tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

report min lenght of tokenized data (#1186) [skip ci]

Browse Source
This commit is contained in:
Wing Lian
2024-01-24 09:17:50 -05:00
committed by GitHub
parent 02f2c720fc
commit d85d4942cf
1 changed files with 2 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/axolotl/utils/trainer.py
View File

@@ -110,6 +110,8 @@ def process_datasets_for_packing(cfg, train_dataset, eval_dataset):
drop_long = partial(drop_long_seq, sequence_len=cfg.sequence_len) drop_long = partial(drop_long_seq, sequence_len=cfg.sequence_len)
with zero_first(is_main_process()): with zero_first(is_main_process()):
if cfg.is_preprocess: if cfg.is_preprocess:
min_input_len = np.min(get_dataset_lengths(train_dataset))
LOG.debug(f"min_input_len: {min_input_len}", main_process_only=True)
max_input_len = np.max(get_dataset_lengths(train_dataset)) max_input_len = np.max(get_dataset_lengths(train_dataset))
LOG.debug(f"max_input_len: {max_input_len}", main_process_only=True) LOG.debug(f"max_input_len: {max_input_len}", main_process_only=True)