Skip to content

Commit ac86bff

Browse files
njhillnjhill
authored
Revert "[Core] Performance: Use list[np.ndarray] instead of list[list… (#28773)
1 parent edfe498 commit ac86bff

File tree

12 files changed

+76
-102
lines changed

12 files changed

+76
-102
lines changed

‎tests/v1/core/test_async_scheduler.py‎

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -2,7 +2,6 @@
22
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
33
from collections import deque
44

5-
import numpy as np
65
import pytest
76

87
from vllm.v1.core.sched.output import SchedulerOutput
@@ -22,7 +21,7 @@ def _make_model_runner_output(
2221
return ModelRunnerOutput(
2322
req_ids=req_ids,
2423
req_id_to_index={req_id: i for i, req_id in enumerate(req_ids)},
25-
sampled_token_ids=[np.array([i]) for i in range(len(req_ids))],
24+
sampled_token_ids=[[i] for i in range(len(req_ids))],
2625
logprobs=None,
2726
prompt_logprobs_dict={},
2827
pooler_output=[],

‎tests/v1/core/test_scheduler.py‎

Lines changed: 32 additions & 44 deletions
Original file line numberDiff line numberDiff line change
@@ -3,7 +3,6 @@
33
import dataclasses
44
from unittest.mock import Mock
55

6-
import numpy as np
76
import pytest
87
import torch
98

@@ -170,7 +169,7 @@ def test_schedule_partial_requests():
170169
req_id_to_index=req_to_index,
171170
# Only the first request has a sampled token id because
172171
# the rest requests are still being prefilled.
173-
sampled_token_ids=[np.array([0]), np.array([]), np.array([])],
172+
sampled_token_ids=[[0], [], []],
174173
logprobs=None,
175174
prompt_logprobs_dict={},
176175
pooler_output=[],
@@ -217,7 +216,7 @@ def test_no_mm_input_chunking():
217216
model_runner_output = ModelRunnerOutput(
218217
req_ids=[request.request_id for request in requests],
219218
req_id_to_index=req_to_index,
220-
sampled_token_ids=[np.array([]) for _ in range(len(requests))],
219+
sampled_token_ids=[[] for _ in range(len(requests))],
221220
logprobs=None,
222221
prompt_logprobs_dict={},
223222
pooler_output=[],
@@ -277,7 +276,7 @@ def test_schedule_concurrent_partial_requests(enable_prefix_caching: bool):
277276
model_runner_output = ModelRunnerOutput(
278277
req_ids=[request.request_id for request in requests],
279278
req_id_to_index=req_to_index,
280-
sampled_token_ids=[np.array([]) for _ in range(len(requests))],
279+
sampled_token_ids=[[] for _ in range(len(requests))],
281280
logprobs=None,
282281
prompt_logprobs_dict={},
283282
pooler_output=[],
@@ -301,8 +300,7 @@ def test_schedule_concurrent_partial_requests(enable_prefix_caching: bool):
301300
model_runner_output = ModelRunnerOutput(
302301
req_ids=[request.request_id for request in requests],
303302
req_id_to_index=req_to_index,
304-
sampled_token_ids=[np.array([0]), np.array([0])]
305-
+ [np.array([]) for _ in range(len(requests) - 2)],
303+
sampled_token_ids=[[0], [0]] + [[] for _ in range(len(requests) - 2)],
306304
logprobs=None,
307305
prompt_logprobs_dict={},
308306
pooler_output=[],
@@ -349,8 +347,8 @@ def test_stop_via_update_from_output():
349347
req_ids=[req.request_id for req in requests],
350348
req_id_to_index={req.request_id: i for i, req in enumerate(requests)},
351349
sampled_token_ids=[
352-
np.array([EOS_TOKEN_ID]),
353-
np.array([10, 11]),
350+
[EOS_TOKEN_ID],
351+
[10, 11],
354352
], # First request hits EOS, second continues
355353
logprobs=None,
356354
prompt_logprobs_dict={},
@@ -394,10 +392,7 @@ def test_stop_via_update_from_output():
394392
model_output = ModelRunnerOutput(
395393
req_ids=[req.request_id for req in requests],
396394
req_id_to_index={req.request_id: i for i, req in enumerate(requests)},
397-
sampled_token_ids=[
398-
np.array([10, 42, 12]),
399-
np.array([13, 14]),
400-
], # First request hits stop token
395+
sampled_token_ids=[[10, 42, 12], [13, 14]], # First request hits stop token
401396
logprobs=None,
402397
prompt_logprobs_dict={},
403398
pooler_output=[],
@@ -441,10 +436,7 @@ def test_stop_via_update_from_output():
441436
model_output = ModelRunnerOutput(
442437
req_ids=[req.request_id for req in requests],
443438
req_id_to_index={req.request_id: i for i, req in enumerate(requests)},
444-
sampled_token_ids=[
445-
np.array([10, 11, 12]),
446-
np.array([13]),
447-
], # First request exceeds max_tokens
439+
sampled_token_ids=[[10, 11, 12], [13]], # First request exceeds max_tokens
448440
logprobs=None,
449441
prompt_logprobs_dict={},
450442
pooler_output=[],
@@ -483,7 +475,7 @@ def test_stop_via_update_from_output():
483475
model_output = ModelRunnerOutput(
484476
req_ids=[requests[0].request_id],
485477
req_id_to_index={requests[0].request_id: 0},
486-
sampled_token_ids=[np.array([EOS_TOKEN_ID, 10, 11])],
478+
sampled_token_ids=[[EOS_TOKEN_ID, 10, 11]],
487479
logprobs=None,
488480
prompt_logprobs_dict={},
489481
pooler_output=[],
@@ -624,7 +616,7 @@ def test_schedule_concurrent_batches(
624616
model_runner_output = ModelRunnerOutput(
625617
req_ids=[requests[0].request_id],
626618
req_id_to_index={requests[0].request_id: 0},
627-
sampled_token_ids=[np.array([0])],
619+
sampled_token_ids=[[0]],
628620
logprobs=None,
629621
prompt_logprobs_dict={},
630622
pooler_output=[],
@@ -641,7 +633,7 @@ def test_schedule_concurrent_batches(
641633
model_runner_output = ModelRunnerOutput(
642634
req_ids=[requests[1].request_id],
643635
req_id_to_index={requests[1].request_id: 0},
644-
sampled_token_ids=[np.array([0])],
636+
sampled_token_ids=[[0]],
645637
logprobs=None,
646638
prompt_logprobs_dict={},
647639
pooler_output=[],
@@ -678,7 +670,7 @@ def test_preempt_during_execution():
678670
model_runner_output0 = ModelRunnerOutput(
679671
req_ids=[requests[0].request_id],
680672
req_id_to_index={requests[0].request_id: 0},
681-
sampled_token_ids=[np.array([0])],
673+
sampled_token_ids=[[0]],
682674
logprobs=None,
683675
prompt_logprobs_dict={},
684676
pooler_output=[],
@@ -695,7 +687,7 @@ def test_preempt_during_execution():
695687
model_runner_output1 = ModelRunnerOutput(
696688
req_ids=[requests[1].request_id],
697689
req_id_to_index={requests[1].request_id: 0},
698-
sampled_token_ids=[np.array([42])],
690+
sampled_token_ids=[[42]],
699691
logprobs=None,
700692
prompt_logprobs_dict={},
701693
pooler_output=[],
@@ -712,18 +704,14 @@ def test_preempt_during_execution():
712704
@pytest.mark.parametrize(
713705
"spec_tokens,output_tokens,expected",
714706
[
715-
([[1, 2, 3]], [np.array([1, 2, 3, 4])], (1, 3, 3, [1, 1, 1])), # perfect match
716-
([[1, 2, 3]], [np.array([1, 5])], (1, 3, 1, [1, 0, 0])), # early mismatch
717-
(
718-
[[1, 2], [3]],
719-
[np.array([1, 2, 5]), np.array([3, 4])],
720-
(2, 3, 3, [2, 1]),
721-
), # multiple sequences
722-
([[1]], [np.array([1, 2])], (1, 1, 1, [1])), # single token sequence
723-
([[]], [np.array([5])], (0, 0, 0, [0])), # empty sequence
707+
([[1, 2, 3]], [[1, 2, 3, 4]], (1, 3, 3, [1, 1, 1])), # perfect match
708+
([[1, 2, 3]], [[1, 5]], (1, 3, 1, [1, 0, 0])), # early mismatch
709+
([[1, 2], [3]], [[1, 2, 5], [3, 4]], (2, 3, 3, [2, 1])), # multiple sequences
710+
([[1]], [[1, 2]], (1, 1, 1, [1])), # single token sequence
711+
([[]], [[5]], (0, 0, 0, [0])), # empty sequence
724712
(
725713
[[1, 2, 3], [4, 5, 6]],
726-
[np.array([1, 2, 7]), np.array([4, 8])],
714+
[[1, 2, 7], [4, 8]],
727715
(2, 6, 3, [2, 1, 0]),
728716
), # multiple mismatches
729717
],
@@ -757,7 +745,7 @@ def test_schedule_spec_decoding_stats(spec_tokens, output_tokens, expected):
757745
model_runner_output = ModelRunnerOutput(
758746
req_ids=req_ids,
759747
req_id_to_index=req_to_index,
760-
sampled_token_ids=[np.array([0]) for _ in range(len(requests))],
748+
sampled_token_ids=[[0] for _ in range(len(requests))],
761749
logprobs=None,
762750
prompt_logprobs_dict={},
763751
pooler_output=[],
@@ -984,7 +972,7 @@ def test_kv_connector_basic(is_async: bool):
984972
MODEL_RUNNER_OUTPUT = ModelRunnerOutput(
985973
req_ids=req_ids,
986974
req_id_to_index=req_to_index,
987-
sampled_token_ids=[np.array([1000])] * len(req_ids),
975+
sampled_token_ids=[[1000]] * len(req_ids),
988976
logprobs=None,
989977
prompt_logprobs_dict={},
990978
pooler_output=[],
@@ -1037,7 +1025,7 @@ def test_kv_connector_basic(is_async: bool):
10371025
MODEL_RUNNER_OUTPUT = ModelRunnerOutput(
10381026
req_ids=req_ids,
10391027
req_id_to_index=req_to_index,
1040-
sampled_token_ids=[np.array([1000])] * len(req_ids),
1028+
sampled_token_ids=[[1000]] * len(req_ids),
10411029
logprobs=None,
10421030
prompt_logprobs_dict={},
10431031
pooler_output=[],
@@ -1100,7 +1088,7 @@ def test_external_prefix_cache_metrics():
11001088
MODEL_RUNNER_OUTPUT = ModelRunnerOutput(
11011089
req_ids=[r.request_id for r in requests],
11021090
req_id_to_index={r.request_id: i for i, r in enumerate(requests)},
1103-
sampled_token_ids=[np.array([1000])] * NUM_REQUESTS,
1091+
sampled_token_ids=[[1000]] * NUM_REQUESTS,
11041092
logprobs=None,
11051093
prompt_logprobs_dict={},
11061094
pooler_output=[],
@@ -1166,7 +1154,7 @@ def test_kv_connector_unable_to_allocate(use_ec_connector, ec_role):
11661154
MODEL_RUNNER_OUTPUT = ModelRunnerOutput(
11671155
req_ids=req_ids,
11681156
req_id_to_index=req_to_index,
1169-
sampled_token_ids=[np.array([1000])] * len(req_ids),
1157+
sampled_token_ids=[[1000]] * len(req_ids),
11701158
logprobs=None,
11711159
prompt_logprobs_dict={},
11721160
pooler_output=[],
@@ -1251,7 +1239,7 @@ def test_kv_connector_handles_preemption(use_ec_connector, ec_role):
12511239
MODEL_RUNNER_OUTPUT = ModelRunnerOutput(
12521240
req_ids=req_ids,
12531241
req_id_to_index=req_to_index,
1254-
sampled_token_ids=[np.array([1000])] * len(req_ids),
1242+
sampled_token_ids=[[1000]] * len(req_ids),
12551243
logprobs=None,
12561244
prompt_logprobs_dict={},
12571245
pooler_output=[],
@@ -1344,7 +1332,7 @@ def make_output(scheduler: Scheduler):
13441332
return ModelRunnerOutput(
13451333
req_ids=[req.request_id for req in scheduler.running],
13461334
req_id_to_index={req.request_id: i for i, req in enumerate(scheduler.running)},
1347-
sampled_token_ids=[np.array([1000])] * len(scheduler.running),
1335+
sampled_token_ids=[[1000]] * len(scheduler.running),
13481336
logprobs=None,
13491337
prompt_logprobs_dict={},
13501338
pooler_output=[],
@@ -1761,7 +1749,7 @@ def test_priority_scheduling_preemption():
17611749
req_id_to_index={
17621750
req.request_id: i for i, req in enumerate(low_priority_requests)
17631751
},
1764-
sampled_token_ids=[np.array([100]) for _ in low_priority_requests],
1752+
sampled_token_ids=[[100] for _ in low_priority_requests],
17651753
logprobs=None,
17661754
prompt_logprobs_dict={},
17671755
pooler_output=[],
@@ -1830,7 +1818,7 @@ def test_priority_scheduling_no_preemption_when_space_available():
18301818
req_id_to_index={
18311819
req.request_id: i for i, req in enumerate(low_priority_requests)
18321820
},
1833-
sampled_token_ids=[np.array([100]) for _ in low_priority_requests],
1821+
sampled_token_ids=[[100] for _ in low_priority_requests],
18341822
logprobs=None,
18351823
prompt_logprobs_dict={},
18361824
pooler_output=[],
@@ -2076,7 +2064,7 @@ def test_priority_scheduling_heap_property():
20762064
model_output = ModelRunnerOutput(
20772065
req_ids=[req.req_id],
20782066
req_id_to_index={req.req_id: 0},
2079-
sampled_token_ids=[np.array([100])],
2067+
sampled_token_ids=[[100]],
20802068
logprobs=None,
20812069
prompt_logprobs_dict={},
20822070
pooler_output=[],
@@ -2162,7 +2150,7 @@ def test_priority_scheduling_preemption_and_resumption_when_out_of_kv(
21622150
model_output = ModelRunnerOutput(
21632151
req_ids=[request_low.request_id],
21642152
req_id_to_index={request_low.request_id: 0},
2165-
sampled_token_ids=[np.array([100])],
2153+
sampled_token_ids=[[100]],
21662154
# spec_token_ids=None,
21672155
logprobs=None,
21682156
prompt_logprobs_dict={},
@@ -2193,7 +2181,7 @@ def test_priority_scheduling_preemption_and_resumption_when_out_of_kv(
21932181
model_output = ModelRunnerOutput(
21942182
req_ids=[req.request_id for req in requests],
21952183
req_id_to_index={req.request_id: i for i, req in enumerate(requests)},
2196-
sampled_token_ids=[np.array([100]) for _ in requests],
2184+
sampled_token_ids=[[100] for _ in requests],
21972185
# spec_token_ids=None,
21982186
logprobs=None,
21992187
prompt_logprobs_dict={},
@@ -2219,7 +2207,7 @@ def test_priority_scheduling_preemption_and_resumption_when_out_of_kv(
22192207
model_output = ModelRunnerOutput(
22202208
req_ids=[req.request_id for req in requests],
22212209
req_id_to_index={req.request_id: i for i, req in enumerate(requests)},
2222-
sampled_token_ids=[np.array([]), np.array([100])],
2210+
sampled_token_ids=[[], [100]],
22232211
# spec_token_ids=None,
22242212
logprobs=None,
22252213
prompt_logprobs_dict={},

‎tests/v1/kv_connector/unit/utils.py‎

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,6 @@
77
from itertools import chain, count
88
from typing import Any
99

10-
import numpy as np
1110
import torch
1211

1312
from vllm import SamplingParams
@@ -229,7 +228,7 @@ def create_model_runner_output(
229228

230229
# Make sampled tokens.
231230
sampled_token = EOS_TOKEN_ID if use_eos else token_id
232-
sampled_token_ids = [np.array([sampled_token]) for _ in req_ids]
231+
sampled_token_ids = [[sampled_token] for _ in req_ids]
233232

234233
kv_connector_output = (
235234
None

‎tests/v1/spec_decode/test_eagle.py‎

Lines changed: 1 addition & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -3,7 +3,6 @@
33

44
from unittest import mock
55

6-
import numpy as np
76
import pytest
87
import torch
98

@@ -113,9 +112,7 @@ def test_prepare_next_token_ids():
113112
sampled_token_ids_tensor = torch.tensor(
114113
sampled_token_ids, dtype=torch.int32, device=device
115114
)
116-
sampled_token_ids_cpu = [
117-
np.array([i for i in seq if i != -1]) for seq in sampled_token_ids
118-
]
115+
sampled_token_ids_cpu = [[i for i in seq if i != -1] for seq in sampled_token_ids]
119116

120117
expected_next_token_ids_cpu = [1, 4, 30, 40]
121118
expected_next_token_ids_tensor = torch.tensor(

0 commit comments

Comments
 (0)