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.

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

[Iluvatar] llama2 7b case #343

Merged
merged 13 commits into from
Dec 18, 2023

Conversation

clveryang
Copy link
Contributor

No description provided.

zhangsanfeng2022 and others added 8 commits November 27, 2023 13:13
* 【kunlunxin】Longformer update config

* update memory used for kunlunxin and solve file conflict.

* update 2x8 memory used.

* update kunlunxin docs.

* update kunlunxin 1x1 memory docs.

* Change to a larger batch size.

* Remove empty line in test_config.py

* Update README.

* Update test_conf.py, add comma

---------

Co-authored-by: zhangsanfeng2022 <[email protected]>
Co-authored-by: root <[email protected]>
Co-authored-by: Zhou Yu <[email protected]>
* update

* update

* update

* update

* update

* update

---------

Co-authored-by: 魏杰 <[email protected]>
@clveryang clveryang changed the title [Iluvatar] llama27 b case [Iluvatar] llama2 7b case Nov 29, 2023

- ##### 软件环境
- OS版本:Ubuntu 20.04
- Docker 版本:20.10.18
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

20.10.21,请参考flagperf自动打印出来的sys_info.log

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

问题已经修复

seqlength = 2048
batchsize = 4
datafilename = "openwebtext_llama2_100M.npy"
theoryflops = 312000000000000.0
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

此值目前填的是312Tflops,为A100的值。请修改为天数bf16算力。此值影响MFU计算,不影响吞吐量计算

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ok

epochs = 1
flashattn = True
gradient_checkpointing = True
use_cache = false
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

拼写错误

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

收到

@shh2000 shh2000 merged commit 41a67bd into FlagOpen:iluvatar_deepspeed Dec 18, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

7 participants