Skip to content

[BE] remove old pytorch version warning on strided sharding since 2.5 is official released #665

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

Merged
merged 2 commits into from
Oct 30, 2024
Merged

[BE] remove old pytorch version warning on strided sharding since 2.5 is official released #665

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ab922b6
[BE] remove old pytorch version warning on strided sharding since 2.5…
XilunWu Oct 30, 2024
fd1cacc
Update on "[BE] remove old pytorch version warning on strided shardin…
XilunWu Oct 30, 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
7 changes: 0 additions & 7 deletions torchtitan/parallelisms/parallelize_llama.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,6 @@
from torchtitan.config_manager import JobConfig, TORCH_DTYPE_MAP
from torchtitan.logging import logger
from torchtitan.parallelisms.parallel_dims import ParallelDims
from torchtitan.parallelisms.utils import check_strided_sharding_enabled


def parallelize_llama(
Expand Down Expand Up @@ -330,12 +329,6 @@ def apply_fsdp(
if cpu_offload:
fsdp_config["offload_policy"] = CPUOffloadPolicy()

# TODO: remove this check once PyTorch 2.5 is released. We can safely assume
# that users won't use a nightly build which is older than 20240809 by then.
if tp_enabled:
# check if strided sharding is enabled, which is necessary for 2D/3D DCP
check_strided_sharding_enabled()

for layer_id, transformer_block in model.layers.items():
if pp_enabled:
# For PP, do not reshard after forward to avoid per-microbatch
Expand Down
30 changes: 0 additions & 30 deletions torchtitan/parallelisms/utils.py
View file
Open in desktop

This file was deleted.

Loading