From 54818d4cbc1976d09f7dd510e4fa13e5f8ce06e3 Mon Sep 17 00:00:00 2001
From: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
Date: Sun, 24 Mar 2024 16:37:25 +0100
Subject: [PATCH] server: bench: init

---
 .github/workflows/bench.yml | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/bench.yml b/.github/workflows/bench.yml
index 445bc34940e5cf..eef9f7d279c91d 100644
--- a/.github/workflows/bench.yml
+++ b/.github/workflows/bench.yml
@@ -55,7 +55,7 @@ jobs:
               -DLLAMA_CUBLAS=ON \
               -DCUDAToolkit_ROOT=/usr/local/cuda \
               -DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc \
-              -DCMAKE_CUDA_ARCHITECTURES=80 \
+              -DCMAKE_CUDA_ARCHITECTURES=75 \
               -DLLAMA_FATAL_WARNINGS=OFF \
               -DLLAMA_ALL_WARNINGS=OFF \
               -DCMAKE_BUILD_TYPE=Release;
@@ -77,7 +77,7 @@ jobs:
         id: server_bench
         run: |
           build/bin/server \
-              --host localhost \
+              --host 0.0.0.0 \
               --port 8080 \
               --hf-repo ggml-org/models	 \
               --hf-file phi-2/ggml-model-q4_0.gguf \
@@ -95,5 +95,9 @@ jobs:
             sleep 0.1
           done
 
+          while [[ "$(curl -s -o /dev/null -w ''%{http_code}'' localhost:8080)" != "200" ]]; do
+            sleep 0.5;
+          done
+
           cd examples/server/bench
           ../../../k6 run script.js --duration 10m --iterations 500 --vus 8