Skip to content

Turn off cuda malloc by default when --fast autotune is turned on. #10393

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

Merged
comfyanonymous merged 1 commit into from
Oct 19, 2025
Merged

Turn off cuda malloc by default when --fast autotune is turned on. #10393

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
3 changes: 3 additions & 0 deletions comfy/model_management.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -371,6 +371,9 @@ def amd_min_version(device=None, min_rdna_version=0):
except:
pass

if torch.cuda.is_available() and torch.backends.cudnn.is_available() and PerformanceFeature.AutoTune in args.fast:
torch.backends.cudnn.benchmark = True

try:
if torch_version_numeric >= (2, 5):
torch.backends.cuda.allow_fp16_bf16_reduction_math_sdp(True)
Expand Down
3 changes: 0 additions & 3 deletions comfy/ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,9 +67,6 @@ def scaled_dot_product_attention(q, k, v, *args, **kwargs):

cast_to = comfy.model_management.cast_to #TODO: remove once no more references

if torch.cuda.is_available() and torch.backends.cudnn.is_available() and PerformanceFeature.AutoTune in args.fast:
torch.backends.cudnn.benchmark = True

def cast_to_input(weight, input, non_blocking=False, copy=True):
return comfy.model_management.cast_to(weight, input.dtype, input.device, non_blocking=non_blocking, copy=copy)

Expand Down
7 changes: 4 additions & 3 deletions cuda_malloc.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
import os
import importlib.util
from comfy.cli_args import args
from comfy.cli_args import args, PerformanceFeature
import subprocess

#Can't use pytorch to get the GPU names because the cuda malloc has to be set before the first import.
Expand Down Expand Up @@ -75,8 +75,9 @@ def cuda_malloc_supported():
spec.loader.exec_module(module)
version = module.__version__

if int(version[0]) >= 2 and "+cu" in version: #enable by default for torch version 2.0 and up only on cuda torch
args.cuda_malloc = cuda_malloc_supported()
if int(version[0]) >= 2 and "+cu" in version: # enable by default for torch version 2.0 and up only on cuda torch
if PerformanceFeature.AutoTune not in args.fast: # Autotune has issues with cuda malloc
args.cuda_malloc = cuda_malloc_supported()
except:
pass

Expand Down