biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-12 17:51:31 +01:00
committed by GitHub
parent 9bb38130cb
commit 8fcaaf6a16
944 changed files with 9490 additions and 10121 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
tests/kernels/moe/modular_kernel_tools/mk_objects.py
View File

@@ -1,7 +1,6 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from dataclasses import dataclass
from typing import Optional, Union
import torch
@@ -43,25 +42,25 @@ from vllm.utils.flashinfer import has_flashinfer_cutlass_fused_moe
@dataclass
class TestMoEQuantConfig:
quant_dtype: Union[torch.dtype, str, None]
quant_dtype: torch.dtype | str | None
per_out_ch_quant: bool
per_act_token_quant: bool
block_shape: Optional[list[int]]
block_shape: list[int] | None
@dataclass
class PrepareFinalizeInfo:
activation_format: mk.FusedMoEActivationFormat
supported_dtypes: list[Union[torch.dtype, str]]
supported_dtypes: list[torch.dtype | str]
blocked_quantization_support: bool
backend: Optional[str]
backend: str | None
supports_apply_weight_on_input: bool = True
@dataclass
class ExpertInfo:
activation_format: mk.FusedMoEActivationFormat
supported_dtypes: list[Union[torch.dtype, str]]
supported_dtypes: list[torch.dtype | str]
blocked_quantization_support: bool
supports_chunking: bool
supports_expert_map: bool
@@ -78,7 +77,7 @@ MK_FUSED_EXPERT_TYPES: list[mk.FusedMoEPermuteExpertsUnpermute] = []
standard_format = mk.FusedMoEActivationFormat.Standard
batched_format = mk.FusedMoEActivationFormat.BatchedExperts
common_float_types: list[Union[torch.dtype, str]] = [
common_float_types: list[torch.dtype | str] = [
torch.float8_e4m3fn,
torch.bfloat16,
torch.float16,
@@ -92,9 +91,9 @@ fp8_types = [torch.float8_e4m3fn]
def register_prepare_and_finalize(
kind,
activation_format: mk.FusedMoEActivationFormat,
supported_dtypes: list[Union[torch.dtype, str]],
supported_dtypes: list[torch.dtype | str],
blocked_quantization_support: bool,
backend: Optional[str],
backend: str | None,
force_multigpu: bool = False,
supports_apply_weight_on_input: bool = True,
):
@@ -121,7 +120,7 @@ def register_prepare_and_finalize(
def register_experts(
kind,
activation_format: mk.FusedMoEActivationFormat,
supported_dtypes: list[Union[torch.dtype, str]],
supported_dtypes: list[torch.dtype | str],
blocked_quantization_support: bool,
supports_chunking: bool,
supports_expert_map: bool,
@@ -340,7 +339,7 @@ if cutlass_fp4_supported():
supports_expert_map=False,
)
MK_QUANT_CONFIGS: list[Optional[TestMoEQuantConfig]] = [
MK_QUANT_CONFIGS: list[TestMoEQuantConfig | None] = [
None,
# per-channel / per-column weights and per-tensor activations
TestMoEQuantConfig(
@@ -395,7 +394,7 @@ if cutlass_fp4_supported() or has_flashinfer_cutlass_fused_moe():
def make_prepare_finalize(
prepare_finalize_type: mk.FusedMoEPrepareAndFinalize,
backend: Optional[str],
backend: str | None,
moe: FusedMoEConfig,
quant_config: FusedMoEQuantConfig,
) -> mk.FusedMoEPrepareAndFinalize: