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

Jump to bottom

fix distributed devices #612

Merged
merged 3 commits into from
Sep 21, 2023
Merged

fix distributed devices #612

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2129036
fix distributed devices
maximegmd Sep 20, 2023
b670694
Update distributed.py
maximegmd Sep 20, 2023
a75a3bc
Update distributed.py
maximegmd Sep 20, 2023
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
16 changes: 12 additions & 4 deletions src/axolotl/utils/distributed.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,7 +77,9 @@ def gather_scalar_from_all_ranks(fn, world_size=1): # pylint: disable=invalid-n
value_scalar = fn()
if not is_distributed():
return [value_scalar]
value_tensor = torch.tensor(value_scalar, device=dist.get_rank()).float()
value_tensor = torch.tensor(
value_scalar, device=torch.cuda.current_device()
).float()

if not is_main_process():
dist.gather(value_tensor, dst=0)
Expand Down Expand Up @@ -137,9 +139,13 @@ def compute_and_broadcast(fn): # pylint: disable=invalid-name
"""
if is_main_process():
value_scalar = fn()
value_tensor = torch.tensor(value_scalar, device=dist.get_rank()).float()
value_tensor = torch.tensor(
value_scalar, device=torch.cuda.current_device()
).float()
else:
value_tensor = torch.tensor(0.0, device=dist.get_rank()) # Placeholder tensor
value_tensor = torch.tensor(
0.0, device=torch.cuda.current_device()
) # Placeholder tensor

# Broadcast the tensor to all processes.
barrier()
Expand All @@ -164,7 +170,9 @@ def gather_from_all_ranks(fn, world_size=1): # pylint: disable=invalid-name
- A list of computed values from all ranks if on the gathering rank, otherwise None.
"""
value_scalar = fn()
value_tensor = torch.tensor(value_scalar, device=dist.get_rank()).float()
value_tensor = torch.tensor(
value_scalar, device=torch.cuda.current_device()
).float()

# Placeholder tensor for gathering results
if is_main_process():
Expand Down