Skip to content

Issues: BaguaSys/bagua

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

29 Open 117 Closed
29 Open 117 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

Importing Bagua modules do no work
#718 opened Jul 4, 2024 by silverCore97
fix Lightning failing Bagua test for manual optimization
#661 opened Jan 3, 2023 by wangraying
1
Programs get blocked when using multiple nodes training.
#653 opened Nov 22, 2022 by zhaone
1
Support Pytorch 1.13
#652 opened Nov 8, 2022 by wangraying
error with pip install
#614 opened May 26, 2022 by cmlakhan
1
how to do gradient accumulation in bugua
#571 opened Mar 24, 2022 by CaRRotOne
1
Questions about grad hook in bagua
#476 opened Jan 10, 2022 by ProHuper
add communication support for more datatypes
#427 opened Dec 14, 2021 by wangraying
support amp with fused optimizer
#379 opened Nov 16, 2021 by wangraying
2
async algorithm conflict with fused optimizer, can only support flattened inplac... todo 🗒️
#353 opened Nov 2, 2021 by todo bot
remove this after process group destroy supported todo 🗒️
#352 opened Nov 2, 2021 by todo bot
fused_step may miss init_post_optimizer_step_hook todo 🗒️
#351 opened Nov 2, 2021 by todo bot
async algorithm conflict with fused optimizer, can only support flattened inplac... todo 🗒️
#349 opened Nov 2, 2021 by todo bot
How to use bagua to replace torch.distributed.destroy_process_group()
#324 opened Oct 25, 2021 by lixiangMindSpore
5
PyTorch API compatible communication primitives and process group enhancement New feature or request
#320 opened Oct 22, 2021 by NOBLES5E
create landing page and optimization decision tree diagram documentation Improvements or additions to documentation
#315 opened Oct 19, 2021 by NOBLES5E
@NOBLES5E
autotune add first phase CI, and blacklist algorithms that do not need bucket size autotune
#311 opened Oct 19, 2021 by NOBLES5E
@shjwudp
barrier with broadcast enhancement New feature or request
#301 opened Oct 16, 2021 by NOBLES5E
@wangraying
combine **inplace API todo 🗒️
#289 opened Oct 15, 2021 by todo bot
remove the dependency on torch process group todo 🗒️
#288 opened Oct 15, 2021 by todo bot
Create param groups to handle expert + data case (e.g. param.group = moe_group) todo 🗒️
#234 opened Sep 28, 2021 by todo bot
@liuhatry
Create param groups to handle expert + data case (e.g. param.group = moe_group) todo 🗒️
#233 opened Sep 28, 2021 by todo bot
revisit allreduce for moe.gate... todo 🗒️
#232 opened Sep 28, 2021 by todo bot
support all Al backends including CPU bagua-core
#195 opened Sep 15, 2021 by NOBLES5E
support megatron and moe enhancement New feature or request
#181 opened Sep 9, 2021 by NOBLES5E
@liuhatry
ProTip! no:milestone will show everything without a milestone.