note: this version is only for publication, we will re-organize and clean our code for better research use later

Please follow official websites to set-up environments.

Apart from the environments above, we also need following packages:

pytorch == 1.6.0
lightgbm

You can run the metadl test using this repo directly which will help you handle the environments above automatically.

First, you need to activate your environments:

$ conda activate metadl

Anywhere outside this folder, make a new folder and cd to it. Then you can run the official code:

$ mkdir paths/to/your/runtime/folder
$ cd paths/to/your/runtime/folder
$ python -m metadl.core.run --meta_dataset_dir=abs/path/to/dataset --code_dir=abs/path/to/this/folder

@AutoMLGroup/Meta_Learners