mirror of https://github.com/hpcaitech/ColossalAI
update GPT-2 experiment result (#666)
parent
17e73e62cc
commit
ce8a3eae5b
|
@ -86,9 +86,10 @@ Colossal-AI为您提供了一系列并行训练组件。我们的目标是让您
|
|||
|
||||
- 降低11倍GPU显存占用,或超线性扩展(张量并行)
|
||||
|
||||
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/Colossal-AI%20with%20ZeRO.jpg" width=393>
|
||||
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/(updated)GPT-2.png" width=800>
|
||||
|
||||
- 能训练接近11倍大小的模型(ZeRO)
|
||||
- 用相同的硬件条件训练24倍大的模型
|
||||
- 超3倍的吞吐量
|
||||
|
||||
### BERT
|
||||
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/BERT.png" width=800/>
|
||||
|
|
|
@ -87,9 +87,10 @@ distributed training in a few lines.
|
|||
|
||||
- 11x lower GPU memory consumption, and superlinear scaling efficiency with Tensor Parallelism
|
||||
|
||||
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/Colossal-AI%20with%20ZeRO.jpg" width=393>
|
||||
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/(updated)GPT-2.png" width=800>
|
||||
|
||||
- 10.7x larger model size on the same hardware
|
||||
- 24x larger model size on the same hardware
|
||||
- over 3x acceleration
|
||||
### BERT
|
||||
<img src="https://raw.githubusercontent.com/hpcaitech/public_assets/main/colossalai/img/BERT.png" width=800/>
|
||||
|
||||
|
|
Loading…
Reference in New Issue