提交 d3162339 编写于 作者: Y Yang Yang(Tony) 提交者: Abhinav Arora

Update parallel_do.md (#8498)

上级 d50016b2
...@@ -24,7 +24,7 @@ A vanilla implementation of parallel_do can be shown as the following (`|` means ...@@ -24,7 +24,7 @@ A vanilla implementation of parallel_do can be shown as the following (`|` means
``` ```
In the forward pass In the forward pass
| Split input onto different devices | Split input onto different devices
| Copy parameter to onto different devices | Copy parameter onto different devices
|||| Compute forward pass in parallel |||| Compute forward pass in parallel
| Merge output from different devices | Merge output from different devices
...@@ -87,7 +87,7 @@ block2 { ...@@ -87,7 +87,7 @@ block2 {
} }
``` ```
## Proformance Imporvement ## Performance Imporvement
There are serial places we can make this parallel_do faster. There are serial places we can make this parallel_do faster.
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册