[TPU] Call torch._sync(param) during weight loading (#9437)

This commit is contained in:
Woosuk Kwon 2024-10-17 09:00:11 -07:00 committed by GitHub
parent 5e443b594f
commit 8e1cddcd44
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194

View File

@ -3,6 +3,7 @@ from typing import Any, Dict, Optional
import torch
from vllm.platforms import current_platform
from vllm.utils import seed_everything
@ -28,4 +29,25 @@ def set_weight_attrs(
for key, value in weight_attrs.items():
assert not hasattr(
weight, key), (f"Overwriting existing tensor attribute: {key}")
# NOTE(woosuk): During weight loading, we often do something like:
# narrowed_tensor = param.data.narrow(0, offset, len)
# narrowed_tensor.copy_(real_weight)
# expecting narrowed_tensor and param.data to share the same storage.
# However, on TPUs, narrowed_tensor will lazily propagate to the base
# tensor, which is param.data, leading to the redundant memory usage.
# This sometimes causes OOM errors during model loading. To avoid this,
# we sync the param tensor after its weight loader is called.
# TODO(woosuk): Remove this hack once we have a better solution.
if current_platform.is_tpu() and key == "weight_loader":
value = _make_synced_weight_loader(value)
setattr(weight, key, value)
def _make_synced_weight_loader(original_weight_loader):
def _synced_weight_loader(param, *args, **kwargs):
original_weight_loader(param, *args, **kwargs)
torch._sync(param)
return _synced_weight_loader