@@ -53,6 +53,10 @@ These two operators need the Multi-GPU context support.
...
@@ -53,6 +53,10 @@ These two operators need the Multi-GPU context support.
Need to notice that Allreduce operator force GPUs synchronized at that point. Every device only need runs sub-graph in a loop style forever, the whole training process in asynchronous or synchronous mode depends on the Allreduce point in the graph.
Need to notice that Allreduce operator force GPUs synchronized at that point. Every device only need runs sub-graph in a loop style forever, the whole training process in asynchronous or synchronous mode depends on the Allreduce point in the graph.
For the simplest implement, when each GPU compute the gradient of `W`, followed with a `AllReduce` operator, accumulate the `dW` to full batch of data, then run the optimize process individually and apply the gradient to its `W`.
In fact, in the way of every GPU optimized full batch of data, wasted (n-1) GPU compute resources. We will enhance it in the next stage.
### Benefits
### Benefits
- can easily move the optimize sub-graph to parameter server, multi-GPU feature can be compatible with distributed support design.
- can easily move the optimize sub-graph to parameter server, multi-GPU feature can be compatible with distributed support design.
