Fork me on GitHub
Slack Archive

This page is not created by, affiliated with, or supported by Slack Technologies, Inc.

2019-12-02

Channels

#data-science
<
2019-12-02
>
gigasquid00:12:13

Looks like it’s bringing jvm people together which is a good thing

gigasquid01:12:46

@blueberry - would it be theoretically possible to make a backend for Deep Diamond? like https://github.com/awslabs/djl/issues/2

blueberry16:12:26

Theoretically, probably yes, but I really don't see the point. My idea with DD is to provide simple and low-overhead layer in Clojure that talks directly to Nvidia's and Intel's fast DL operations. I don't see the point of passing through several intermediate layers and doing the same with huge overhead and increase in complexity. Of course, points would be that "supports API X" could be slapped on it, but I have limited resources and have to concentrate on simple and elegant solution that makes sense for small teams.

gigasquid01:12:40

Just thinking out loud trying to see what (if any) opportunities that DJL opens up for the Clojure community. For example, perhaps being able to benchmark RESNET easily between Deep Diamond, MXNet, Tensorflow, and ND4J would be useful.

viesti06:12:01

I think it is good that there is a story for having access to multiple frameworks and a quite direct native binding