From d6c8426b9136bcf86b77cbc03db060811cb11c51 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Dani=C3=ABl=20de=20Kok?= Date: Tue, 12 Nov 2024 09:17:37 +0000 Subject: [PATCH] benchmark: fix prefill throughput --- benchmark/src/generation.rs | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/benchmark/src/generation.rs b/benchmark/src/generation.rs index 63fc7808..60d96f70 100644 --- a/benchmark/src/generation.rs +++ b/benchmark/src/generation.rs @@ -180,7 +180,7 @@ async fn prefill( let latency = start_time.elapsed(); // Compute throughput from latency and batch size - let throughput = batch_size as f64 / latency.as_secs_f64(); + let throughput = (batch_size * sequence_length) as f64 / latency.as_secs_f64(); // Decode batch cannot be empty let decode_batch = decode_batch.expect("decode_batch is None. This is a bug.");