提交 9b167504 编写于 作者: D dzhwinter

"remove AllReduce2 comments"

上级 ebd0cf14
......@@ -53,8 +53,13 @@ Need to notice that AllReduce operator force GPUs synchronized at that point. Th
As it shown in the picture, 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`.
- **AllReduce2**
If we use the NCCL2 AllReduce primitive, every GPU optimized full batch of data, wasted (n-1) GPU compute resources. In addition, AllReduce will only utilize the communicate resource during synchronization, then update the gradient will be a seperated phase. In fact, we can amortize the update gradient time cost into the communicating phase.
- Every parameter has its root card. That card will call **Reduce** operator and collect the gradients from GPUs.
- The whole model's parameter will be hashed to different root card, ensure the load balance between GPUs.
Then we have another version AllReduce operator. Other part keep the same with before.
- **AllReduce**
Need to note that our AllReduce operator is a ring-base AllReduce implementation. If we use the NCCL2 AllReduce primitive, every GPU optimized full batch of data, wasted (n-1) GPU compute resources. In addition, NCCL2 built-in AllReduce will only utilize the communicating resource during synchronization, then update the gradient will be a subsequent phase. In fact, we can amortize the update gradient time cost into the communicating phase. The process is
1. Every parameter has its root card. That card will responsible for aggregating the gradients from GPUs.
2. The whole model's parameter will be hashed to different root card, ensure the load balance between GPUs.
3. Logically neighberhood card will start send parameter to the next one. After one round, the parameter main card will aggregate the full gradients.
4. Then the root card will optimize the parameter.
5. This parameter card will send its optimized result to its neighberhood, then the neighberhood will send parameter to its next one.
6. Finish the sychronization round.
The total time cost will be 2 * (n-1) * per-parameter-send-time, we reach the goal of amortize the upgrade time into communicating phase.
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