Skip to content

Added torch distributed barrier #2

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
Am1n3e merged 1 commit into from
Nov 24, 2021
Merged

Added torch distributed barrier #2

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
9 changes: 2 additions & 7 deletions megatron/data/biencoder_dataset_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -187,13 +187,8 @@ def get_block_samples_mapping(block_dataset, title_dataset, data_prefix, num_epo
'(seconds): {:4f}'.format(
time.time() - start_time))

# This should be a barrier but nccl barrier assumes
# device_index=rank which is not the case for model
# parallel case
counts = torch.cuda.LongTensor([1])
torch.distributed.all_reduce(counts, group=mpu.get_data_parallel_group())
assert counts[0].item() == torch.distributed.get_world_size(
group=mpu.get_data_parallel_group())
# Wait until rank 0 generate the index file.
torch.distributed.barrier(device_ids=[int(os.environ['LOCAL_RANK'])])

# Load indexed dataset.
print_rank_0(' > loading indexed mapping from {}'.format(
Expand Down
12 changes: 3 additions & 9 deletions megatron/data/dataset_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -699,15 +699,9 @@ def get_samples_mapping(indexed_dataset,
print_rank_0(' > elasped time to build and save samples mapping '
'(seconds): {:4f}'.format(
time.time() - start_time))
# This should be a barrier but nccl barrier assumes
# device_index=rank which is not the case for model
# parallel case
counts = torch.cuda.LongTensor([1])
torch.distributed.all_reduce(counts, group=mpu.get_data_parallel_group())
torch.distributed.all_reduce(counts, group=mpu.get_pipeline_model_parallel_group())
assert counts[0].item() == (
torch.distributed.get_world_size() //
torch.distributed.get_world_size(group=mpu.get_tensor_model_parallel_group()))

# Wait until rank 0 generate the index file.
torch.distributed.barrier(device_ids=[int(os.environ['LOCAL_RANK'])])

# Load indexed dataset.
print_rank_0(' > loading indexed mapping from {}'.format(
Expand Down
11 changes: 2 additions & 9 deletions megatron/data/gpt_dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -299,15 +299,8 @@ def _build_index_mappings(name, data_prefix, documents, sizes,
print_rank_0(' > elasped time to build and save shuffle-idx mapping'
' (seconds): {:4f}'.format(time.time() - start_time))

# This should be a barrier but nccl barrier assumes
# device_index=rank which is not the case for model
# parallel case
counts = torch.cuda.LongTensor([1])
torch.distributed.all_reduce(counts, group=mpu.get_data_parallel_group())
torch.distributed.all_reduce(counts, group=mpu.get_pipeline_model_parallel_group())
assert counts[0].item() == (
torch.distributed.get_world_size() //
torch.distributed.get_world_size(group=mpu.get_tensor_model_parallel_group()))
# Wait until rank 0 generate the index file.
torch.distributed.barrier(device_ids=[int(os.environ['LOCAL_RANK'])])

# Load mappings.
start_time = time.time()
Expand Down
9 changes: 2 additions & 7 deletions megatron/data/realm_dataset_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -177,13 +177,8 @@ def get_block_samples_mapping(block_dataset, title_dataset, data_prefix, num_epo
'(seconds): {:4f}'.format(
time.time() - start_time))

# This should be a barrier but nccl barrier assumes
# device_index=rank which is not the case for model
# parallel case
counts = torch.cuda.LongTensor([1])
torch.distributed.all_reduce(counts, group=mpu.get_data_parallel_group())
assert counts[0].item() == torch.distributed.get_world_size(
group=mpu.get_data_parallel_group())
# Wait until rank 0 generate the index file.
torch.distributed.barrier(device_ids=[int(os.environ['LOCAL_RANK'])])

# Load indexed dataset.
print_rank_0(' > loading indexed mapping from {}'.format(
Expand Down