From 97955f063e39c3b580a70ed42c2baa77a0efbeeb Mon Sep 17 00:00:00 2001 From: liaogang Date: Fri, 12 May 2017 00:23:54 +0800 Subject: [PATCH] rename majel --- paddle/{tensor => majel}/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) rename paddle/{tensor => majel}/README.md (99%) diff --git a/paddle/tensor/README.md b/paddle/majel/README.md similarity index 99% rename from paddle/tensor/README.md rename to paddle/majel/README.md index b37ed72216e..55398530567 100644 --- a/paddle/tensor/README.md +++ b/paddle/majel/README.md @@ -9,7 +9,7 @@ In this week, we discussed several potential weaknesses of PaddlePaddle caused b Consequently, we decide to refactor PaddlePaddle step-by-step. First, refactor and replace Matrix/Vector to Tensor, a modern terminology in the deep learning system. Fortunately, we can learn from Majel how to define a Tensor. -To simplify heterogeneous resource allocation in any dimensions (1-9) and types (double, float, float16), Majel consists of several primitives such as `Dim`, `Place` and `Array`, all of them are standard C++ classes. +To simplify heterogeneous resource allocation in any dimensions (1-9) and types (double, float, float16), Majel consists of several primitives such as `Dim`, `Place` and `Array`, all of them are standard C++ class templates. 1. `Place`: memory location [i.e. CPU/GPU]. 2. `Allocation`: heterogeneous resource allocator [i.e. 20MB in GPU]. -- GitLab