Skip to main content
brandony
Projects
neox
Reports
Test context parallelism correctness
Log in
Sign up
Share
Comment
Star
Test context parallelism correctness
Brandon Yang
Created on October 1
|
Last edited on October 1
Comment
Section 1
timers/backward-backward
timers/backward-backward
200
400
600
800
1k
Step
0.04
0.06
0.08
0.1
0.12
0.14
0.16
GPU708E-0
GPU708E-0
GPU708E-0
GPU708E-0
timers/optimizer
timers/optimizer
200
400
600
800
1k
Step
0.01
0.02
0.03
0.04
0.05
GPU708E-0
GPU708E-0
GPU708E-0
GPU708E-0
timers/batch generator
timers/batch generator
200
400
600
800
1k
Step
0.01
0.02
0.03
0.04
0.05
0.06
GPU708E-0
GPU708E-0
GPU708E-0
GPU708E-0
timers/backward
timers/backward
200
400
600
800
1k
Step
0.04
0.06
0.08
0.1
0.12
0.14
0.16
GPU708E-0
GPU708E-0
GPU708E-0
GPU708E-0
timers/forward
timers/forward
200
400
600
800
1k
Step
0.5
1
1.5
2
2.5
3
GPU708E-0
GPU708E-0
GPU708E-0
GPU708E-0
timers/backward-allreduce
timers/backward-allreduce
200
400
600
800
1k
Step
-2
-1
0
1
2
GPU708E-0
GPU708E-0
GPU708E-0
GPU708E-0
Run set
4
Run set
4
Run set
4
Run set
4
Run set
4
Add a comment