transformers
Enhancing SFT Training Efficiency Using Packing and FlashAttention2 with Position IDs
#31629
Merged
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Overview
Commits
13
Changes
View On
GitHub
Enhancing SFT Training Efficiency Using Packing and FlashAttention2 with Position IDs
#31629
ArthurZucker
merged 13 commits into
huggingface:main
from
RhuiDih:dev/fa_packing_posid
ArthurZucker
commented on 2024-07-10
RhuiDih
force pushed
from
cf6271fa
to
c3451dbc
1 year ago
ArthurZucker
commented on 2024-07-15
ArthurZucker
approved these changes on 2024-07-19
fxmarty
commented on 2024-07-19
add DataCollatorBatchFlattening
df3c9b29
Update data_collator.py
dfe08de7
change name
8120b3a3
new FA2 flow if position_ids is provided
0598510e
add comments
1ff23436
minor fix
f97ab716
minor fix data collator
08e1f2cb
add test cases for models
48ce3d20
add test case for data collator
6c6b1688
remove extra code
00e7abf6
RhuiDih
force pushed
to
00e7abf6
1 year ago
formating for ruff check and check_repo.py
3a872933
ruff format
c60b720d
custom_init_isort.py
90305596
ArthurZucker
approved these changes on 2024-07-23
ArthurZucker
merged
9cf4f2aa
into main
1 year ago
Login to write a write a comment.
Login via GitHub
Reviewers
ArthurZucker
fxmarty
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub