Skip to content

block swap with no_grad context #1988

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

Draft
wants to merge 2 commits into
base: sd3
Choose a base branch
from
Draft

block swap with no_grad context #1988

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c898e4e
fix: optimize weight device swapping with no_grad context
kohya-ss Mar 17, 2025
0e8b5d4
Merge branch 'sd3' into no-grad-block-swap
kohya-ss May 24, 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
27 changes: 14 additions & 13 deletions library/custom_offloading_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,19 +42,20 @@ def swap_weight_devices_cuda(device: torch.device, layer_to_cpu: nn.Module, laye

torch.cuda.current_stream().synchronize() # this prevents the illegal loss value

stream = torch.cuda.Stream()
with torch.cuda.stream(stream):
# cuda to cpu
for module_to_cpu, module_to_cuda, cuda_data_view, cpu_data_view in weight_swap_jobs:
cuda_data_view.record_stream(stream)
module_to_cpu.weight.data = cuda_data_view.data.to("cpu", non_blocking=True)
with torch.no_grad():
stream = torch.cuda.Stream()
with torch.cuda.stream(stream):
# cuda to cpu
for module_to_cpu, module_to_cuda, cuda_data_view, cpu_data_view in weight_swap_jobs:
cuda_data_view.record_stream(stream)
module_to_cpu.weight.data = cuda_data_view.data.to("cpu", non_blocking=True)

stream.synchronize()
stream.synchronize()

# cpu to cuda
for module_to_cpu, module_to_cuda, cuda_data_view, cpu_data_view in weight_swap_jobs:
cuda_data_view.copy_(module_to_cuda.weight.data, non_blocking=True)
module_to_cuda.weight.data = cuda_data_view
# cpu to cuda
for module_to_cpu, module_to_cuda, cuda_data_view, cpu_data_view in weight_swap_jobs:
cuda_data_view.copy_(module_to_cuda.weight.data, non_blocking=True)
module_to_cuda.weight.data = cuda_data_view

stream.synchronize()
torch.cuda.current_stream().synchronize() # this prevents the illegal loss value
Expand All @@ -75,14 +76,14 @@ def swap_weight_devices_no_cuda(device: torch.device, layer_to_cpu: nn.Module, l
for module_to_cpu, module_to_cuda, cuda_data_view, cpu_data_view in weight_swap_jobs:
module_to_cpu.weight.data = cuda_data_view.data.to("cpu", non_blocking=True)

synchronize_device()
synchronize_device(device)

# cpu to device
for module_to_cpu, module_to_cuda, cuda_data_view, cpu_data_view in weight_swap_jobs:
cuda_data_view.copy_(module_to_cuda.weight.data, non_blocking=True)
module_to_cuda.weight.data = cuda_data_view

synchronize_device()
synchronize_device(device)


def weighs_to_device(layer: nn.Module, device: torch.device):
Expand Down