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
@@ -1,7 +1,6 @@
# SPDX-License-Identifier: Apache-2.0
import itertools
from typing import List
import pytest
import torch
@@ -43,7 +42,7 @@ def test_empty_seq_group():
enable_chunked_prefill=False,
enforce_eager=True,
)
seq_group_metadata_list: List[SequenceGroupMetadata] = []
seq_group_metadata_list: list[SequenceGroupMetadata] = []
model_input = model_runner._prepare_model_input_tensors(
seq_group_metadata_list)
(
@@ -103,9 +102,9 @@ def test_prepare_prompt(batch_size):
enforce_eager=True,
)
seq_lens: List[int] = []
encoder_seq_lens: List[int] = []
seq_group_metadata_list: List[SequenceGroupMetadata] = []
seq_lens: list[int] = []
encoder_seq_lens: list[int] = []
seq_group_metadata_list: list[SequenceGroupMetadata] = []
block_tables = {0: [1]}
cross_block_table = [2]
for i in range(batch_size):
@@ -295,9 +294,9 @@ def test_prepare_decode(batch_size, multiple_seqs_per_seq_group):
enforce_eager=True,
)
seq_lens: List[int] = []
encoder_seq_lens: List[int] = []
seq_group_metadata_list: List[SequenceGroupMetadata] = []
seq_lens: list[int] = []
encoder_seq_lens: list[int] = []
seq_group_metadata_list: list[SequenceGroupMetadata] = []
block_tables = {
0: [1],
1: [3]
@@ -503,9 +502,9 @@ def test_prepare_decode_cuda_graph(batch_size, multiple_seqs_per_seq_group):
} if multiple_seqs_per_seq_group else {
0: [1]
}
seq_lens: List[int] = []
encoder_seq_lens: List[int] = []
seq_group_metadata_list: List[SequenceGroupMetadata] = []
seq_lens: list[int] = []
encoder_seq_lens: list[int] = []
seq_group_metadata_list: list[SequenceGroupMetadata] = []
cross_block_table = [2]
expanded_batch_size = 0
+5 -6
View File
@@ -1,7 +1,6 @@
# SPDX-License-Identifier: Apache-2.0
import dataclasses
from typing import List, Tuple, Type
import torch
@@ -27,15 +26,15 @@ class MockAttentionBackend(AttentionBackend):
raise NotImplementedError
@staticmethod
def get_metadata_cls() -> Type["AttentionMetadata"]:
def get_metadata_cls() -> type["AttentionMetadata"]:
return AttentionMetadata
@staticmethod
def get_builder_cls() -> Type["AttentionMetadataBuilder"]:
def get_builder_cls() -> type["AttentionMetadataBuilder"]:
return AttentionMetadataBuilder
@staticmethod
def get_state_cls() -> Type["CommonAttentionState"]:
def get_state_cls() -> type["CommonAttentionState"]:
return CommonAttentionState
@staticmethod
@@ -44,7 +43,7 @@ class MockAttentionBackend(AttentionBackend):
block_size: int,
num_kv_heads: int,
head_size: int,
) -> Tuple[int, ...]:
) -> tuple[int, ...]:
raise NotImplementedError
@staticmethod
@@ -57,7 +56,7 @@ class MockAttentionBackend(AttentionBackend):
@staticmethod
def copy_blocks(
kv_caches: List[torch.Tensor],
kv_caches: list[torch.Tensor],
src_to_dists: torch.Tensor,
) -> None:
pass
+9 -11
View File
@@ -1,7 +1,5 @@
# SPDX-License-Identifier: Apache-2.0
from typing import List
import pytest
import torch
@@ -42,8 +40,8 @@ def test_prepare_prompt(batch_size):
enable_chunked_prefill=False,
)
seq_lens: List[int] = []
seq_group_metadata_list: List[SequenceGroupMetadata] = []
seq_lens: list[int] = []
seq_group_metadata_list: list[SequenceGroupMetadata] = []
block_tables = {0: [1]}
for i in range(batch_size):
# make sure all tokens fit into one block
@@ -159,8 +157,8 @@ def test_prepare_decode_cuda_graph(batch_size):
enable_chunked_prefill=False,
)
context_lens: List[int] = []
seq_group_metadata_list: List[SequenceGroupMetadata] = []
context_lens: list[int] = []
seq_group_metadata_list: list[SequenceGroupMetadata] = []
# Assume each seq group finishes prefill.
for i in range(batch_size):
# make sure all tokens fit into one block
@@ -265,7 +263,7 @@ def test_empty_seq_group():
dtype="float16",
enforce_eager=False,
)
seq_group_metadata_list: List[SequenceGroupMetadata] = []
seq_group_metadata_list: list[SequenceGroupMetadata] = []
model_input = model_runner._prepare_model_input_tensors(
seq_group_metadata_list)
input_tokens, input_positions, attn_metadata = (
@@ -315,10 +313,10 @@ def test_hybrid_batches(batch_size, enforce_eager, distributed_init):
)
# Add prefill requests.
seq_lens: List[int] = []
seq_group_metadata_list: List[SequenceGroupMetadata] = []
prefill_metadata_list: List[SequenceGroupMetadata] = []
decode_metadata_list: List[SequenceGroupMetadata] = []
seq_lens: list[int] = []
seq_group_metadata_list: list[SequenceGroupMetadata] = []
prefill_metadata_list: list[SequenceGroupMetadata] = []
decode_metadata_list: list[SequenceGroupMetadata] = []
block_tables = {0: [1]}
prefill_batch_size = batch_size // 2
decode_batch_size = batch_size - prefill_batch_size