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 14, 05:30 UTC
[CI] Add Blackwell LM Eval Small Models test to nightly (#26052)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Oct 5, 2025 at 16:59 UTC
60bc25e74c872ec3df2beee103e11f953c2dfe33
2 files changed
+16
-2
.buildkite/test-pipeline.yaml
+12
-2
@@ -835,11 +835,11 @@ steps:
835
- pytest -v -s tests/kernels/moe/test_flashinfer.py
836
- pytest -v -s tests/compile/test_silu_mul_quant_fusion.py
837
838
-- label: GPT-OSS Eval (Blackwell)
838
+- label: Blackwell GPT-OSS Eval
839
timeout_in_minutes: 60
840
working_dir: "/vllm-workspace/"
841
gpu: b200
842
- optional: true # disable while debugging
842
+ optional: true # run on nightlies
843
source_file_dependencies:
844
- tests/evals/gpt_oss
845
- vllm/model_executor/models/gpt_oss.py
@@ -866,6 +866,16 @@ steps:
866
commands:
867
- pytest -s -v tests/quantization/test_blackwell_moe.py
868
869
+- label: Blackwell LM Eval Small Models
870
+ timeout_in_minutes: 75
871
+ gpu: b200
872
+ optional: true # run on nightlies
873
+ source_file_dependencies:
874
+ - csrc/
875
+ - vllm/model_executor/layers/quantization
876
+ commands:
877
+ - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-blackwell.txt --tp-size=1
878
+
879
##### 1 GPU test #####
880
##### multi gpus test #####
881
tests/evals/gsm8k/configs/models-blackwell.txt
new
+4
@@ -0,0 +1,4 @@
1
+Qwen3-0.6B-FP8.yaml
2
+Qwen2.5-VL-3B-Instruct-FP8-dynamic.yaml
3
+Qwen1.5-MoE-W4A16-CT.yaml
4
+DeepSeek-V2-Lite-Instruct-FP8.yaml