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/test_cutlass_moe.py
View File

@@ -3,7 +3,6 @@
import copy
import dataclasses
from math import prod
from typing import Optional
import pytest
import torch
@@ -85,16 +84,16 @@ class MOETensors:
@dataclasses.dataclass
class MOETensors8Bit(MOETensors):
# quantized
a_q: Optional[torch.Tensor] = None # a -> a_q
w1_q: Optional[torch.Tensor] = None # w1 -> w1_q
w2_q: Optional[torch.Tensor] = None # w2 -> w2_q
a_scale: Optional[torch.Tensor] = None
w1_scale: Optional[torch.Tensor] = None
w2_scale: Optional[torch.Tensor] = None
a_q: torch.Tensor | None = None # a -> a_q
w1_q: torch.Tensor | None = None # w1 -> w1_q
w2_q: torch.Tensor | None = None # w2 -> w2_q
a_scale: torch.Tensor | None = None
w1_scale: torch.Tensor | None = None
w2_scale: torch.Tensor | None = None
# dequantized
a_d: Optional[torch.Tensor] = None # a -> a_q -> a_d
w1_d: Optional[torch.Tensor] = None # w1 -> w1_q -> w1_d
w2_d: Optional[torch.Tensor] = None # w2 -> w2_q -> w2_d
a_d: torch.Tensor | None = None # a -> a_q -> a_d
w1_d: torch.Tensor | None = None # w1 -> w1_q -> w1_d
w2_d: torch.Tensor | None = None # w2 -> w2_q -> w2_d
@staticmethod
def make_moe_tensors_8bit(
@@ -209,7 +208,7 @@ def run_8_bit(
topk_ids: torch.Tensor,
per_act_token: bool,
per_out_ch: bool,
num_local_experts: Optional[int] = None,
num_local_experts: int | None = None,
) -> torch.Tensor:
assert not any(
[
@@ -280,7 +279,7 @@ def test_cutlass_moe_8_bit_no_graph(
per_act_token: bool,
per_out_ch: bool,
monkeypatch,
ep_size: Optional[int] = None,
ep_size: int | None = None,
):
current_platform.seed_everything(7)
monkeypatch.setenv("VLLM_FUSED_MOE_CHUNK_SIZE", "8192")