Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/seujung/SNAIL-gluon

Implementation of SNAIL(A Simple Neural Attentive Meta-Learner) with Gluon
https://github.com/seujung/SNAIL-gluon

gluon metalearning mxnet

Last synced: about 2 months ago
JSON representation

Implementation of SNAIL(A Simple Neural Attentive Meta-Learner) with Gluon

Host: GitHub
URL: https://github.com/seujung/SNAIL-gluon
Owner: seujung
Created: 2018-07-11T23:36:06.000Z (over 6 years ago)
Default Branch: master
Last Pushed: 2019-02-22T13:40:12.000Z (almost 6 years ago)
Last Synced: 2024-07-31T23:44:54.495Z (4 months ago)
Topics: gluon, metalearning, mxnet
Language: Python
Size: 16.5 MB
Stars: 11
Watchers: 3
Forks: 0
Open Issues: 2
Metadata Files:
- Readme: README.md

Awesome Lists containing this project

awesome-few-shot-meta-learning - code (MXNet? Gluon)
Awesome-MXNet - SNAIL

README

        # SNAIL with Gluon

---

Gluon inplementation of [A Simple Neural Attentive Meta-Learniner](https://openreview.net/pdf?id=B1DmUzWAW)

##### network structore

![net_structure](assets/net_structure.png)

##### building block structure

![block_structure](assets/blocks.png)

## Requirements

- Python 3.6.1

- mxnet 1.3.1

- mxboard 0.1.0

- tqdm 4.29.0

## Application

-  Omniglot

## Usage

- arguments

  - batch_size : Define batch size (defualt=64)

  - epochs : Define total epoches (default=50)

  - N : the nunber of N-way (default=10)

  - K : the number of K-shot (default=5)

  - iterations : the number of data iteration (default=1000)

  - input_dims : embedding dimension of input data (default=64)

  - download : download omniglot dataset (default=False)

  - GPU_COUNT : use gpu count  (default=1)

  - logdir : location of mxboard log file (default=./log)

  - modeldir : location of model parameter file (default=./models)

###### default setting

```

python main.py

``` 

or

###### manual setting

```

python main.py --batch_size=24 --epochs=200 ..

```

## Results

##### 10-way 5-shot case

![perf_acc](assets/perf_acc.png)

## Reference

- https://github.com/sagelywizard/snail

- https://github.com/eambutu/snail-pytorch