Skip to content

Commit d18d1fa

Browse files
committed
feat(torchtitan): add mock HuggingFace dataset support for offline testing
1 parent b39b880 commit d18d1fa

File tree

1 file changed

+1
-1
lines changed

1 file changed

+1
-1
lines changed

primus/modules/trainer/torchtitan/patch_utils.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -59,7 +59,7 @@ def mock_load_dataset(path: str, *args, **kwargs) -> Dataset:
5959
if "validation" in path.lower():
6060
return _create_mock_text_dataset(num_samples=32)
6161
else:
62-
return _create_mock_token_dataset(seq_len=2048, vocab_size=32000, num_samples=256)
62+
return _create_mock_token_dataset(seq_len=8192, vocab_size=32000, num_samples=256)
6363

6464
datasets.load_dataset = mock_load_dataset
6565
logger.warning("[PrimusPath][Dataset] Patched datasets.load_dataset successfully.")

0 commit comments

Comments
 (0)