From 6028879f56a9b8c2ac1b0d14270f38998c8ec0f2 Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Tue, 19 Sep 2023 23:50:05 +0300 Subject: [PATCH] parallel : print misses on each request --- examples/parallel/parallel.cpp | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/examples/parallel/parallel.cpp b/examples/parallel/parallel.cpp index b674a0344..0f1df36d6 100644 --- a/examples/parallel/parallel.cpp +++ b/examples/parallel/parallel.cpp @@ -326,12 +326,10 @@ int main(int argc, char ** argv) { const auto t_main_end = ggml_time_us(); - LOG_TEE("\033[1mClient %3d, seq %4d, prompt %4d t, response %4d t, time %5.2f s, speed: PP %5.2f t/s, TG %5.2f t/s, AVG %5.2f t/s \033[0m: \n\nInput: %s\nResponse: %s\n\n", + LOG_TEE("\033[1mClient %3d, seq %4d, prompt %4d t, response %4d t, time %5.2f s, cache miss %d \033[0m: \n\nInput: %s\nResponse: %s\n\n", client.id, client.seq_id, client.n_prompt, client.n_decoded, (t_main_end - client.t_start_prompt) / 1e6, - (double) (client.n_prompt ) / (client.t_start_gen - client.t_start_prompt) * 1e6, - (double) (client.n_decoded ) / (t_main_end - client.t_start_gen) * 1e6, - (double) (client.n_decoded + client.n_prompt) / (t_main_end - client.t_start_prompt) * 1e6, + n_cache_miss, ::trim(client.input).c_str(), ::trim(client.response).c_str());