Skip to content

Reset AG_pipeline bucket status after validation step. #3155

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

Draft
vasunvidia wants to merge 1 commit into
base: main
Choose a base branch
from
+31 −1
Draft

Reset AG_pipeline bucket status after validation step. #3155

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 5 additions & 1 deletion megatron/core/distributed/fsdp/src/megatron_fsdp/megatron_fsdp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1164,7 +1164,7 @@ def reset_mixed_precision_policy(self, mixed_precision_policy: MixedPrecisionPol
self.mp_policy = mp_policy_reset
self.param_and_grad_buffer.mp_policy = mp_policy_reset

def start_param_sync(self, *unused, force_sync: bool = False, force_dispatch: bool = False):
def start_param_sync(self, *unused, force_sync: bool = False, force_dispatch: bool = False, sync_and_return: bool = False):
"""
Initiates param sync (all-gather) communication operations for all model parameters.

Expand All @@ -1177,6 +1177,10 @@ def start_param_sync(self, *unused, force_sync: bool = False, force_dispatch: bo
other settings.
force_dispatch (bool, optional): force dispatch regardless of other settings.
"""
if sync_and_return:
self.synchronize_param_gather()
return

self._replace_param_with_raw_if_needed()

if not force_sync and self.ddp_config.overlap_param_gather:
Expand Down
26 changes: 26 additions & 0 deletions megatron/core/pipeline_parallel/schedules.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -734,6 +734,15 @@ def forward_backward_no_pipelining(

if getattr(config, 'fine_grained_activation_offloading', False):
off_interface.reset()
# Reset all_gather_pipeline bucket status before next validation iteration
if forward_only:
for model_chunk in [model]:
if (
model_chunk.ddp_config.overlap_param_gather
and model_chunk.ddp_config.use_megatron_fsdp
):
model_chunk.start_param_sync(sync_and_return=True)


if config.timers is not None:
config.timers('forward-backward').stop()
Expand Down Expand Up @@ -1954,6 +1963,15 @@ def pp_post_backward(input_tensor_grad, vp_stage=None):

if getattr(config, 'fine_grained_activation_offloading', False):
off_interface.reset()
# Reset all_gather_pipeline bucket status before next validation iteration
if forward_only:
for model_chunk in model:
if (
model_chunk.ddp_config.overlap_param_gather
and model_chunk.ddp_config.use_megatron_fsdp
):
model_chunk.start_param_sync(sync_and_return=True)

# Restore config.grad_sync_func and config.param_sync_func.
if forward_only:
config.grad_sync_func, config.param_sync_func = grad_sync_func, param_sync_func
Expand Down Expand Up @@ -2361,6 +2379,14 @@ def enable_grad_sync():

if getattr(config, 'fine_grained_activation_offloading', False):
off_interface.reset()
# Reset all_gather_pipeline bucket status before next validation iteration
if forward_only:
for model_chunk in [model]:
if (
model_chunk.ddp_config.overlap_param_gather
and model_chunk.ddp_config.use_megatron_fsdp
):
model_chunk.start_param_sync(sync_and_return=True)

if config.timers is not None:
config.timers('forward-backward').stop()
Expand Down