File size: 989 Bytes
3d4d40d
29c9647
cf49f13
 
 
 
3d4d40d
29c9647
2a9343e
 
 
 
 
3d4d40d
1
2
3
4
5
6
7
8
9
10
11
12
13
14
# Pipeline Parallellism with Controllable Memory

Pipeline Parallelism with Controllable Memory creates a framework on designing pipeline schedules and uses the framework to find memory optimal efficient schedules.

From our findings, we need approximately 1/3 memory under ideal conditions (F, B and W have same runtime), and 1/2 memory to create zero bubble schedule in realistic scenarios (with the necessary condition being W + 2B ≥ 2F and W + 2F ≥ 2B ).

Check out our paper at [Arxiv](https://arxiv.org/abs/2405.15362).

| Comparison assuming T_F=T_B=T_W                       | 1F1B    | V-Min  | V-Half     | V-ZB |
| ----------------------------------------------------- | ------- |------- | ---------- | ---- |
| Bubble Rate                                           | (p-1)/m | ????   | ???        |   0  |
| Activation Memory <br> (Compared to 1F1B)             |    p   |  (p+4)//3 | (p+2)//2 |   p  |

Bubble Rate here is calculated as (1 - longest stage time/(F+B+W)/m).