edit

floyd data

Manage your data sets on Floyd. The subcommands are:

Command Description
floyd data init Initialize a dataset
floyd data upload Create a new dataset version
floyd data status List all your datasets
floyd data clone Clone an existing dataset
floyd data delete Delete your datasets
floyd data add Add job output to dataset
floyd data output View contents of a dataset

floyd data init

Initialize a Floyd dataset.

Usage

floyd data init DATASET_NAME

Options

Name, shorthand Default Description
DATASET_NAME Name of the dataset (Pick a name from the dataset page in web dashboard)

Description

Floyd can manage your experiment datasets and make them available when running your projects.This command initializes the current directory and tracks all files and subdirectories. Make sure the dataset name you enter here already exists in Floyd. In case the dataset name does not exist, the CLI will open the create dataset page in your browser.

Example

Initialize a floyd dataset in your data directory.

$ cd /data/mnist
$ floyd data init mnist
Data source "mnist" initialized in current directory


floyd data upload

Upload a new version of dataset

Usage

floyd data upload [OPTIONS]

Options

Name, shorthand Default Description
--message <message_str> Attach a message to the specific version of a dataset.

Description

Upload contents of the current directory as a new version of the dataset. This data can now be referred to in the run command. At run time the data will be available at the /input path.

Floyd also versions your data so you can choose any specific version to use in your runs.

Currently this command does NOT respect a .floydignore file. This functionality may be added in the future.

Similarly to flody run command, you can attach a message to the specific version of the dataset.

Example

# e.g. 1
$ floyd data upload
Creating data source. Uploading files ...
NAME
--------------------
alice/mnist-data:1


# e.g. 2
$ floyd data upload --message mnist-edited
Creating data source. Uploading files ...
NAME
--------------------
alice/mnist-data:2
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. See here for more details.


floyd data status

View your datasets on Floyd

Usage

floyd data status [NAME or ID]

Options

Name, shorthand Default Description
NAME or ID Name or ID of your data.

Description

Lists all your datasets on Floyd with more details.

This command can take a shortened job name.

Example

$ floyd data status
DATA NAME                         CREATED         STATUS    DISK USAGE
--------------------------------  --------------  --------  ------------
mckay/datasets/zeroes/1            57 seconds ago  valid     180.0 KB
mckay/datasets/mnist/1             2 minutes ago   valid     10.0 KB

floyd data clone

Clone the output of a job or a dataset.

Usage

floyd data clone [OPTIONS] JOB_NAME or DATASET_NAME

Options

Name, shorthand Default Description
JOB_NAME or DATASET_NAME Name of your job or dataset.
--path <dir_path>, -p <dir_path> Download files in a specific path from job output or dataset.

Description

Use this command to clone an existing job output or floydhub dataset. The output or data used for the job/dataset is downloaded to the current directory. This will override any existing file or directory in the process.

Note: You do not have to be in the project directory to run this command. This command can take a shortened job name.

Example

Example1: Download the full Job.

$ floyd data clone floydhub/projects/deep-photo-styletransfer/4
Downloading the tar file to the current directory ...
Untarring the contents of the file ...
Cleaning up the tar file ...

Example2: Download the content of the models directory from the deep-photo-styletransfer/4 job.

$ floyd data clone -p models floydhub/projects/deep-photo-styletransfer/4
Downloading the tar file to the current directory ...
Untarring the contents of the file ...
Cleaning up the tar file ...


floyd data delete

Delete datasets on FloydHub

Usage

floyd data delete [OPTIONS] [NAMES or IDS]

Options

Name, shorthand Default Description
NAMES or IDS One or more Names or IDs of your data.
--yes, -y False Skip delete confirmation step

Description

Deletes your datasets from FloydHub. This data will no longer be accessible.

Note: You do not have to be in the project directory to run this command. This command can take a shortened job name.

Example

$ floyd data delete mckay/datasets/mnist/1
Delete Data: mckay/datasets/mnist/1? [y/N]: y
Data deleted

floyd data add

Copy job output data to a dataset

Usage

floyd data add ID

Description

Add the contents of a job output to a dataset. This will appear as a new version of data in the dataset. This command is useful if you want to save the output of specific jobs under a dataset. Note: You will be charged for the disk usage separately. This new data can now be referred to in the run command.

This command can take a shortened job name.

Example

$ floyd data add mckay/projects/mnist/1
DATA NAME                         CREATED    STATUS    DISK USAGE
--------------------------------  ---------  --------  ------------
mckay/datasets/mnist/1            just now   valid     10.0 KB
Floyd will generate a new version for the added data. This can be used in your future experiments, if needed, using this data id. See here for more details.


floyd data output

View datasets

Usage

floyd data output [OPTIONS] ID

Options

Name, shorthand Default Description
--url, -u Only print the URL. The data directory can be viewed in the browser.
NAME or ID NAME or ID of your data.

Description

The output command gives the url to access a dataset. This command by default opens the data url in your default browser.

Example

$ floyd data output mckay/datasets/mnist/1
Opening output directory in your browser ...

Help make this document better

This guide, as well as the rest of our docs, are open-source and available on GitHub. We welcome your contributions.