From 1b4b2a18b80c5e265b93ab6f83dde494eadfaa4a Mon Sep 17 00:00:00 2001 From: "Kevin H. Luu" Date: Wed, 2 Sep 2026 01:10:14 -0700 Subject: [PATCH] [CI] Shard H100 MoE refactor integration tests (#52352) Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com> --- .buildkite/test_areas/lm_eval.yaml | 6 +++++- .../configs/moe-refactor/config-h100-shard-0.txt | 2 ++ .../configs/moe-refactor/config-h100-shard-1.txt | 3 +++ .../configs/moe-refactor/config-h100-shard-2.txt | 3 +++ .../configs/moe-refactor/config-h100-shard-3.txt | 4 ++++ .../evals/gsm8k/configs/moe-refactor/config-h100.txt | 12 ------------ 6 files changed, 17 insertions(+), 13 deletions(-) create mode 100644 tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-0.txt create mode 100644 tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-1.txt create mode 100644 tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-2.txt create mode 100644 tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-3.txt delete mode 100644 tests/evals/gsm8k/configs/moe-refactor/config-h100.txt diff --git a/.buildkite/test_areas/lm_eval.yaml b/.buildkite/test_areas/lm_eval.yaml index 86b6bf0042..1b43c707eb 100644 --- a/.buildkite/test_areas/lm_eval.yaml +++ b/.buildkite/test_areas/lm_eval.yaml @@ -177,8 +177,12 @@ steps: device: h100 optional: true num_devices: 2 + parallelism: 4 commands: - - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/moe-refactor/config-h100.txt + # Timing-balanced static split of the former config-h100.txt (measured on + # build 83851): shards ~12.4/13.6/12.4/14.5m of eval time. New configs must + # be added to exactly one config-h100-shard-*.txt. + - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/moe-refactor/config-h100-shard-$$BUILDKITE_PARALLEL_JOB.txt - label: ":nvidia: (B200) MoE Refactor Integration TEMPORARY Shard %N" key: moe-refactor-integration-test-b200-temporary diff --git a/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-0.txt b/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-0.txt new file mode 100644 index 0000000000..75d5184f96 --- /dev/null +++ b/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-0.txt @@ -0,0 +1,2 @@ +Qwen3-30B-A3B-Fp8-AutoFp8-deepgemm.yaml +Qwen3-30B-A3B-BF16-fi-cutlass.yaml diff --git a/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-1.txt b/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-1.txt new file mode 100644 index 0000000000..2f2b177967 --- /dev/null +++ b/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-1.txt @@ -0,0 +1,3 @@ +Qwen3-30B-A3B-Fp8-AutoFp8-marlin.yaml +Qwen3-30B-A3B-Fp8-AutoFp8-triton.yaml +Qwen3-30B-A3B-BF16-triton.yaml diff --git a/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-2.txt b/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-2.txt new file mode 100644 index 0000000000..0ff9a70b5d --- /dev/null +++ b/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-2.txt @@ -0,0 +1,3 @@ +Qwen3-30B-A3B-Fp8-AutoFp8-fi-cutlass.yaml +Qwen3-30B-A3B-Fp8-CT-Block-deepgemm.yaml +Qwen3-30B-A3B-Fp8-CT-Block-triton.yaml diff --git a/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-3.txt b/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-3.txt new file mode 100644 index 0000000000..d99e3d1407 --- /dev/null +++ b/tests/evals/gsm8k/configs/moe-refactor/config-h100-shard-3.txt @@ -0,0 +1,4 @@ +Mixtral-8x7B-Fp8-AutoFp8-triton.yaml +Qwen3-30B-A3B-Fp8-CT-Block-marlin.yaml +Qwen3-30B-A3B-Fp8-CT-Channel-marlin.yaml +Qwen3-30B-A3B-Fp8-CT-Channel-vllm-cutlass.yaml diff --git a/tests/evals/gsm8k/configs/moe-refactor/config-h100.txt b/tests/evals/gsm8k/configs/moe-refactor/config-h100.txt deleted file mode 100644 index 7397fc4e46..0000000000 --- a/tests/evals/gsm8k/configs/moe-refactor/config-h100.txt +++ /dev/null @@ -1,12 +0,0 @@ -Mixtral-8x7B-Fp8-AutoFp8-triton.yaml -Qwen3-30B-A3B-Fp8-AutoFp8-deepgemm.yaml -Qwen3-30B-A3B-Fp8-AutoFp8-fi-cutlass.yaml -Qwen3-30B-A3B-Fp8-AutoFp8-marlin.yaml -Qwen3-30B-A3B-Fp8-AutoFp8-triton.yaml -Qwen3-30B-A3B-Fp8-CT-Block-deepgemm.yaml -Qwen3-30B-A3B-Fp8-CT-Block-marlin.yaml -Qwen3-30B-A3B-Fp8-CT-Block-triton.yaml -Qwen3-30B-A3B-Fp8-CT-Channel-marlin.yaml -Qwen3-30B-A3B-Fp8-CT-Channel-vllm-cutlass.yaml -Qwen3-30B-A3B-BF16-fi-cutlass.yaml -Qwen3-30B-A3B-BF16-triton.yaml