LuGY
|
105c5301c3
|
[zero]added hybrid adam, removed loss scale in adam (#527)
* [zero]added hybrid adam, removed loss scale of adam
* remove useless code
|
3 years ago |
Jiarui Fang
|
8d8c5407c0
|
[zero] refactor model data tracing (#522)
|
3 years ago |
Frank Lee
|
3601b2bad0
|
[test] fixed rerun_on_exception and adapted test cases (#487)
|
3 years ago |
Jiarui Fang
|
4d322b79da
|
[refactor] remove old zero code (#517)
|
3 years ago |
LuGY
|
6a3f9fda83
|
[cuda] modify the fused adam, support hybrid of fp16 and fp32 (#497)
|
3 years ago |
Jiarui Fang
|
920c5889a7
|
[zero] add colo move inline (#521)
|
3 years ago |
ver217
|
7be397ca9c
|
[log] polish disable_existing_loggers (#519)
|
3 years ago |
Jiarui Fang
|
0bebda6ea5
|
[zero] fix init device bug in zero init context unittest (#516)
|
3 years ago |
fastalgo
|
a513164379
|
Update README.md (#514)
|
3 years ago |
Jiarui Fang
|
7ef3507ace
|
[zero] show model data cuda memory usage after zero context init. (#515)
|
3 years ago |
ver217
|
a2e61d61d4
|
[zero] zero init ctx enable rm_torch_payload_on_the_fly (#512)
* enable rm_torch_payload_on_the_fly
* polish docstr
|
3 years ago |
Jiarui Fang
|
81145208d1
|
[install] run with out rich (#513)
|
3 years ago |
HELSON
|
0f2d219162
|
[MOE] add MOEGPT model (#510)
|
3 years ago |
Jiarui Fang
|
bca0c49a9d
|
[zero] use colo model data api in optimv2 (#511)
|
3 years ago |
Jiarui Fang
|
9330be0f3c
|
[memory] set cuda mem frac (#506)
|
3 years ago |
Frank Lee
|
97933b6710
|
[devops] recover tsinghua pip source due to proxy issue (#509)
|
3 years ago |
Jiarui Fang
|
0035b7be07
|
[memory] add model data tensor moving api (#503)
|
3 years ago |
Frank Lee
|
65ad47c35c
|
[devops] remove tsinghua source for pip (#507)
|
3 years ago |
Frank Lee
|
44f7bcb277
|
[devops] remove tsinghua source for pip (#505)
|
3 years ago |
binmakeswell
|
af56c1d024
|
fix discussion button in issue template (#504)
|
3 years ago |
Jiarui Fang
|
a445e118cf
|
[polish] polish singleton and global context (#500)
|
3 years ago |
ver217
|
9ec1ce6ab1
|
[zero] sharded model support the reuse of fp16 shard (#495)
* sharded model supports reuse fp16 shard
* rename variable
* polish code
* polish code
* polish code
|
3 years ago |
HELSON
|
f24b5ed201
|
[MOE] remove old MoE legacy (#493)
|
3 years ago |
ver217
|
c4c02424f3
|
[zero] sharded model manages ophooks individually (#492)
|
3 years ago |
HELSON
|
c9023d4078
|
[MOE] support PR-MOE (#488)
|
3 years ago |
ver217
|
a9ecb4b244
|
[zero] polish sharded optimizer v2 (#490)
|
3 years ago |
ver217
|
62b0a8d644
|
[zero] sharded optim support hybrid cpu adam (#486)
* sharded optim support hybrid cpu adam
* update unit test
* polish docstring
|
3 years ago |
Jiarui Fang
|
b334822163
|
[zero] polish sharded param name (#484)
* [zero] polish sharded param name
* polish code
* polish
* polish code
* polish
* polsih
* polish
|
3 years ago |
ver217
|
9caa8b6481
|
docs get correct release version (#489)
|
3 years ago |
HELSON
|
d7ea63992b
|
[MOE] add FP32LinearGate for MOE in NaiveAMP context (#480)
|
3 years ago |
github-actions[bot]
|
353566c198
|
Automated submodule synchronization (#483)
Co-authored-by: github-actions <github-actions@github.com>
|
3 years ago |
Jiarui Fang
|
65c0f380c2
|
[format] polish name format for MOE (#481)
|
3 years ago |
ver217
|
8d3250d74b
|
[zero] ZeRO supports pipeline parallel (#477)
|
3 years ago |
Sze-qq
|
7f5e4592eb
|
Update Experiment result about Colossal-AI with ZeRO (#479)
* [readme] add experimental visualisation regarding ColossalAI with ZeRO (#476)
* Hotfix/readme (#478)
* add experimental visualisation regarding ColossalAI with ZeRO
* adjust newly-added figure size
|
3 years ago |
Frank Lee
|
83a847d058
|
[test] added rerun on exception for testing (#475)
* [test] added rerun on exception function
* polish code
|
3 years ago |
ver217
|
d70f43dd7a
|
embedding remove attn mask (#474)
|
3 years ago |
HELSON
|
7544347145
|
[MOE] add unitest for MOE experts layout, gradient handler and kernel (#469)
|
3 years ago |
ver217
|
1559c0df41
|
fix attn mask shape of gpt (#472)
|
3 years ago |
ver217
|
3cb3fc275e
|
zero init ctx receives a dp process group (#471)
|
3 years ago |
ver217
|
7e30068a22
|
[doc] update rst (#470)
* update rst
* remove empty rst
|
3 years ago |
HELSON
|
aff9d354f7
|
[MOE] polish moe_env (#467)
|
3 years ago |
HELSON
|
bccbc15861
|
[MOE] changed parallelmode to dist process group (#460)
|
3 years ago |
Frank Lee
|
8f9617c313
|
[release] update version (#465)
|
3 years ago |
Frank Lee
|
2963565ff8
|
[test] fixed release workflow step (#464)
|
3 years ago |
Frank Lee
|
292590e0fa
|
[test] fixed release workflow condition (#463)
|
3 years ago |
Frank Lee
|
90bd97b9c0
|
[devops] fixed workflow bug (#462)
|
3 years ago |
ver217
|
304263c2ce
|
fix gpt attention mask (#461)
|
3 years ago |
ver217
|
fc8e6db005
|
[doc] Update docstring for ZeRO (#459)
* polish sharded model docstr
* polish sharded optim docstr
* polish zero docstr
* polish shard strategy docstr
|
3 years ago |
HELSON
|
84fd7c1d4d
|
add moe context, moe utilities and refactor gradient handler (#455)
|
3 years ago |
Frank Lee
|
af185b5519
|
[test] fixed amp convergence comparison test (#454)
|
3 years ago |