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

fix signal_file_path to avoid race condition #1253

Closed
wants to merge 2 commits into from
Closed

fix signal_file_path to avoid race condition #1253

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3e77715
fix signal_file_path in locks
ofivite Jun 5, 2024
c40faf8
replace SLURM_JOB_ID with SHA
ofivite Jun 5, 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
3 changes: 2 additions & 1 deletion llmfoundry/data/finetuning/dataloader.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -532,9 +532,10 @@ def _download_remote_hf_dataset(remote_path: str, split: str) -> str:

# Since we don't know exactly what the extension will be, since it is one of a list
# use a signal file to wait for instead of the desired file
slurm_job_id = int(os.getenv('SLURM_JOB_ID', -1))
signal_file_path = os.path.join(
finetune_dir,
f'.node_{dist.get_node_rank()}_local_rank0_completed',
f'.node_{dist.get_node_rank()}_slurm_job_id{slurm_job_id}_local_rank0_completed',
)
if dist.get_local_rank() == 0:
try:
Expand Down
3 changes: 2 additions & 1 deletion llmfoundry/data/finetuning/tasks.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -814,7 +814,8 @@ def build_from_hf(
Returns:
Dataset: The tokenized dataset.
"""
signal_file_path = f'.node_{dist.get_node_rank()}_local_rank0_data_prep_completed'
slurm_job_id = int(os.getenv('SLURM_JOB_ID', -1))
signal_file_path = f'.node_{dist.get_node_rank()}_slurm_job_id{slurm_job_id}_local_rank0_data_prep_completed'

# Non local rank 0 ranks will wait here for local rank 0 to finish the data processing.
# Once local rank 0 is done, the datasets are all cached on disk, and all other ranks
Expand Down
3 changes: 2 additions & 1 deletion llmfoundry/models/hf/hf_causal_lm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -339,7 +339,8 @@ def _autoset_attn_implementation_monkeypatch(
f'init_device="{init_device}" must be either "cpu" or "meta".',
)

signal_file_path = f'.node_{dist.get_node_rank()}_local_rank0_completed'
slurm_job_id = int(os.getenv('SLURM_JOB_ID', -1))
signal_file_path = f'.node_{dist.get_node_rank()}_local_rank0_slurm_job_id{slurm_job_id}_completed'
if dist.get_local_rank() == 0:
with open(signal_file_path, 'wb') as f:
f.write(b'local_rank0_completed_download')
Expand Down
3 changes: 2 additions & 1 deletion llmfoundry/utils/builders.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -453,7 +453,8 @@ def build_tokenizer(
os.environ['TRANSFORMERS_NO_ADVISORY_WARNINGS'] = '1'
os.environ['TOKENIZERS_PARALLELISM'] = 'false'

signal_file_path = f'.node_{dist.get_node_rank()}_local_rank0_completed_tokenizer_setup'
slurm_job_id = int(os.getenv('SLURM_JOB_ID', -1))
signal_file_path = f'.node_{dist.get_node_rank()}_local_rank0_slurm_job_id{slurm_job_id}_completed_tokenizer_setup'

if dist.is_available() and dist.is_initialized(
) and dist.get_world_size() > 1:
Expand Down
Loading