Skip to content

Issues: laekov/fastmoe

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

20 Open 93 Closed
20 Open 93 Closed
Author
Filter by author
Label
Filter by label
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Milestones
Filter by milestone
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

CUDA memory increases after each loss.backward()
#201 opened Mar 22, 2024 by sreetamasarkar
6
A bug in switch_gate
#199 opened Mar 7, 2024 by Heihaierr
6
About switch_gate
#198 opened Mar 6, 2024 by Heihaierr
1
multi-node problem
#197 opened Mar 4, 2024 by Qianshaowei
1
Example to run Megatron
#196 opened Feb 26, 2024 by Juanhui28
3
[BUG] AttributeError: module 'fmoe_cuda' has no attribute 'assign_pos_'
#195 opened Feb 17, 2024 by pangsg
3
fastmoe是否支持微调，page-attention，flasahattention和kvcache，混合精度等
#192 opened Feb 5, 2024 by PowerDispatch
4
请问fastmoe能被集成到VLLM里吗
#191 opened Feb 4, 2024 by pangsg
4
我们有线上沟通的群吗
#189 opened Feb 4, 2024 by PowerDispatch
1
开启Smart schedule时报错Segmentation fault bug Something isn't working
#184 opened Dec 30, 2023 by Xingzhi107
@laekov
8
pytest error
#183 opened Dec 29, 2023 by R-QinQ
3
ImportError: cannot import name 'get_args' from 'megatron'
#181 opened Dec 14, 2023 by peter-fei
5
During inference, the output of noisy gate is nan.
#180 opened Dec 3, 2023 by zqhang
5
ModuleNotFoundError: No module named 'fmoe_cuda'
#177 opened Nov 6, 2023 by Taskii-Lei
1
how to use balance loss?
#176 opened Oct 26, 2023 by Heihaierr
1
MoE L2 norm reduce in Megatron
#169 opened Aug 10, 2023 by blankde
3
No overlapping observed when enabling Smart Scheduling
#168 opened Aug 4, 2023 by chenyu-jiang
@zms1999
8
Only 204 unique tokens (vocabulary size) in enwik8 (transformer-XL example)
#163 opened Jul 7, 2023 by chenwydj
3
Mixture of Expert in Vison Task (Segmentation )
#161 opened Jun 7, 2023 by deep-matter
2
Adding Expert Prototyping to FastMoE enhancement New feature or request
#69 opened Aug 23, 2021 by JustinLin610
@Sengxian
1
ProTip! Type g i on any issue or pull request to go back to the issue listing page.