From 25b4e0b1b65c2425056689afb5cd9426a3faf2ba Mon Sep 17 00:00:00 2001
From: Misha Chornyi <mchornyi@nvidia.com>
Date: Thu, 25 Sep 2025 20:21:08 -0700
Subject: [PATCH] Update L0_multi_gpu_vllm script in order to be able access
 resources within container. Those resources no longer shipped with
 'pytorch-triton' wheel package.

---
 ci/L0_multi_gpu_vllm/multi_lora/test.sh   | 3 +++
 ci/L0_multi_gpu_vllm/vllm_backend/test.sh | 3 +++
 2 files changed, 6 insertions(+)

diff --git a/ci/L0_multi_gpu_vllm/multi_lora/test.sh b/ci/L0_multi_gpu_vllm/multi_lora/test.sh
index 8cf2c3fa..c045c4fc 100755
--- a/ci/L0_multi_gpu_vllm/multi_lora/test.sh
+++ b/ci/L0_multi_gpu_vllm/multi_lora/test.sh
@@ -41,6 +41,9 @@ EXPECTED_NUM_TESTS=2
 GENERATE_ENDPOINT="localhost:8000/v2/models/vllm_llama_multi_lora/generate"
 CHECK_FOR_ERROR=true
 
+export C_INCLUDE_PATH=/usr/local/cuda/include:$C_INCLUDE_PATH
+export TRITON_PTXAS_PATH=/usr/local/cuda/bin/ptxas
+
 make_api_call() {
     local endpoint="$1"
     local data="$2"
diff --git a/ci/L0_multi_gpu_vllm/vllm_backend/test.sh b/ci/L0_multi_gpu_vllm/vllm_backend/test.sh
index 0609bebf..e4de2ad2 100755
--- a/ci/L0_multi_gpu_vllm/vllm_backend/test.sh
+++ b/ci/L0_multi_gpu_vllm/vllm_backend/test.sh
@@ -36,6 +36,9 @@ CLIENT_PY="./vllm_multi_gpu_test.py"
 SAMPLE_MODELS_REPO="../../../samples/model_repository"
 EXPECTED_NUM_TESTS=1
 
+export C_INCLUDE_PATH=/usr/local/cuda/include:$C_INCLUDE_PATH
+export TRITON_PTXAS_PATH=/usr/local/cuda/bin/ptxas
+
 ### Helpers
 function validate_file_contains() {
     local KEY="${1}"