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

For xla tensors, use an alternative way to get a unique id #25802

Merged
merged 2 commits into from
Aug 31, 2023
Merged

For xla tensors, use an alternative way to get a unique id #25802

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
75f9203
For xla tensors, use an alternative way to get a unique id
qihqi Aug 28, 2023
cef75a7
add is_torch_tpu_available check
qihqi Aug 28, 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
15 changes: 13 additions & 2 deletions src/transformers/pytorch_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
from safetensors.torch import storage_ptr, storage_size
from torch import nn

from .utils import logging
from .utils import is_torch_tpu_available, logging


ALL_LAYERNORM_LAYERS = [nn.LayerNorm]
Expand Down Expand Up @@ -285,4 +285,15 @@ def id_tensor_storage(tensor: torch.Tensor) -> Tuple[torch.device, int, int]:
guaranteed to be unique and constant for this tensor's storage during its lifetime. Two tensor storages with
non-overlapping lifetimes may have the same id.
"""
return tensor.device, storage_ptr(tensor), storage_size(tensor)
if tensor.device.type == "xla" and is_torch_tpu_available():
# NOTE: xla tensors dont have storage
# use some other unique id to distinguish.
# this is a XLA tensor, it must be created using torch_xla's
# device. So the following import is safe:
qihqi marked this conversation as resolved.
Show resolved Hide resolved
import torch_xla

unique_id = torch_xla._XLAC._xla_get_tensor_id(tensor)
else:
unique_id = storage_ptr(tensor)

return tensor.device, unique_id, storage_size(tensor)