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

Jump to bottom

streaming multipack for pretraining dataset #959

Merged
merged 9 commits into from
Jan 6, 2024
Merged

streaming multipack for pretraining dataset #959

Changes from 1 commit
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
8ed5bcb
[Feat] streaming multipack
Dec 15, 2023
da9aee1
WIP make continued pretraining work w multipack
winglian Jan 5, 2024
36b244d
fix up hadrcoding, lint
winglian Jan 5, 2024
680cbe2
fix dict check
winglian Jan 5, 2024
789c972
update test for updated pretraining multipack code
winglian Jan 5, 2024
2a49248
fix hardcoded data collator fix for multipack pretraining
winglian Jan 5, 2024
7c3be2e
fix the collator to be the max length for multipack pretraining
winglian Jan 5, 2024
bea8bee
don't bother with latest tag for test
winglian Jan 5, 2024
5a321c3
cleanup docker build/test
winglian Jan 5, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
Prev Previous commit
Next Next commit
fix dict check
  • Loading branch information
@winglian
winglian committed Jan 5, 2024
commit 680cbe21bb5d7f1ca996301cf9d4130b1566efc7
6 changes: 3 additions & 3 deletions src/axolotl/utils/data.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,9 +71,9 @@ def prepare_dataset(cfg, tokenizer):
else:
path = cfg.pretraining_dataset
name = None
if isinstance(dict, cfg.pretraining_dataset):
path = cfg.pretraining_dataset.path
name = cfg.pretraining_dataset.name
if isinstance(cfg.pretraining_dataset, dict):
path = cfg.pretraining_dataset["path"]
name = cfg.pretraining_dataset["name"]

train_dataset = load_pretraining_dataset(
path,
Expand Down