[CI] Shard (H200 MIG 35GB / MI355 DPX) Entrypoints Integration (Pooling) into named jobs (#58652)
Signed-off-by: Thang Nguyen <thangnguyenvn647@gmail.com> Signed-off-by: khluu <khluu000@gmail.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: khluu <khluu000@gmail.com>
Thang Nguyen committed
Sep 28, 2026 at 03:21 UTC
413ae4b1677bd34398c63840c3f29c0b35bf3da9
22 files changed
+77
-9
.buildkite/test_areas/entrypoints.yaml
+74
-6
@@ -230,10 +230,10 @@ steps:
230
depends_on:
231
- image-build-amd
232
233
-- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (Pooling)"
233
+- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (Pooling Scoring Cross Encoder)"
234
device: h200_35gb
235
- key: entrypoints-integration-pooling
236
- timeout_in_minutes: 75
235
+ key: entrypoints-integration-pooling-scoring-cross-encoder
236
+ timeout_in_minutes: 25
237
working_dir: "/vllm-workspace/tests"
238
source_file_dependencies:
239
- vllm/
@@ -241,13 +241,81 @@ steps:
241
- tests/entrypoints/pooling
242
commands:
243
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
244
- - pytest -v -s entrypoints/pooling
244
+ - pytest -v -s entrypoints/pooling/scoring/cross_encoder
245
mirror:
246
amd:
247
- label: ":amd: (MI355 DPX) Entrypoints Integration (Pooling)"
247
+ label: ":amd: (MI355 DPX) Entrypoints Integration (Pooling Scoring Cross Encoder)"
248
dind: false
249
device: mi355_dpx
250
- timeout_in_minutes: 65
250
+ timeout_in_minutes: 30
251
+ working_dir: "/vllm-workspace/tests"
252
+ depends_on:
253
+ - image-build-amd
254
+
255
+- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (Pooling Scoring Bi-Encoder and Late Interaction)"
256
+ device: h200_35gb
257
+ key: entrypoints-integration-pooling-scoring-bi-encoder-late-interaction
258
+ timeout_in_minutes: 25
259
+ working_dir: "/vllm-workspace/tests"
260
+ source_file_dependencies:
261
+ - vllm/
262
+ - "!vllm/distributed/kv_transfer/"
263
+ - tests/entrypoints/pooling
264
+ commands:
265
+ - export VLLM_WORKER_MULTIPROC_METHOD=spawn
266
+ - pytest -v -s entrypoints/pooling/scoring/bi_encoder entrypoints/pooling/scoring/late_interaction entrypoints/pooling/scoring/unit
267
+ mirror:
268
+ amd:
269
+ label: ":amd: (MI355 DPX) Entrypoints Integration (Pooling Scoring Bi-Encoder and Late Interaction)"
270
+ dind: false
271
+ device: mi355_dpx
272
+ timeout_in_minutes: 30
273
+ working_dir: "/vllm-workspace/tests"
274
+ depends_on:
275
+ - image-build-amd
276
+
277
+- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (Pooling Embed)"
278
+ device: h200_35gb
279
+ key: entrypoints-integration-pooling-embed
280
+ timeout_in_minutes: 35
281
+ working_dir: "/vllm-workspace/tests"
282
+ source_file_dependencies:
283
+ - vllm/
284
+ - "!vllm/distributed/kv_transfer/"
285
+ - tests/entrypoints/pooling
286
+ commands:
287
+ - export VLLM_WORKER_MULTIPROC_METHOD=spawn
288
+ - pytest -v -s entrypoints/pooling/embed
289
+ mirror:
290
+ amd:
291
+ label: ":amd: (MI355 DPX) Entrypoints Integration (Pooling Embed)"
292
+ dind: false
293
+ device: mi355_dpx
294
+ timeout_in_minutes: 35
295
+ working_dir: "/vllm-workspace/tests"
296
+ depends_on:
297
+ - image-build-amd
298
+
299
+- label: ":nvidia: (H200 MIG 35GB) Entrypoints Integration (Pooling Basic, Classify, Reward and Token)"
300
+ device: h200_35gb
301
+ key: entrypoints-integration-pooling-basic-classify-reward-token
302
+ timeout_in_minutes: 35
303
+ working_dir: "/vllm-workspace/tests"
304
+ source_file_dependencies:
305
+ - vllm/
306
+ - "!vllm/distributed/kv_transfer/"
307
+ - tests/entrypoints/pooling
308
+ commands:
309
+ - export VLLM_WORKER_MULTIPROC_METHOD=spawn
310
+ - pytest -v -s entrypoints/pooling/basic entrypoints/pooling/classify entrypoints/pooling/reward entrypoints/pooling/token_classify entrypoints/pooling/token_embed
311
+ - "find entrypoints/pooling -maxdepth 1 -name 'test_*.py' | xargs -r pytest -v -s"
312
+ - "find entrypoints/pooling/scoring -maxdepth 1 -name 'test_*.py' | xargs -r pytest -v -s"
313
+ mirror:
314
+ amd:
315
+ label: ":amd: (MI355 DPX) Entrypoints Integration (Pooling Basic, Classify, Reward and Token)"
316
+ dind: false
317
+ device: mi355_dpx
318
+ timeout_in_minutes: 35
319
working_dir: "/vllm-workspace/tests"
320
depends_on:
321
- image-build-amd
tests/entrypoints/pooling/basic/test_factories.py
renamed
tests/entrypoints/pooling/basic/test_io_processor.py
renamed
tests/entrypoints/pooling/basic/test_utils.py
renamed
tests/entrypoints/pooling/scoring/bi_encoder/__init__.py
tests/entrypoints/pooling/scoring/bi_encoder/test_bi_encoder_offline.py
renamed
tests/entrypoints/pooling/scoring/bi_encoder/test_bi_encoder_online.py
renamed
tests/entrypoints/pooling/scoring/cross_encoder/__init__.py
tests/entrypoints/pooling/scoring/cross_encoder/test_cross_encoder_correctness_mteb.py
renamed
tests/entrypoints/pooling/scoring/cross_encoder/test_cross_encoder_offline.py
renamed
tests/entrypoints/pooling/scoring/cross_encoder/test_cross_encoder_online.py
renamed
tests/entrypoints/pooling/scoring/cross_encoder/test_cross_encoder_online_vision.py
renamed
tests/entrypoints/pooling/scoring/late_interaction/__init__.py
tests/entrypoints/pooling/scoring/late_interaction/test_late_interaction_offline.py
renamed
+1
-1
@@ -5,7 +5,7 @@ import weakref
5
6
import pytest
7
8
-from .util import ColBERTScoringHfRunner
8
+from ..util import ColBERTScoringHfRunner
9
10
MODEL_NAME = "answerdotai/answerai-colbert-small-v1"
11
COLBERT_DIM = 96
tests/entrypoints/pooling/scoring/late_interaction/test_late_interaction_offline_vision.py
renamed
+1
-1
@@ -5,7 +5,7 @@ import weakref
5
6
import pytest
7
8
-from .util import make_base64_image, make_image_mm_param
8
+from ..util import make_base64_image, make_image_mm_param
9
10
MODEL_NAME = "vidore/colpali-v1.3-hf"
11
tests/entrypoints/pooling/scoring/late_interaction/test_late_interaction_online.py
renamed
+1
-1
@@ -8,7 +8,7 @@ import requests
8
from tests.utils import RemoteOpenAIServer
9
from vllm.entrypoints.pooling.scoring.protocol import RerankResponse, ScoreResponse
10
11
-from .util import ColBERTScoringHfRunner
11
+from ..util import ColBERTScoringHfRunner
12
13
MODEL_NAME = "answerdotai/answerai-colbert-small-v1"
14
COLBERT_DIM = 96