@tduncan / Nvidia-Optimization-Stack-VLLM / commits / 78fa18910e

ci: add MIG slice size to H200 job labels (#54420)

Signed-off-by: khluu <khluu000@gmail.com>

Kevin H. Luu committed Aug 30, 2026 at 04:35 UTC 78fa18910ee2fe64b2624e9358f622e1b50b14bf
23 files changed +85 -85
.buildkite/test_areas/basic_correctness.yaml
+1 -1
@@ -2,7 +2,7 @@ group: Basic Correctness
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Basic Correctness"
5 +- label: ":nvidia: (H200 MIG 18GB) Basic Correctness"
6 key: basic-correctness
7 timeout_in_minutes: 68
8 device: h200_18gb
.buildkite/test_areas/benchmarks.yaml
+1 -1
@@ -2,7 +2,7 @@ group: Benchmarks
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Benchmarks CLI"
5 +- label: ":nvidia: (H200 MIG 18GB) Benchmarks CLI"
6 key: benchmarks-cli-test
7 timeout_in_minutes: 45
8 device: h200_18gb
.buildkite/test_areas/cuda.yaml
+2 -2
@@ -2,7 +2,7 @@ group: CUDA
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) CUDA Platform"
5 +- label: ":nvidia: (H200 MIG 18GB) CUDA Platform"
6 key: platform-tests
7 timeout_in_minutes: 20
8 device: h200_18gb
@@ -30,7 +30,7 @@ steps:
30 - vllm/
31 - tests/cuda
32
33 -- label: ":nvidia: (H200) CUDAGraph"
33 +- label: ":nvidia: (H200 MIG 35GB) CUDAGraph"
34 device: h200_35gb
35 key: cudagraph
36 timeout_in_minutes: 30
.buildkite/test_areas/engine.yaml
+3 -3
@@ -2,7 +2,7 @@ group: Engine
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Engine Core"
5 +- label: ":nvidia: (H200 MIG 18GB) Engine Core"
6 key: engine
7 timeout_in_minutes: 30
8 device: h200_18gb
@@ -55,7 +55,7 @@ steps:
55 depends_on:
56 - image-build-amd
57
58 -- label: ":nvidia: (H200) E2E Scheduling"
58 +- label: ":nvidia: (H200 MIG 18GB) E2E Scheduling"
59 key: e2e-scheduling-1-gpu
60 timeout_in_minutes: 53
61 device: h200_18gb
@@ -72,7 +72,7 @@ steps:
72 depends_on:
73 - image-build-amd
74
75 -- label: ":nvidia: (H200) E2E Core"
75 +- label: ":nvidia: (H200 MIG 35GB) E2E Core"
76 device: h200_35gb
77 key: e2e-core-1-gpu
78 timeout_in_minutes: 40
.buildkite/test_areas/entrypoints.yaml
+11 -11
@@ -2,7 +2,7 @@ group: Entrypoints
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Entrypoints Unit"
5 +- label: ":nvidia: (H200 MIG 35GB) Entrypoints Unit"
6 device: h200_35gb
7 key: entrypoints-unit-tests
8 timeout_in_minutes: 40
@@ -25,7 +25,7 @@ steps:
25 depends_on:
26 - image-build-amd
27
28 -- label: ":nvidia: (H200) Entrypoints Integration (LLM)"
28 +- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (LLM)"
29 device: h200_35gb
30 key: entrypoints-integration-llm
31 timeout_in_minutes: 60
@@ -48,7 +48,7 @@ steps:
48 depends_on:
49 - image-build-amd
50
51 -- label: ":nvidia: (H200) Entrypoints Integration (API Server)"
51 +- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (API Server)"
52 key: entrypoints-integration-api-server
53 device: h200_35gb
54 timeout_in_minutes: 75
@@ -72,7 +72,7 @@ steps:
72 depends_on:
73 - image-build-amd
74
75 -- label: ":nvidia: (H200) Entrypoints Integration (API Server OpenAI - Part 1)"
75 +- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (API Server OpenAI - Part 1)"
76 device: h200_35gb
77 key: entrypoints-integration-api-server-openai-part-1
78 timeout_in_minutes: 68
@@ -94,7 +94,7 @@ steps:
94 depends_on:
95 - image-build-amd
96
97 -- label: ":nvidia: (H200) Entrypoints Integration (API Server OpenAI - Part 2)"
97 +- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (API Server OpenAI - Part 2)"
98 device: h200_35gb
99 key: entrypoints-integration-api-server-openai-part-2
100 timeout_in_minutes: 83
@@ -117,7 +117,7 @@ steps:
117 depends_on:
118 - image-build-amd
119
120 -- label: ":nvidia: (H200) Entrypoints Integration (API Server Generate)"
120 +- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (API Server Generate)"
121 device: h200_35gb
122 key: entrypoints-integration-api-server-generate
123 timeout_in_minutes: 50
@@ -143,7 +143,7 @@ steps:
143 depends_on:
144 - image-build-amd
145
146 -- label: ":nvidia: (H200) Entrypoints Integration (Responses API)"
146 +- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (Responses API)"
147 device: h200_35gb
148 key: entrypoints-integration-responses-api
149 timeout_in_minutes: 50
@@ -163,7 +163,7 @@ steps:
163 depends_on:
164 - image-build-amd
165
166 -- label: ":nvidia: (H200) Entrypoints Integration (Speech to Text)"
166 +- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (Speech to Text)"
167 device: h200_35gb
168 key: entrypoints-integration-speech_to_text
169 timeout_in_minutes: 45
@@ -176,7 +176,7 @@ steps:
176 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
177 - pytest -v -s entrypoints/speech_to_text
178
179 -- label: ":nvidia: (H200) Entrypoints Integration (Multimodal)"
179 +- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (Multimodal)"
180 device: h200_35gb
181 key: entrypoints-integration-multimodal
182 timeout_in_minutes: 45
@@ -197,7 +197,7 @@ steps:
197 depends_on:
198 - image-build-amd
199
200 -- label: ":nvidia: (H200) Entrypoints Integration (Pooling)"
200 +- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (Pooling)"
201 device: h200_35gb
202 key: entrypoints-integration-pooling
203 timeout_in_minutes: 75
@@ -210,7 +210,7 @@ steps:
210 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
211 - pytest -v -s entrypoints/pooling
212
213 -- label: ":nvidia: (H200) OpenAI API Correctness"
213 +- label: ":nvidia: (H200 MIG 18GB) OpenAI API Correctness"
214 key: openai-api-correctness
215 timeout_in_minutes: 20
216 device: h200_18gb
.buildkite/test_areas/expert_parallelism.yaml
+1 -1
@@ -2,7 +2,7 @@ group: Expert Parallelism
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) EPLB Algorithm"
5 +- label: ":nvidia: (H200 MIG 18GB) EPLB Algorithm"
6 key: eplb-algorithm
7 timeout_in_minutes: 20
8 device: h200_18gb
.buildkite/test_areas/jit_monitor.yaml
+1 -1
@@ -2,7 +2,7 @@ group: JIT Monitor
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) No Runtime JITs E2E"
5 +- label: ":nvidia: (H200 MIG 35GB) No Runtime JITs E2E"
6 key: jit-monitor-no-runtime-jit
7 device: h200_35gb
8 timeout_in_minutes: 45
.buildkite/test_areas/kernels.yaml
+3 -3
@@ -2,7 +2,7 @@ group: Kernels
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) vLLM IR"
5 +- label: ":nvidia: (H200 MIG 18GB) vLLM IR"
6 key: vllm-ir-tests
7 timeout_in_minutes: 35
8 device: h200_18gb
@@ -14,7 +14,7 @@ steps:
14 - pytest -v -s tests/ir
15 - pytest -v -s tests/kernels/ir
16
17 -- label: ":nvidia: (H200) Core Operation Kernels Shard %N"
17 +- label: ":nvidia: (H200 MIG 35GB) Core Operation Kernels Shard %N"
18 device: h200_35gb
19 key: kernels-core-operation-test
20 timeout_in_minutes: 120
@@ -225,7 +225,7 @@ steps:
225 depends_on:
226 - image-build-amd
227
228 -- label: ":nvidia: (H200) Mamba Kernels"
228 +- label: ":nvidia: (H200 MIG 35GB) Mamba Kernels"
229 device: h200_35gb
230 key: kernels-mamba-test
231 timeout_in_minutes: 60
.buildkite/test_areas/lm_eval.yaml
+4 -4
@@ -2,7 +2,7 @@ group: LM Eval
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) LM Eval Small Models"
5 +- label: ":nvidia: (H200 MIG 35GB) LM Eval Small Models"
6 device: h200_35gb
7 key: lm-eval-small-models
8 timeout_in_minutes: 45
@@ -291,7 +291,7 @@ steps:
291 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/humming/config-act-fp8.txt
292 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/humming/config-act-int8.txt
293
294 -- label: ":nvidia: (H200) LM Eval TurboQuant KV Cache"
294 +- label: ":nvidia: (H200 MIG 18GB) LM Eval TurboQuant KV Cache"
295 key: lm-eval-turboquant-kv-cache
296 timeout_in_minutes: 55
297 device: h200_18gb
@@ -347,7 +347,7 @@ steps:
347 - uv pip install --system 'gpt-oss[eval]==0.0.5'
348 - pytest -s -v evals/gpt_oss/test_gpqa_correctness.py --config-list-file=configs/models-spark.txt
349
350 -- label: ":nvidia: (H200) LM Eval KV-Offload"
350 +- label: ":nvidia: (H200 MIG 35GB) LM Eval KV-Offload"
351 key: kv-offload-small
352 timeout_in_minutes: 30
353 device: h200_35gb
@@ -388,7 +388,7 @@ steps:
388 commands:
389 - pytest -s -v evals/gsm8k/test_gsm8k_offloading.py -k "deepseek-v4-flash"
390
391 -- label: ":nvidia: (H200) MRCR Eval Small Models"
391 +- label: ":nvidia: (H200 MIG 35GB) MRCR Eval Small Models"
392 device: h200_35gb
393 timeout_in_minutes: 25
394 source_file_dependencies:
.buildkite/test_areas/lora.yaml
+1 -1
@@ -2,7 +2,7 @@ group: LoRA
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) LoRA Shard %N"
5 +- label: ":nvidia: (H200 MIG 35GB) LoRA Shard %N"
6 device: h200_35gb
7 key: lora
8 timeout_in_minutes: 40
.buildkite/test_areas/misc.yaml
+8 -8
@@ -2,7 +2,7 @@ group: Miscellaneous
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) V1 Sample + Logits"
5 +- label: ":nvidia: (H200 MIG 18GB) V1 Sample + Logits"
6 key: v1-sample-logits
7 timeout_in_minutes: 83
8 device: h200_18gb
@@ -40,7 +40,7 @@ steps:
40 depends_on:
41 - image-build-amd
42
43 -- label: ":nvidia: (H200) V1 Core + KV + Metrics"
43 +- label: ":nvidia: (H200 MIG 35GB) V1 Core + KV + Metrics"
44 device: h200_35gb
45 key: v1-core-kv-metrics
46 timeout_in_minutes: 80
@@ -132,7 +132,7 @@ steps:
132 - pytest -v -s -m 'cpu_test' v1/ec_connector/unit
133 - pytest -v -s -m 'cpu_test' v1/metrics
134
135 -- label: ":nvidia: (H200) Extract Hidden States Integration"
135 +- label: ":nvidia: (H200 MIG 18GB) Extract Hidden States Integration"
136 key: extract-hidden-states-integration
137 timeout_in_minutes: 20
138 device: h200_18gb
@@ -162,7 +162,7 @@ steps:
162 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
163 - pytest -v -s -m 'distributed' tests/v1/kv_connector/extract_hidden_states_integration
164
165 -- label: ":nvidia: (H200) Regression"
165 +- label: ":nvidia: (H200 MIG 18GB) Regression"
166 key: regression
167 timeout_in_minutes: 30
168 device: h200_18gb
@@ -185,7 +185,7 @@ steps:
185 - pytest -v -s test_regression.py
186 working_dir: "/vllm-workspace/tests" # optional
187
188 -- label: ":nvidia: (H200) Examples"
188 +- label: ":nvidia: (H200 MIG 35GB) Examples"
189 device: h200_35gb
190 key: examples
191 timeout_in_minutes: 40
@@ -268,7 +268,7 @@ steps:
268 depends_on:
269 - image-build-amd
270
271 -- label: ":nvidia: (H200) Python-only Installation"
271 +- label: ":nvidia: (H200 MIG 18GB) Python-only Installation"
272 key: python-only-installation
273 depends_on: ~
274 optional: true
@@ -293,7 +293,7 @@ steps:
293 - setup.py
294 - vllm/platforms/rocm.py
295
296 -- label: ":nvidia: (H200) Async Engine, Inputs, Utils, Worker"
296 +- label: ":nvidia: (H200 MIG 35GB) Async Engine, Inputs, Utils, Worker"
297 device: h200_35gb
298 key: async-engine-inputs-utils-worker
299 timeout_in_minutes: 25
@@ -440,7 +440,7 @@ steps:
440 - pytest -v -s v1/determinism/test_cutlass_batch_invariance.py
441 - pytest -v -s v1/determinism/test_online_batch_invariance.py
442
443 -- label: ":nvidia: (H200) Acceptance Length (Large Models)"
443 +- label: ":nvidia: (H200 MIG 35GB) Acceptance Length (Large Models)"
444 device: h200_35gb
445 key: acceptance-length-test-large-models
446 timeout_in_minutes: 20
.buildkite/test_areas/model_executor.yaml
+1 -1
@@ -2,7 +2,7 @@ group: Model Executor
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Model Executor"
5 +- label: ":nvidia: (H200 MIG 35GB) Model Executor"
6 device: h200_35gb
7 key: model-executor
8 timeout_in_minutes: 60
.buildkite/test_areas/model_runner_v2.yaml
+3 -3
@@ -2,7 +2,7 @@ group: Model Runner V2
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Model Runner V2 Core"
5 +- label: ":nvidia: (H200 MIG 35GB) Model Runner V2 Core"
6 device: h200_35gb
7 key: model-runner-v2-core-tests
8 timeout_in_minutes: 35
@@ -24,7 +24,7 @@ steps:
24 # Temporary hack filter to exclude ngram spec decoding based tests.
25 - pytest -v -s entrypoints/llm/test_struct_output_generate.py -k "xgrammar and not speculative_config6 and not speculative_config7 and not speculative_config8 and not speculative_config0"
26
27 -- label: ":nvidia: (H200) Model Runner V2 Examples"
27 +- label: ":nvidia: (H200 MIG 35GB) Model Runner V2 Examples"
28 device: h200_35gb
29 key: model-runner-v2-examples
30 timeout_in_minutes: 35
@@ -126,7 +126,7 @@ steps:
126 - pytest -v -s distributed/test_pp_cudagraph.py -k "not ray"
127 - pytest -v -s v1/distributed/test_pp_dp_v2.py
128
129 -- label: ":nvidia: (H200) Model Runner V2 Spec Decode"
129 +- label: ":nvidia: (H200 MIG 35GB) Model Runner V2 Spec Decode"
130 device: h200_35gb
131 key: model-runner-v2-spec-decode
132 timeout_in_minutes: 50
.buildkite/test_areas/models_basic.yaml
+3 -3
@@ -2,7 +2,7 @@ group: Models - Basic
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Basic Models (Initialization)"
5 +- label: ":nvidia: (H200 MIG 18GB) Basic Models (Initialization)"
6 key: basic-models-tests-initialization
7 timeout_in_minutes: 25
8 device: h200_18gb
@@ -23,7 +23,7 @@ steps:
23 depends_on:
24 - image-build-amd
25
26 -- label: ":nvidia: (H200) Basic Models (Extra Initialization) Shard %N"
26 +- label: ":nvidia: (H200 MIG 35GB) Basic Models (Extra Initialization) Shard %N"
27 device: h200_35gb
28 key: basic-models-tests-extra-initialization
29 timeout_in_minutes: 100
@@ -38,7 +38,7 @@ steps:
38 - pytest -v -s models/test_initialization.py -k 'not test_can_initialize_small_subset' --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
39 parallelism: 4
40
41 -- label: ":nvidia: (H200) Basic Models (Other)"
41 +- label: ":nvidia: (H200 MIG 35GB) Basic Models (Other)"
42 device: h200_35gb
43 key: basic-models-tests-other
44 timeout_in_minutes: 35
.buildkite/test_areas/models_language.yaml
+7 -7
@@ -2,7 +2,7 @@ group: Models - Language
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Language Models (Standard)"
5 +- label: ":nvidia: (H200 MIG 18GB) Language Models (Standard)"
6 key: language-models-tests-standard
7 timeout_in_minutes: 30
8 device: h200_18gb
@@ -23,7 +23,7 @@ steps:
23 depends_on:
24 - image-build-amd
25
26 -- label: ":nvidia: (H200) Language Models (Extra Standard) Shard %N"
26 +- label: ":nvidia: (H200 MIG 35GB) Language Models (Extra Standard) Shard %N"
27 device: h200_35gb
28 key: language-models-tests-extra-standard
29 timeout_in_minutes: 40
@@ -57,7 +57,7 @@ steps:
57 - tests/models/language/pooling/test_classification.py
58 - vllm/_aiter_ops.py
59 - vllm/platforms/rocm.py
60 -- label: ":nvidia: (H200) Language Models (Hybrid) Shard %N"
60 +- label: ":nvidia: (H200 MIG 35GB) Language Models (Hybrid) Shard %N"
61 device: h200_35gb
62 key: language-models-tests-hybrid
63 timeout_in_minutes: 65
@@ -104,7 +104,7 @@ steps:
104 - CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
105 - pytest -v -s models/language/generation -m hybrid_model -k 'granite-4.0-tiny-preview'
106
107 -- label: ":nvidia: (H200) Language Models (Extended Generation)"
107 +- label: ":nvidia: (H200 MIG 35GB) Language Models (Extended Generation)"
108 device: h200_35gb
109 key: language-models-test-extended-generation
110 timeout_in_minutes: 80
@@ -133,7 +133,7 @@ steps:
133 - CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
134 - pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
135
136 -- label: ":nvidia: (H200) Language Models (PPL)"
136 +- label: ":nvidia: (H200 MIG 18GB) Language Models (PPL)"
137 key: language-models-test-ppl
138 timeout_in_minutes: 30
139 device: h200_18gb
@@ -171,7 +171,7 @@ steps:
171 - vllm/platforms/rocm.py
172 - tests/models/language/generation_ppl_test
173
174 -- label: ":nvidia: (H200) Language Models (Extended Pooling) Shard %N"
174 +- label: ":nvidia: (H200 MIG 35GB) Language Models (Extended Pooling) Shard %N"
175 device: h200_35gb
176 key: language-models-test-extended-pooling
177 timeout_in_minutes: 120
@@ -195,7 +195,7 @@ steps:
195 commands:
196 - pytest -v -s models/language/pooling -m 'not core_model' --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
197
198 -- label: ":nvidia: (H200) Language Models (MTEB)"
198 +- label: ":nvidia: (H200 MIG 18GB) Language Models (MTEB)"
199 key: language-models-test-mteb
200 timeout_in_minutes: 68
201 device: h200_18gb
.buildkite/test_areas/models_multimodal.yaml
+11 -11
@@ -2,7 +2,7 @@ group: Models - Multimodal
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Multimodal Models (Standard) 1: qwen2"
5 +- label: ":nvidia: (H200 MIG 18GB) Multimodal Models (Standard) 1: qwen2"
6 key: multi-modal-models-standard-1-qwen2
7 timeout_in_minutes: 68
8 device: h200_18gb
@@ -22,7 +22,7 @@ steps:
22 depends_on:
23 - image-build-amd
24
25 -- label: ":nvidia: (H200) Multimodal Models (Standard) 2: qwen3 + gemma"
25 +- label: ":nvidia: (H200 MIG 18GB) Multimodal Models (Standard) 2: qwen3 + gemma"
26 key: multi-modal-models-standard-2-qwen3-gemma
27 timeout_in_minutes: 75
28 device: h200_18gb
@@ -43,7 +43,7 @@ steps:
43 depends_on:
44 - image-build-amd
45
46 -- label: ":nvidia: (H200) Multimodal Models (Standard) 3: llava + qwen2_vl"
46 +- label: ":nvidia: (H200 MIG 35GB) Multimodal Models (Standard) 3: llava + qwen2_vl"
47 device: h200_35gb
48 key: multi-modal-models-standard-3-llava-qwen2-vl
49 timeout_in_minutes: 40
@@ -62,7 +62,7 @@ steps:
62 depends_on:
63 - image-build-amd
64
65 -- label: ":nvidia: (H200) Multimodal Models (Standard) 4: other + whisper"
65 +- label: ":nvidia: (H200 MIG 35GB) Multimodal Models (Standard) 4: other + whisper"
66 device: h200_35gb
67 key: multi-modal-models-standard-4-other-whisper
68 timeout_in_minutes: 75
@@ -99,7 +99,7 @@ steps:
99 - pytest -v -s models/multimodal/processing --ignore models/multimodal/processing/test_tensor_schema.py --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
100 parallelism: 4
101
102 -- label: ":nvidia: (H200) Multimodal Processor"
102 +- label: ":nvidia: (H200 MIG 18GB) Multimodal Processor"
103 key: multi-modal-processor
104 timeout_in_minutes: 98
105 device: h200_18gb
@@ -119,7 +119,7 @@ steps:
119 depends_on:
120 - image-build-amd
121
122 -- label: ":nvidia: (H200) Multimodal Accuracy Eval (Small Models)"
122 +- label: ":nvidia: (H200 MIG 35GB) Multimodal Accuracy Eval (Small Models)"
123 device: h200_35gb
124 key: multi-modal-accuracy-eval-small-models
125 timeout_in_minutes: 30
@@ -145,7 +145,7 @@ steps:
145 - vllm/platforms/rocm.py
146 - vllm/model_executor/model_loader/
147
148 -- label: ":nvidia: (H200) Multimodal Models (Extended Generation 1)"
148 +- label: ":nvidia: (H200 MIG 35GB) Multimodal Models (Extended Generation 1)"
149 device: h200_35gb
150 key: multi-modal-models-extended-generation-1
151 optional: true
@@ -166,7 +166,7 @@ steps:
166 depends_on:
167 - image-build-amd
168
169 -- label: ":nvidia: (H200) Multimodal Models (PPL)"
169 +- label: ":nvidia: (H200 MIG 35GB) Multimodal Models (PPL)"
170 device: h200_35gb
171 key: multi-modal-models-extended-ppl
172 optional: true
@@ -184,7 +184,7 @@ steps:
184 depends_on:
185 - image-build-amd
186
187 -- label: ":nvidia: (H200) Multimodal Models (Extended Generation 2) Shard %N"
187 +- label: ":nvidia: (H200 MIG 35GB) Multimodal Models (Extended Generation 2) Shard %N"
188 device: h200_35gb
189 key: multi-modal-models-extended-generation-2
190 parallelism: 4
@@ -204,7 +204,7 @@ steps:
204 depends_on:
205 - image-build-amd
206
207 -- label: ":nvidia: (H200) Multimodal Models (Extended Generation 3)"
207 +- label: ":nvidia: (H200 MIG 35GB) Multimodal Models (Extended Generation 3)"
208 device: h200_35gb
209 key: multi-modal-models-extended-generation-3
210 optional: true
@@ -223,7 +223,7 @@ steps:
223 depends_on:
224 - image-build-amd
225
226 -- label: ":nvidia: (H200) Multimodal Models (Extended Pooling)"
226 +- label: ":nvidia: (H200 MIG 18GB) Multimodal Models (Extended Pooling)"
227 key: multi-modal-models-extended-pooling
228 optional: true
229 device: h200_18gb
.buildkite/test_areas/plugins.yaml
+2 -2
@@ -62,7 +62,7 @@ steps:
62 - vllm/platforms/rocm.py
63
64
65 -- label: ":nvidia: (H200) GGUF Plugin"
65 +- label: ":nvidia: (H200 MIG 18GB) GGUF Plugin"
66 key: gguf-plugin
67 device: h200_18gb
68 timeout_in_minutes: 30
@@ -75,7 +75,7 @@ steps:
75 - pip install "vllm-gguf-plugin >= 0.0.2"
76 - pytest -v -s plugins_tests/gguf
77
78 -- label: ":nvidia: (H200) BitsAndBytes Plugin"
78 +- label: ":nvidia: (H200 MIG 18GB) BitsAndBytes Plugin"
79 key: bitsandbytes-plugin
80 device: h200_18gb
81 timeout_in_minutes: 30
.buildkite/test_areas/pytorch.yaml
+4 -4
@@ -2,7 +2,7 @@ group: PyTorch
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) PyTorch Compilation"
5 +- label: ":nvidia: (H200 MIG 35GB) PyTorch Compilation"
6 device: h200_35gb
7 key: pytorch-compilation-unit-tests
8 timeout_in_minutes: 60
@@ -54,7 +54,7 @@ steps:
54 - csrc/
55 - vllm/platforms/rocm.py
56
57 -- label: ":nvidia: (H200) PyTorch Compilation H100 Cases"
57 +- label: ":nvidia: (H200 MIG 18GB) PyTorch Compilation H100 Cases"
58 key: pytorch-compilation-unit-tests-h100
59 timeout_in_minutes: 30
60 device: h200_18gb
@@ -121,7 +121,7 @@ steps:
121 commands:
122 - pytest -s -v compile/passes --ignore compile/passes/distributed
123
124 -- label: ":nvidia: (H200) PyTorch Fullgraph"
124 +- label: ":nvidia: (H200 MIG 35GB) PyTorch Fullgraph"
125 device: h200_35gb
126 key: pytorch-fullgraph-test
127 timeout_in_minutes: 90
@@ -195,7 +195,7 @@ steps:
195 commands:
196 - pytest -s -v compile/fullgraph/test_full_cudagraph.py
197
198 -- label: ":nvidia: (H200) PyTorch Nightly Dependency Override Check"
198 +- label: ":nvidia: (H200 MIG 18GB) PyTorch Nightly Dependency Override Check"
199 key: pytorch-nightly-dependency-override-check
200 # if this test fails, it means the nightly torch version is not compatible with some
201 # of the dependencies. Please check the error message and add the package to whitelist
.buildkite/test_areas/quantization.yaml
+3 -3
@@ -2,7 +2,7 @@ group: Quantization
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Quantization Shard %N"
5 +- label: ":nvidia: (H200 MIG 35GB) Quantization Shard %N"
6 device: h200_35gb
7 key: quantization
8 timeout_in_minutes: 40
@@ -21,7 +21,7 @@ steps:
21 # parameter. It was not exercised by the previous L4 job.
22 - VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s quantization/ --ignore quantization/test_blackwell_moe.py -k 'not test_compressed_tensors_w4a8_fp8' --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
23
24 -- label: ":nvidia: (H200) Quantized Fusions"
24 +- label: ":nvidia: (H200 MIG 35GB) Quantized Fusions"
25 device: h200_35gb
26 key: quantized-fusions
27 timeout_in_minutes: 20
@@ -52,7 +52,7 @@ steps:
52 commands:
53 - pytest -s -v tests/quantization/test_blackwell_moe.py
54
55 -- label: ":nvidia: (H200) Quantized Models"
55 +- label: ":nvidia: (H200 MIG 35GB) Quantized Models"
56 device: h200_35gb
57 key: quantized-models-test
58 timeout_in_minutes: 65
.buildkite/test_areas/ray_compat.yaml
+1 -1
@@ -2,7 +2,7 @@ group: Ray Compatibility
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Ray Dependency Compatibility Check"
5 +- label: ":nvidia: (H200 MIG 18GB) Ray Dependency Compatibility Check"
6 key: ray-dependency-compatibility-check
7 # Informational only — does not block the pipeline.
8 # If this fails, it means the PR introduces a dependency that
.buildkite/test_areas/rust_frontend.yaml
+4 -4
@@ -2,7 +2,7 @@ group: Rust Frontend E2E
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Rust Frontend OpenAI Coverage"
5 +- label: ":nvidia: (H200 MIG 18GB) Rust Frontend OpenAI Coverage"
6 timeout_in_minutes: 30
7 device: h200_18gb
8 working_dir: "/vllm-workspace/tests"
@@ -60,7 +60,7 @@ steps:
60 - tests/v1/sample/test_logprobs_e2e.py
61 - vllm/platforms/rocm.py
62
63 -- label: ":nvidia: (H200) Rust Frontend Serve/Admin Coverage"
63 +- label: ":nvidia: (H200 MIG 18GB) Rust Frontend Serve/Admin Coverage"
64 timeout_in_minutes: 25
65 device: h200_18gb
66 working_dir: "/vllm-workspace/tests"
@@ -109,7 +109,7 @@ steps:
109 - tests/entrypoints/serve/tokenize/test_tokenization.py
110 - vllm/platforms/rocm.py
111
112 -- label: ":nvidia: (H200) Rust Frontend Core Correctness"
112 +- label: ":nvidia: (H200 MIG 18GB) Rust Frontend Core Correctness"
113 timeout_in_minutes: 20
114 device: h200_18gb
115 working_dir: "/vllm-workspace/tests"
@@ -123,7 +123,7 @@ steps:
123 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
124 - pytest -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine
125
126 -- label: ":nvidia: (H200) Rust Frontend Tool Use"
126 +- label: ":nvidia: (H200 MIG 35GB) Rust Frontend Tool Use"
127 device: h200_35gb
128 timeout_in_minutes: 25
129 working_dir: "/vllm-workspace/tests"
.buildkite/test_areas/samplers.yaml
+1 -1
@@ -2,7 +2,7 @@ group: Samplers
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) Samplers"
5 +- label: ":nvidia: (H200 MIG 35GB) Samplers"
6 device: h200_35gb
7 key: samplers-test
8 timeout_in_minutes: 40
.buildkite/test_areas/spec_decode.yaml
+9 -9
@@ -2,7 +2,7 @@ group: Spec Decode
2 depends_on:
3 - image-build
4 steps:
5 -- label: ":nvidia: (H200) V1 Spec Decode"
5 +- label: ":nvidia: (H200 MIG 35GB) V1 Spec Decode"
6 device: h200_35gb
7 key: v1-spec-decode
8 timeout_in_minutes: 40
@@ -30,7 +30,7 @@ steps:
30 depends_on:
31 - image-build-amd
32
33 -- label: ":nvidia: (H200) Spec Decode Eagle 1: DeepSeek + Qwen"
33 +- label: ":nvidia: (H200 MIG 35GB) Spec Decode Eagle 1: DeepSeek + Qwen"
34 key: spec-decode-eagle-1-deepseek-qwen
35 timeout_in_minutes: 25
36 device: h200_35gb
@@ -57,7 +57,7 @@ steps:
57 - tests/v1/e2e/spec_decode/
58 - vllm/platforms/rocm.py
59
60 -- label: ":nvidia: (H200) Spec Decode Eagle 2: Llama 3 + Qwen VL + Other"
60 +- label: ":nvidia: (H200 MIG 35GB) Spec Decode Eagle 2: Llama 3 + Qwen VL + Other"
61 key: spec-decode-eagle-2-llama3-qwen-vl-other
62 timeout_in_minutes: 25
63 device: h200_35gb
@@ -96,7 +96,7 @@ steps:
96 commands:
97 - pytest -v -s v1/e2e/spec_decode/eagle/
98
99 -- label: ":nvidia: (H200) Spec Decode Speculators + MTP"
99 +- label: ":nvidia: (H200 MIG 35GB) Spec Decode Speculators + MTP"
100 key: spec-decode-speculators-mtp
101 timeout_in_minutes: 50
102 device: h200_35gb
@@ -141,7 +141,7 @@ steps:
141 - pytest -v -s v1/e2e/spec_decode/speculators/
142 - pytest -v -s v1/e2e/spec_decode/mtp/
143
144 -- label: ":nvidia: (H200) Spec Decode N-Gram + Suffix"
144 +- label: ":nvidia: (H200 MIG 35GB) Spec Decode N-Gram + Suffix"
145 key: spec-decode-ngram-suffix
146 timeout_in_minutes: 20
147 device: h200_35gb
@@ -170,7 +170,7 @@ steps:
170 - tests/v1/e2e/spec_decode/
171 - vllm/platforms/rocm.py
172
173 -- label: ":nvidia: (H200) Spec Decode Draft Model"
173 +- label: ":nvidia: (H200 MIG 18GB) Spec Decode Draft Model"
174 key: spec-decode-draft-model
175 timeout_in_minutes: 45
176 device: h200_18gb
@@ -269,7 +269,7 @@ steps:
269 - tests/v1/e2e/spec_decode/test_mtp_parallel_load.py
270 - vllm/platforms/rocm.py
271
272 -- label: ":nvidia: (H200) Spec Decode AL DFlash Nightly"
272 +- label: ":nvidia: (H200 MIG 35GB) Spec Decode AL DFlash Nightly"
273 key: spec-decode-dflash-nightly
274 timeout_in_minutes: 90
275 device: h200_35gb
@@ -305,7 +305,7 @@ steps:
305 - tests/v1/e2e/spec_decode/
306 - vllm/platforms/rocm.py
307
308 -- label: ":nvidia: (H200) Spec Decode AL DSpark Nightly"
308 +- label: ":nvidia: (H200 MIG 35GB) Spec Decode AL DSpark Nightly"
309 key: spec-decode-dspark-nightly
310 timeout_in_minutes: 60
311 device: h200_35gb
@@ -345,7 +345,7 @@ steps:
345 - tests/v1/e2e/spec_decode/
346 - vllm/platforms/rocm.py
347
348 -- label: ":nvidia: (H200) Spec Decode AL MTP + Other Acceptance Nightly"
348 +- label: ":nvidia: (H200 MIG 35GB) Spec Decode AL MTP + Other Acceptance Nightly"
349 key: spec-decode-mtp-other-acceptance-nightly
350 timeout_in_minutes: 60
351 device: h200_35gb