3 ms·
My question was more about usability or interchangeability. Could TF get replaced by Mxnet in a typical deep learning project?
by StreamBright 7y ago
My question was more about usability or interchangeability. Could TF get replaced by Mxnet in a typical deep learning project?
- solidasparagus 7y agoIn terms of features and functionality, they are very much interchangeable. A new project could be written in any of the three major frameworks and be equally good. The only standout feature I'm aware of is that TF has the best support for doing inference on devices, but that won't be true forever. In terms of actually migrating a codebase from one to the other, the APIs are different enough in small ways that it would be a large amount of effort.