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

xpu use allgather #8697

Merged
merged 2 commits into from
Jul 3, 2024
Merged

xpu use allgather #8697

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b18ca2e
xpu use allgather
FeixLiu Jul 2, 2024
7912d27
xpu use allgather
FeixLiu Jul 3, 2024
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
8 changes: 6 additions & 2 deletions paddlenlp/transformers/conversion_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,7 @@
from paddle import Tensor
from paddle.nn import Layer

from paddlenlp.utils.distributed import distributed_gather
from paddlenlp.utils.distributed import distributed_allgather, distributed_gather
from paddlenlp.utils.env import CONFIG_NAME, PADDLE_WEIGHTS_NAME, PYTORCH_WEIGHTS_NAME
from paddlenlp.utils.import_utils import (
is_package_available,
Expand All @@ -46,6 +46,7 @@
)
from paddlenlp.utils.log import logger
from paddlenlp.utils.serialization import load_torch
from paddlenlp.utils.tools import get_env_device

if TYPE_CHECKING:
from paddlenlp.transformers import PretrainedConfig, PretrainedModel
Expand Down Expand Up @@ -1269,7 +1270,10 @@
for key in state_dict.keys():
tensor = state_dict[key]
if key in name_action_mappings:
ret = distributed_gather(tensor, group=mp_group, offload=True)
if get_env_device() == "xpu":
ret = distributed_allgather(tensor, group=mp_group, offload=True)

Check warning on line 1274 in paddlenlp/transformers/conversion_utils.py

View check run for this annotation

Codecov / codecov/patch

paddlenlp/transformers/conversion_utils.py#L1273-L1274 

Added lines #L1273 - L1274 were not covered by tests
else:
ret = distributed_gather(tensor, group=mp_group, offload=True)

Check warning on line 1276 in paddlenlp/transformers/conversion_utils.py

View check run for this annotation

Codecov / codecov/patch

paddlenlp/transformers/conversion_utils.py#L1276 

Added line #L1276 was not covered by tests
action = name_action_mappings.pop(key)
tensor = action(ret) if is_dst else None
else:
Expand Down
Loading