Update deprecated Python 3.8 typing (#13971)

This commit is contained in:
Harry Mellor
2025-03-02 17:34:51 -08:00
committed by GitHub
parent bf33700ecd
commit cf069aa8aa
300 changed files with 2294 additions and 2347 deletions
+2 -1
View File
@@ -1,6 +1,7 @@
# SPDX-License-Identifier: Apache-2.0
from typing import Callable, Iterable, Optional
from collections.abc import Iterable
from typing import Callable, Optional
import pytest
@@ -1,7 +1,6 @@
# SPDX-License-Identifier: Apache-2.0
import random
from typing import List
import pytest
@@ -137,9 +136,9 @@ def prep_prompts(batch_size: int):
The prompt is just under 10k tokens; sliding window is 4k
so the answer is outside sliding window, but should still be correct.
"""
prompts: List[str] = []
answer: List[int] = []
indices: List[int] = []
prompts: list[str] = []
answer: list[int] = []
indices: list[int] = []
random.seed(1)
for _ in range(batch_size):
idx = random.randint(30, 90)
@@ -158,7 +157,7 @@ def prep_prompts(batch_size: int):
return prompts, answer, indices
def check_answers(indices: List[int], answer: List[int], outputs: List[str]):
def check_answers(indices: list[int], answer: list[int], outputs: list[str]):
answer2 = [int(text[0:2].strip()) for text in outputs]
print(list(zip(indices, zip(answer, answer2))))
numok = 0
@@ -170,7 +169,7 @@ def check_answers(indices: List[int], answer: List[int], outputs: List[str]):
assert frac_ok > 0.7
def check_window(prompts: List[str]):
def check_window(prompts: list[str]):
def inner(llm: LLM):
sliding_window = llm.llm_engine.model_config.get_sliding_window()
+3 -5
View File
@@ -1,7 +1,5 @@
# SPDX-License-Identifier: Apache-2.0
from typing import List
import pytest
from vllm.core.block.block_table import BlockTable
@@ -32,7 +30,7 @@ def test_allocate_naive(block_size: int, sequence_len: int):
token_ids = list(range(sequence_len))
num_blocks_per_alloc = len(list(chunk_list(token_ids, block_size)))
block_tables: List[BlockTable] = []
block_tables: list[BlockTable] = []
for i in range(5):
assert allocator.get_num_free_blocks(
device=Device.GPU) == num_gpu_blocks - i * num_blocks_per_alloc
@@ -77,7 +75,7 @@ def test_allocate_prefix_caching(block_size: int, sequence_len: int):
num_immutable_blocks_per_alloc = len(
chunked_tokens) - num_mutable_blocks_per_alloc
block_tables: List[BlockTable] = []
block_tables: list[BlockTable] = []
for alloc_i in range(1, 6):
block_tables.append(
@@ -272,7 +270,7 @@ def test_append_token_ids_correct_content(block_size: int, sequence_len: int,
)
block_table.allocate(token_ids=token_ids, device=Device.GPU)
appended_so_far: List[int] = []
appended_so_far: list[int] = []
for append in chunk_list(token_ids_to_append, append_size):
block_table.append_token_ids(append)
appended_so_far.extend(append)
+2 -2
View File
@@ -1,6 +1,6 @@
# SPDX-License-Identifier: Apache-2.0
from typing import List, Optional
from typing import Optional
import pytest
@@ -14,7 +14,7 @@ class TestNaiveBlockAllocator:
def create_allocate_lambda(allocate_type: str,
allocator: NaiveBlockAllocator,
prev_block: Optional[Block],
token_ids: List[int]):
token_ids: list[int]):
if allocate_type == "immutable":
allocate_block = lambda: allocator.allocate_immutable_block(
prev_block=prev_block, token_ids=token_ids)
@@ -2,7 +2,7 @@
import math
import random
from typing import List, Optional
from typing import Optional
from unittest.mock import MagicMock
import pytest
@@ -123,11 +123,11 @@ class TestPrefixCachingBlock:
@staticmethod
def create_chain(block_size: int,
token_ids: List[int],
num_empty_trailing_blocks=0) -> List[PrefixCachingBlock]:
token_ids: list[int],
num_empty_trailing_blocks=0) -> list[PrefixCachingBlock]:
"""Helper method which creates a chain of blocks.
"""
blocks: List[PrefixCachingBlock] = []
blocks: list[PrefixCachingBlock] = []
num_blocks = math.ceil(
len(token_ids) / block_size) + num_empty_trailing_blocks
@@ -161,7 +161,7 @@ class TestPrefixCachingBlockAllocator:
@staticmethod
def create_allocate_lambda(allocate_type: str, allocator: BlockAllocator,
prev_block: Optional[Block],
token_ids: List[int]):
token_ids: list[int]):
if allocate_type == "immutable":
allocate_block = lambda: allocator.allocate_immutable_block(
prev_block=prev_block, token_ids=token_ids)
@@ -839,13 +839,13 @@ class TestPrefixCachingBlockAllocator:
@staticmethod
def create_immutable_chain(
block_size: int,
token_ids: List[int],
token_ids: list[int],
allocator: PrefixCachingBlockAllocator,
extra_hash: Optional[int] = None,
) -> List[PrefixCachingBlock]:
) -> list[PrefixCachingBlock]:
"""Helper method which creates a chain of blocks.
"""
blocks: List[Block] = []
blocks: list[Block] = []
num_blocks = math.ceil(len(token_ids) / block_size)
if num_blocks == 0:
+12 -13
View File
@@ -1,6 +1,5 @@
# SPDX-License-Identifier: Apache-2.0
from typing import List
from unittest.mock import MagicMock
import pytest # noqa
@@ -46,7 +45,7 @@ def test_simple():
cache_config.num_cpu_blocks = 8
cache_config.num_gpu_blocks = 8
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
# Add seq groups to scheduler.
for i in range(num_seq_group):
@@ -93,7 +92,7 @@ def test_chunk():
cache_config.num_cpu_blocks = 32
cache_config.num_gpu_blocks = 32
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
# Add seq groups to scheduler.
for i in range(2):
@@ -145,7 +144,7 @@ def test_concurrent_chunking():
cache_config.num_cpu_blocks = 32
cache_config.num_gpu_blocks = 32
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
# Add seq groups to scheduler.
for i in range(2):
@@ -226,8 +225,8 @@ def test_short_prompts_jump_long_prompts_in_queue():
cache_config.num_cpu_blocks = 3200 # large KV cache size for large requests
cache_config.num_gpu_blocks = 3200
scheduler = Scheduler(scheduler_config, cache_config, None)
long_seqs: List[SequenceGroup] = []
short_seqs: List[SequenceGroup] = []
long_seqs: list[SequenceGroup] = []
short_seqs: list[SequenceGroup] = []
# Add 2 large seq groups to scheduler.
for i in range(2):
@@ -368,7 +367,7 @@ def test_complex():
cache_config.num_cpu_blocks = 64
cache_config.num_gpu_blocks = 64
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
# Add seq groups to scheduler.
for i in range(2):
@@ -439,7 +438,7 @@ def test_maximal_decoding():
cache_config.num_cpu_blocks = 8
cache_config.num_gpu_blocks = 8
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
# Add seq groups to scheduler.
for i in range(2):
@@ -533,7 +532,7 @@ def test_prompt_limit():
cache_config.num_cpu_blocks = 16
cache_config.num_gpu_blocks = 16
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
_, seq_group = create_dummy_prompt("1",
prompt_length=48,
@@ -565,7 +564,7 @@ def test_prompt_limit_exceed():
cache_config.num_cpu_blocks = 16
cache_config.num_gpu_blocks = 16
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
_, seq_group = create_dummy_prompt("2",
prompt_length=48,
block_size=block_size)
@@ -699,7 +698,7 @@ def test_chunked_prefill_max_seqs():
cache_config.num_cpu_blocks = 128
cache_config.num_gpu_blocks = 128
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
_, seq_group = create_dummy_prompt("1",
prompt_length=65,
@@ -758,7 +757,7 @@ def test_prefix_caching():
cache_config.num_cpu_blocks = 0
cache_config.num_gpu_blocks = 32
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
# Add seq groups to scheduler.
for i in range(2):
@@ -800,7 +799,7 @@ def test_prefix_caching_with_concurrent_partial_prefills():
cache_config.num_cpu_blocks = 0
cache_config.num_gpu_blocks = 32
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
# Add seq groups to scheduler.
for i in range(2):
+9 -10
View File
@@ -2,7 +2,6 @@
import time
from collections import deque
from typing import List, Set, Tuple
from unittest.mock import MagicMock
import pytest # noqa
@@ -57,7 +56,7 @@ def test_scheduler_abort_seq_group():
# Add multiple seq groups to scheduler.
num_seq_group = 4
request_ids: Set[str] = set()
request_ids: set[str] = set()
for i in range(num_seq_group):
_, seq_group = create_dummy_prompt(str(i), block_size)
scheduler.add_seq_group(seq_group)
@@ -83,7 +82,7 @@ def test_scheduler_schedule_simple():
cache_config.num_cpu_blocks = 8
cache_config.num_gpu_blocks = 8
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
# Add seq groups to scheduler.
for i in range(num_seq_group):
@@ -221,7 +220,7 @@ def test_scheduler_max_seqs():
cache_config.num_gpu_blocks = 8
scheduler = Scheduler(scheduler_config, cache_config, None)
all_seq_groups: List[SequenceGroup] = []
all_seq_groups: list[SequenceGroup] = []
# Add seq groups to scheduler.
for i in range(num_seq_group):
_, seq_group = create_dummy_prompt(str(i),
@@ -480,7 +479,7 @@ def test_prefill_schedule_max_lora():
num_cpu_blocks=64,
num_gpu_blocks=64)
budget = create_token_budget(token_budget=120)
curr_loras: Set[int] = set()
curr_loras: set[int] = set()
for i in range(2):
_, seq_group = create_dummy_prompt(str(i),
prompt_length=60,
@@ -651,8 +650,8 @@ def test_schedule_swapped_max_loras():
block_size=block_size,
num_cpu_blocks=32,
num_gpu_blocks=32)
curr_loras: Set[int] = set()
blocks_to_swap_out: List[Tuple[int, int]] = []
curr_loras: set[int] = set()
blocks_to_swap_out: list[tuple[int, int]] = []
for i in range(2):
_, seq_group = create_dummy_prompt(str(i),
prompt_length=60,
@@ -683,7 +682,7 @@ def test_schedule_swapped_cannot_swap_in():
num_cpu_blocks=32,
num_gpu_blocks=32)
curr_loras = None
blocks_to_swap_out: List[Tuple[int, int]] = []
blocks_to_swap_out: list[tuple[int, int]] = []
for i in range(2):
_, seq_group = create_dummy_prompt(str(i),
prompt_length=60,
@@ -714,7 +713,7 @@ def test_infeasible_swap():
num_cpu_blocks=32,
num_gpu_blocks=32)
curr_loras = None
blocks_to_swap_out: List[Tuple[int, int]] = []
blocks_to_swap_out: list[tuple[int, int]] = []
for i in range(2):
_, seq_group = create_dummy_prompt(str(i),
prompt_length=60,
@@ -752,7 +751,7 @@ def test_schedule_swapped_blocks_to_copy():
block_size=block_size)
scheduler._allocate_and_set_running(seq_group)
append_new_token_seq_group(60, seq_group, 1)
blocks_to_swap_out: List[Tuple[int, int]] = []
blocks_to_swap_out: list[tuple[int, int]] = []
scheduler._swap_out(seq_group, blocks_to_swap_out)
scheduler._add_seq_group_to_swapped(seq_group)
+1 -3
View File
@@ -1,7 +1,5 @@
# SPDX-License-Identifier: Apache-2.0
from typing import List
import pytest # noqa
from vllm.config import CacheConfig, SchedulerConfig
@@ -48,7 +46,7 @@ def test_scheduler_schedule_simple_encoder_decoder():
cache_config.num_cpu_blocks = 16 # enc and dec prompts per seq_group
cache_config.num_gpu_blocks = 16 # enc and dec prompts per seq_group
scheduler = Scheduler(scheduler_config, cache_config, None)
running: List[SequenceGroup] = []
running: list[SequenceGroup] = []
# Add seq groups to scheduler.
req_id_list = []
+10 -11
View File
@@ -2,9 +2,8 @@
import time
from collections import defaultdict
from typing import Any, Dict, List, Optional
from typing import Sequence as GenericSequence
from typing import Tuple
from collections.abc import Sequence as GenericSequence
from typing import Any, Optional
from vllm import SamplingParams
from vllm.core.scheduler import Scheduler, SchedulerOutputs
@@ -20,10 +19,10 @@ def create_dummy_prompt(
block_size: Optional[int] = None,
lora_request: Optional[LoRARequest] = None,
best_of: int = 1,
prompt_tokens: Optional[List[int]] = None,
prompt_tokens: Optional[list[int]] = None,
min_tokens: int = 0,
max_tokens: int = 16,
) -> Tuple[Sequence, SequenceGroup]:
) -> tuple[Sequence, SequenceGroup]:
if not block_size:
block_size = prompt_length
@@ -48,7 +47,7 @@ def create_dummy_prompt(
return prompt, seq_group
def create_dummy_lora_sequence(request_id: int, token_ids: List[int],
def create_dummy_lora_sequence(request_id: int, token_ids: list[int],
block_size: int, lora_int_id: int) -> Sequence:
return Sequence(seq_id=request_id,
inputs=token_inputs(token_ids),
@@ -58,7 +57,7 @@ def create_dummy_lora_sequence(request_id: int, token_ids: List[int],
lora_int_id=lora_int_id))
def create_dummy_sequence(request_id: int, token_ids: List[int],
def create_dummy_sequence(request_id: int, token_ids: list[int],
block_size: int) -> Sequence:
return Sequence(
seq_id=request_id,
@@ -74,7 +73,7 @@ def create_dummy_prompt_encoder_decoder(
block_size: Optional[int] = None,
lora_request: Optional[LoRARequest] = None,
best_of: int = 1,
) -> Tuple[Sequence, Sequence, SequenceGroup]:
) -> tuple[Sequence, Sequence, SequenceGroup]:
if not block_size:
block_size = decoder_prompt_length
@@ -125,7 +124,7 @@ def create_seq_group(
prompt_token_ids = [0] * seq_prompt_len
seqs: List[Sequence] = []
seqs: list[Sequence] = []
for seq_id_offset, output_len in enumerate(seq_output_lens):
seq = Sequence(
seq_id=seq_id_start + seq_id_offset,
@@ -241,7 +240,7 @@ class SchedulerProxy:
def __init__(self, scheduler: Scheduler):
self.scheduler_ = scheduler
self.call_history: Dict[str, List[Any]] = defaultdict(list)
self.call_history: dict[str, list[Any]] = defaultdict(list)
def __getattr__(self, name: str) -> Any:
@@ -253,6 +252,6 @@ class SchedulerProxy:
return wrapper
def last_schedule_ret(
self, ) -> Tuple[List[SequenceGroupMetadata], SchedulerOutputs, Any]:
self, ) -> tuple[list[SequenceGroupMetadata], SchedulerOutputs, Any]:
_, _, ret = self.call_history["schedule"][-1]
return ret