Skip to content

Updated memmap_dtype to automatically choose based on vocab size #795

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

Open
aman-17 wants to merge 11 commits into
base: main
Choose a base branch
from
Open

Updated memmap_dtype to automatically choose based on vocab size #795

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
efcf563
Updated`memmap_dtype` to uint32, added discord and playground to readme
aman-17 Feb 12, 2025
4d84e59
fixed changelog and pytest
aman-17 Feb 12, 2025
376eb29
restored old changes
aman-17 Feb 14, 2025
9aa7ebf
update
aman-17 Feb 15, 2025
23effe4
fixed lint
aman-17 Feb 15, 2025
803c963
fixed tests
aman-17 Feb 21, 2025
00c6bef
forced input_ids to be in range of uint16
aman-17 Feb 21, 2025
167a7ac
fixed uint32 issue
aman-17 Feb 28, 2025
355bd87
resolved README conflict
aman-17 Feb 28, 2025
0a9f7f0
Merge branch 'main' into amanr/bug_fix
aman-17 Mar 4, 2025
cda5141
Merge branch 'main' into amanr/bug_fix
dirkgr Mar 13, 2025
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
4 changes: 4 additions & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0

- MPS support

### Changed

- Updated `memmap_dtype` to `uint32` for compatibility with OLMo-2-1124.

## [v0.6.0](https://github.com/allenai/OLMo/releases/tag/v0.6.0) - 2024-12-17

### Added
Expand Down
5 changes: 4 additions & 1 deletion olmo/data/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
from pathlib import Path
from typing import Any, Dict, List, Optional, cast

import numpy as np
from torch.utils.data import DataLoader, DistributedSampler

from ..aliases import PathOrStr
Expand Down Expand Up @@ -38,10 +39,12 @@ def build_memmap_dataset(
metadata.extend([{"label": label}] * len(label_paths))
else:
raise OLMoConfigurationError("One of DataConfig.paths or DataConfig.datasets is required")

optimal_memmap_dtype = np.uint32 if train_config.model.vocab_size > 2**16 else np.uint16
return MemMapDataset(
*paths,
chunk_size=train_config.model.max_sequence_length,
memmap_dtype=data_config.effective_memmap_dtype,
memmap_dtype=optimal_memmap_dtype,
metadata=metadata,
include_instance_metadata=include_instance_metadata,
pad_token_id=train_config.model.pad_token_id,
Expand Down