DeepSeek/DualPipe
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DualPipe synced 2025-06-26 18:16:46 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

update readme

Browse Source
This commit is contained in:
ljss
2025-03-05 15:51:19 +08:00
parent cfc43619a3
commit f371022947
1 changed files with 8 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
README.md
View File

@@ -19,16 +19,16 @@ DualPipeV is a concise V-shape schedule derived from DualPipe using a "cut-in-ha
![dualpipev](images/dualpipev.png) ![dualpipev](images/dualpipev.png)
Example DualPipeV scheduling for 4 PP ranks and 10 micro-batches. Example DualPipeV scheduling for 4 PP ranks (8 PP stages) and 10 micro-batches.
## Pipeline Bubbles and Memory Usage Comparison (vs. equivalent pp stages) ## Pipeline Bubbles and Memory Usage Comparison (based on the same number of PP stages)
| Method | Bubble | Parameter | Activation | #Devices | | Method | Bubble | Parameter Per Device | Activation Per Device | #Devices |
|-------------|---------------------------------|-----------|------------|----------| |-------------|---------------------------------|----------------------|-----------------------|----------|
| 1F1B | (*PP*-1)(𝐹+𝐵) | 1× | *PP* | *PP* | | 1F1B | (*PP*-1)(𝐹+𝐵) | 1× | *PP* | *PP* |
| ZB1P | (*PP*-1)(𝐹+𝐵-2𝑊) | 1× | *PP* | *PP* | | ZB1P | (*PP*-1)(𝐹+𝐵-2𝑊) | 1× | *PP* | *PP* |
| DualPipe | (*PP*/2-1)(𝐹&𝐵+𝐵-3𝑊) | 2× | *PP*+1 | *PP* | | DualPipe | (*PP*/2-1)(𝐹&𝐵+𝐵-3𝑊) | 2× | *PP*+1 | *PP* |
| DualPipeV | (*PP*/2-1)(𝐹&𝐵+𝐵-3𝑊) | 2× | *PP*+1 | *PP*/2 | | DualPipeV | (*PP*/2-1)(𝐹&𝐵+𝐵-3𝑊) | 2× | *PP*+1 | *PP*/2 |
*PP* denotes the number of pp stages (even). *PP* denotes the number of pp stages (even).
𝐹 denotes the execution time of a forward chunk, 𝐵 denotes the execution time of a 𝐹 denotes the execution time of a forward chunk, 𝐵 denotes the execution time of a