[CI] Shard MoE refactor B200 eval (#52327)
Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com>
Kevin H. Luu committed
Aug 14, 2026 at 15:11 UTC
81e81dac1f3654564c3c48000edce1cdbe248ab9
5 files changed
+22
-2
.buildkite/test_areas/lm_eval.yaml
+3
-2
@@ -178,13 +178,14 @@ steps:
178
commands:
179
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/moe-refactor/config-h100.txt
180
181
-- label: MoE Refactor Integration Test (B200 - TEMPORARY)
181
+- label: MoE Refactor Integration Test (B200 - TEMPORARY) %N
182
key: moe-refactor-integration-test-b200-temporary
183
device: b200-k8s
184
optional: true
185
num_devices: 2
186
+ parallelism: 4
187
commands:
187
- - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/moe-refactor/config-b200.txt
188
+ - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/moe-refactor/config-b200-shard-$$BUILDKITE_PARALLEL_JOB.txt
189
190
- label: MoE Refactor Integration Test (B200 DP - TEMPORARY)
191
key: moe-refactor-integration-test-b200-dp-temporary
tests/evals/gsm8k/configs/moe-refactor/config-b200-shard-0.txt
new
+3
@@ -0,0 +1,3 @@
1
+DeepSeek-V4-Flash-deepgemm-mega-moe.yaml
2
+Qwen3-30B-A3B-NvFp4-ModelOpt-marlin.yaml
3
+Qwen3-30B-A3B-NvFp4-ModelOpt-fi-trtllm.yaml
tests/evals/gsm8k/configs/moe-refactor/config-b200-shard-1.txt
new
+5
@@ -0,0 +1,5 @@
1
+Nemotron-Nano-30B-Fp8-ModelOpt-fi-trtllm.yaml
2
+Nemotron-Nano-30B-NvFp4-ModelOpt-fi-cutlass.yaml
3
+Llama-4-Scout-Fp8-CT-vllm-cutlass.yaml
4
+Qwen3-30B-A3B-NvFp4-ModelOpt-fi-cutlass.yaml
5
+Qwen3-30B-A3B-NvFp4-CT-fi-cutlass.yaml
tests/evals/gsm8k/configs/moe-refactor/config-b200-shard-2.txt
new
+5
@@ -0,0 +1,5 @@
1
+Llama-4-Scout-Fp8-ModelOpt-fi-trtllm.yaml
2
+Qwen3-30B-A3B-NvFp4-CT-fi-trtllm.yaml
3
+Qwen3-30B-A3B-NvFp4-CT-vllm-cutlass.yaml
4
+Llama-4-Scout-BF16-triton.yaml
5
+Qwen3-30B-A3B-NvFp4-ModelOpt-vllm-cutlass.yaml
tests/evals/gsm8k/configs/moe-refactor/config-b200-shard-3.txt
new
+6
@@ -0,0 +1,6 @@
1
+Llama-4-Scout-BF16-fi-cutlass.yaml
2
+Qwen3-30B-A3B-Fp8-AutoFp8-fi-trtllm.yaml
3
+Nemotron-Nano-30B-NvFp4-ModelOpt-vllm-cutlass.yaml
4
+Mixtral-8x7B-BF16-fi-cutlass.yaml
5
+Qwen3-30B-A3B-NvFp4-CT-marlin.yaml
6
+Mixtral-8x7B-BF16-triton.yaml