質問編集履歴
1
追加
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -7,6 +7,10 @@
|
|
|
7
7
|
|
|
8
8
|
|
|
9
9
|
---
|
|
10
|
+
|
|
11
|
+
<<https://www.quora.com/What-are-the-differences-between-static-graph-frameworks-like-Tensorflowand-Caffe-and-dynamic-graph-frameworks-like-PyTorch-and-Chainer>>
|
|
12
|
+
|
|
13
|
+
|
|
10
14
|
|
|
11
15
|
It uses dynamic graph creation, which enables more flexibility in computation. For example, imagine you’re creating a network like HyperNetworks (by David Ha) or “Thin Nets”, the weight matrix of **__the graph would be generated statically, this would massively slow down the framework__**. The ability of PyTorch(『static graph』) to do this on the fly allows this process to go very smoothly.
|
|
12
16
|
|