Floyd is a great cloud computing platform.I have used it for running a lot of my deep learning models. The main advantage of using Floyd is that you don't require a Credit Card and that is the main reason for giving a try to Floyd.

Floyd has a pretty handy documentation at Floydhub which is updated at regular intervals.

Initiallly I wrote this blog when there were a lot of confusions but now it is a very crisp and easy to use docs page. This post contains steps to Get up and running with Floyd on Windows.

Python Machine Learning Setting up floyd

Setting up floyd for using on Windows.

Account

Make your account at Floydhub.

You can get a free account which gives 10GB of storage and 20 hours of CPU time and 2 hours of GPU time. You can upgrade to any of their paid plans if you want to continue using their services.

Installation

$ pip install -U floyd-cli

Login

$ floyd login

Copy and paste your authentication token in your terminal

Run project

Make a project folder

$ mkdir myproject

Change to project folder

$ cd myproject

Make a folder for data

$ mkdir mydata

Make a folder for code

$ mkdir mycode

Paste all your data in mydata

Paste all your code in mycode

Change to data directory

$ cd mydata

Initialise data

$ floyd data init sent-data

Then you can upload your dataset to Floyd.

$ floyd data upload

Floyd will generate a data id for the uploaded dataset. This uploaded dataset can be used in your future experiments, if needed, using this data id.

Output

Creating data source. Uploading files ...
DATA ID                 NAME                    VERSION
----------------------  --------------------  ---------
GY3QRFFUA8KpbnqvroTPPW  alice/sent-data:1            1

Change to main directory

$ cd ..

Change to code directory

$ cd mycode

Run the code

$ floyd run --data <DATA_KEY>:<MOUNT> --env theano-0.9:py2 "python filename.py"

Example code

$ floyd run --data GY3QRFFUA8KpbnqvroTPPW:mydata --env theano-0.9:py2 "python mlp.py"

Check the status of code

$ floyd logs <RUN_ID>

Some common problems:

Should I put my code and data in separate folders

NOTE: Adding the codes in a separate folder prevents re-loading of data in the memory of floyd server. Floyd server synchronizes the WHOLE FOLDER in which the code is located when you run the code using the above command, which is actually a waste of (upload) time and a problem of data redundancy as you have already uploaded the data on the server.

How to link data in my script

x_data = np.load('/nfiles/fileone.npy')
y_data = np.load('/nfiles/filetwo.npy')

Here, nfiles is my MOUNT point.

If you are using Anaconda prompt and you come across the following error after running the command: "pip install -U floyd-cli": scandir could not be installed

$ conda install -c conda-forge scandir=1.5

Contributors:

Aditya Arora
Piyush Jha
Akshita Gupta

Floydhub Documentation