clj-djl 0.1.8

upgrade to DJL 0.15

clj-djl 0.1.6

• Follow up DJL 0.12.0
• Add metamoph from @behrica

Rename new-xxx functions and remove metric.clj

This version updated the new prefixed function, remove this prefix, for example:

• new-trainer -> trainer
• new-default-training-config -> training-config
• new-xavier -> xavier
• new-array-dataset-builder -> array-dataset-builder
• new-gradient-collector -> gradient-collector
• new-topk-accuracy -> topk-accuracy
• new-accuracy -> accuracy
  ...
  But some of the former form are still kept as an alias. May remove in the future.

Update get-metrics function, to get metrics from trainer as map of sequences.

And add one example in the README

Add lots functions in clj-djl.ndarray, learning, and dataset

Added quite many functions in namespace:

• clj-djl.ndarray
• clj-djl.nn
• clj-djl.training
• clj-djl.training.loss

Add -main func to display available engines

1. lein run to display available engines
2. add cv module to load image from url
3. minor fix to model module

Fix java functional Comsumer interface

• Now you can train MNIST dataset with clojure

Initial release

• deployed to clojars.org
• support mnist dataset training with mlp, see example: https://github.com/kimim/clj-djl-examples