Skip to content

[Feature] Allow configuring FlashInfer workspace size (#25342) #25344

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
mishra-krishna wants to merge 2 commits into from
Closed

[Feature] Allow configuring FlashInfer workspace size (#25342) #25344

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
16e65aa
[Feature] Allow configuring FlashInfer workspace size (#25342)
Sep 21, 2025
f0fbde5
Update vllm/v1/attention/backends/flashinfer.py
mishra-krishna Sep 21, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion tests/kernels/attention/test_flashinfer_mla_decode.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import os
import pytest
import torch
import torch.nn.functional as F
Expand All @@ -8,7 +9,8 @@

from vllm.platforms import current_platform

FLASHINFER_WORKSPACE_BUFFER_SIZE = 128 * 1024 * 1024
FLASHINFER_WORKSPACE_BUFFER_SIZE = int(
Copy link
Collaborator

@pavanimajety pavanimajety Sep 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please move this to vllm/envs.py

Copy link
Collaborator

@wangshangsam wangshangsam Oct 25, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1

os.environ.get("VLLM_FLASHINFER_WORKSPACE_BUFFER_SIZE", 128 * 1024 * 1024))

if not current_platform.has_device_capability(100):
pytest.skip(
Expand Down
8 changes: 7 additions & 1 deletion vllm/v1/attention/backends/flashinfer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
"""Attention layer with FlashInfer."""
from __future__ import annotations

import os
from dataclasses import dataclass
from typing import ClassVar, Optional, Union

Expand Down Expand Up @@ -41,7 +42,12 @@
# yapf: enable
from vllm.v1.kv_cache_interface import AttentionSpec

FLASHINFER_WORKSPACE_BUFFER_SIZE = 256 * 1024 * 1024
_default_workspace_size = 256 * 1024 * 1024
_env_val = os.environ.get("VLLM_FLASHINFER_WORKSPACE_BUFFER_SIZE", _default_workspace_size)
try:
FLASHINFER_WORKSPACE_BUFFER_SIZE = int(_env_val)
except (ValueError, TypeError):
FLASHINFER_WORKSPACE_BUFFER_SIZE = _default_workspace_size

FP8_DTYPE = current_platform.fp8_dtype()
FP4_DTYPE = torch.uint8
Expand Down
4 changes: 3 additions & 1 deletion vllm/v1/attention/backends/mla/common.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -188,6 +188,7 @@
"""

import functools
import os
from abc import abstractmethod
from dataclasses import dataclass, field
from typing import ClassVar, Generic, Optional, TypeVar, Union
Expand Down Expand Up @@ -426,7 +427,8 @@ def use_cudnn_prefill() -> bool:
# Currently 394MB, this can be tuned based on GEMM sizes used.
# Chosen to be the same as sglang:
# https://github.com/sgl-project/sglang/blob/766392c6bda2558b61ce6d1c1bfd8081a549e1f1/python/sglang/global_config.py#L37
FLASHINFER_WORKSPACE_BUFFER_SIZE = 394 * 1024 * 1024
FLASHINFER_WORKSPACE_BUFFER_SIZE = int(
os.environ.get("VLLM_FLASHINFER_WORKSPACE_BUFFER_SIZE", 394 * 1024 * 1024))


class MLACommonMetadataBuilder(AttentionMetadataBuilder[M]):
Expand Down
Loading