = PrometheusBuilder::new() let builder = PrometheusBuilder::new()