Skip to content

Issues: pytorch/torchtitan

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

Clear current search query, filters, and sorts
73 Open 512 Closed
73 Open 512 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Issues list

add AC/selective AC to the model
#7 by wanchaol was closed Mar 19, 2024 training techniques enablement
1
@tianyu-l
update readme with local setups
#9 by wanchaol was closed Feb 23, 2024
enable distributed checkpoint save/load
#10 by wanchaol was closed Feb 23, 2024 torchtrain infrastructure building
add Sequence Parallel fp8 support to train llama
#11 by wanchaol was closed May 3, 2024 FP8 Enablement
Add Tensorboard
#17 by wanchaol was closed Feb 23, 2024 torchtrain infrastructure building
2
@tianyu-l
Add ParallelDims helper and --pp_degree option CLA Signed This label is managed by the Meta Open Source bot.
#27 by wconstab was merged Jan 31, 2024 Loading…
6
[torchtrain] add gradient clipping CLA Signed This label is managed by the Meta Open Source bot.
#28 by tianyu-l was merged Jan 31, 2024 Loading…
1
11
Fix sp_degree and 1-D dp mesh CLA Signed This label is managed by the Meta Open Source bot.
#30 by wconstab was merged Feb 1, 2024 Loading…
1
Update dataloading to use packing with fixed seq_length CLA Signed This label is managed by the Meta Open Source bot.
#31 by wanchaol was merged Feb 2, 2024 Loading…
2
Add Sequence Parallelism to llama CLA Signed This label is managed by the Meta Open Source bot.
#32 by wanchaol was merged Feb 7, 2024 Loading…
6
Separate TransformerEmbedding layer CLA Signed This label is managed by the Meta Open Source bot.
#33 by wconstab was merged Feb 2, 2024 Loading…
2
Add linter setup copied from torchtune CLA Signed This label is managed by the Meta Open Source bot.
#34 by wconstab was merged Feb 2, 2024 Loading…
1
Add pytest and CI workflow from torchtune CLA Signed This label is managed by the Meta Open Source bot.
#35 by wconstab was merged Feb 2, 2024 Loading…
7
Fix missing license_header.txt CLA Signed This label is managed by the Meta Open Source bot.
#36 by wconstab was merged Feb 2, 2024 Loading…
Correct the text in the license header CLA Signed This label is managed by the Meta Open Source bot.
#37 by wconstab was merged Feb 2, 2024 Loading…
Apply linters to existing files CLA Signed This label is managed by the Meta Open Source bot.
#38 by wconstab was closed Feb 2, 2024 Loading…
1
Apply linter to existing files CLA Signed This label is managed by the Meta Open Source bot.
#39 by wconstab was merged Feb 2, 2024 Loading…
Remove the accidentally added --compile CLA Signed This label is managed by the Meta Open Source bot.
#42 by fegin was closed Feb 7, 2024 Loading…
1
Add FSDP FP8 support to train llama
#43 by wanchaol was closed May 3, 2024 FP8 Enablement
train scripts on internal cluster
#46 by wanchaol was closed May 3, 2024 torchtrain training on clusters
use FSDP2 to replace FlatParam FSDP
#47 by wanchaol was closed Mar 29, 2024 First OSS release of Torchtrain
@awgu
alpaca dataset data parallel training
#48 by wanchaol was closed Feb 23, 2024 Efficient data loading solution for large datasets
1
@tianyu-l
enable data loading for data parallel training CLA Signed This label is managed by the Meta Open Source bot.
#49 by tianyu-l was merged Feb 8, 2024 Loading…
1
2
Move to cuda unconditionally so pp-only run works CLA Signed This label is managed by the Meta Open Source bot.
#50 by wconstab was merged Feb 10, 2024 Loading…
3
Improve run_llama_train.sh args and add local-ranks-filter CLA Signed This label is managed by the Meta Open Source bot.
#51 by wconstab was merged Feb 10, 2024 Loading…
5
ProTip! Find all open issues with in progress development work with linked:pr.