A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 07:15 UTC
Adding "LoRA Test %N" to AMD production tests (#18929)
Signed-off-by: Yida Wu <yidawu@alumni.cmu.edu>
Concurrensee committed
Jun 2, 2025 at 22:46 UTC
4ce42f92042ef8a24e925fc7121f7c98e51f73ba
2 files changed
+5
-1
.buildkite/scripts/hardware_ci/run-amd-test.sh
+4
@@ -94,6 +94,10 @@ if [[ $commands == *"pytest -v -s compile/test_basic_correctness.py"* ]]; then
94
commands=${commands//"pytest -v -s compile/test_basic_correctness.py"/"VLLM_USE_TRITON_FLASH_ATTN=0 pytest -v -s compile/test_basic_correctness.py"}
95
fi
96
97
+if [[ $commands == *"pytest -v -s lora"* ]]; then
98
+ commands=${commands//"pytest -v -s lora"/"VLLM_ROCM_CUSTOM_PAGED_ATTN=0 pytest -v -s lora"}
99
+fi
100
+
101
#ignore certain kernels tests
102
if [[ $commands == *" kernels/core"* ]]; then
103
commands="${commands} \
.buildkite/test-pipeline.yaml
+1
-1
@@ -287,7 +287,7 @@ steps:
287
- pytest -v -s spec_decode/e2e/test_eagle_correctness.py
288
289
- label: LoRA Test %N # 15min each
290
- mirror_hardwares: [amdexperimental]
290
+ mirror_hardwares: [amdexperimental, amdproduction]
291
source_file_dependencies:
292
- vllm/lora
293
- tests/lora