Skip to content

Allow pinning quantized tensors. #10873

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 2 commits into from
Nov 25, 2025
Merged

Allow pinning quantized tensors. #10873

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
0411153
Allow pinning quant ops tensors.
comfyanonymous Nov 25, 2025
cdc2116
Ruff
comfyanonymous Nov 25, 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
6 changes: 5 additions & 1 deletion comfy/model_management.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1098,13 +1098,14 @@ def cast_to_device(tensor, device, dtype, copy=False):
MAX_PINNED_MEMORY = get_total_memory(torch.device("cpu")) * 0.95
logging.info("Enabled pinned memory {}".format(MAX_PINNED_MEMORY // (1024 * 1024)))

PINNING_ALLOWED_TYPES = set(["Parameter", "QuantizedTensor"])

def pin_memory(tensor):
global TOTAL_PINNED_MEMORY
if MAX_PINNED_MEMORY <= 0:
return False

if type(tensor) is not torch.nn.parameter.Parameter:
if type(tensor).__name__ not in PINNING_ALLOWED_TYPES:
return False

if not is_device_cpu(tensor.device):
Expand All @@ -1124,6 +1125,9 @@ def pin_memory(tensor):
return False

ptr = tensor.data_ptr()
if ptr == 0:
return False

if torch.cuda.cudart().cudaHostRegister(ptr, size, 1) == 0:
PINNED_MEMORY[ptr] = size
TOTAL_PINNED_MEMORY += size
Expand Down
8 changes: 8 additions & 0 deletions comfy/quant_ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -228,6 +228,14 @@ def dequant_arg(arg):
new_kwargs = dequant_arg(kwargs)
return func(*new_args, **new_kwargs)

def data_ptr(self):
return self._qdata.data_ptr()

def is_pinned(self):
return self._qdata.is_pinned()

def is_contiguous(self):
return self._qdata.is_contiguous()

# ==============================================================================
# Generic Utilities (Layout-Agnostic Operations)
Expand Down