Skip to content
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

Using explicit GPU upcast for ZeRO-Offload #6962

Merged
merged 1 commit into from
Jan 21, 2025
Merged

Using explicit GPU upcast for ZeRO-Offload #6962

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
13 changes: 2 additions & 11 deletions deepspeed/runtime/zero/stage3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -546,15 +546,10 @@ def _setup_for_real_optimizer(self):
self.grad_partitions_flat_buffer = get_accelerator().pin_memory(self.grad_partitions_flat_buffer)

offset = 0
max_partition_numel = 0
for param in all_params:
self.__param_id_to_grad_partition[param.ds_id] = self.grad_partitions_flat_buffer.narrow(
0, offset, param.partition_numel())
offset += param.partition_numel()
max_partition_numel = max(max_partition_numel, param.partition_numel())
if self.offload_optimizer:
self.pinned_grad_buffer: Tensor = get_accelerator().pin_memory(
torch.empty(max_partition_numel, device=self.device))

def _link_all_hp_params(self):
for p in self.module.parameters():
Expand Down Expand Up @@ -1510,13 +1505,9 @@ def partition_grads(self, params_to_release: List[Parameter], grad_partitions: L
offload_fp32_gradients[i].append(grad_buffer.float())
offload_fp32_offsets[i].append(dest_offset)
else:
buffer_numel = grad_buffer.numel()
fp32_grad_tensor = self.fp32_partitioned_groups_flat[i].grad.narrow(
0, dest_offset, buffer_numel)
self.pinned_grad_buffer[:buffer_numel].copy_(
grad_buffer.to(dtype=torch.float32, non_blocking=True))
get_accelerator().synchronize()
fp32_grad_tensor.copy_(self.pinned_grad_buffer[:buffer_numel], non_blocking=True)
0, dest_offset, grad_buffer.numel())
fp32_grad_tensor.copy_(grad_buffer.float())

# free the gradient
if not get_accelerator().is_synchronized_device():
Expand Down