You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@horn.apache.org by "Edward J. Yoon (JIRA)" <ji...@apache.org> on 2016/04/20 08:30:25 UTC

[jira] [Commented] (HORN-13) Convert MNIST dataset into a sequence file

    [ https://issues.apache.org/jira/browse/HORN-13?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15249360#comment-15249360 ] 

Edward J. Yoon commented on HORN-13:
------------------------------------

You'll be able to convert the MNIST dataset like below:

% mvn clean install -DskipTests
% bin/horn jar target/horn-0.1.0-SNAPSHOT.jar org.apache.horn.utils.MNISTConverter
Usage: TRAINING_DATA LABELS_DATA OUTPUT_PATH
ex) train-images.idx3-ubyte train-labels.idx1-ubyte /tmp/mnist.seq

then, train with following commands:

# MNIST image feature dimension (img shape: 28*28)
# and label dimension 10 (0-9 digits)

% bin/horn jar target/horn-0.1.0-SNAPSHOT.jar org.apache.horn.examples.MultiLayerPerceptron model_path .... 784 10 ...


> Convert MNIST dataset into a sequence file
> ------------------------------------------
>
>                 Key: HORN-13
>                 URL: https://issues.apache.org/jira/browse/HORN-13
>             Project: Apache Horn
>          Issue Type: Sub-task
>            Reporter: Edward J. Yoon
>            Assignee: Edward J. Yoon
>
> MNIST digit classification will be used as a MLP example. In this task, I'll add util that converts MNIST dataset into a HDFS sequence file format.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)