floyd run

Run your project on Floyd.


floyd run [OPTIONS] [COMMAND]


Name, shorthand Default Description
--gpu/--cpu cpu If specified, runs the job on a GPU (G1) instance or CPU (C1) instance. See instance specifications on the pricing page.
--data <ID:mount> ID of the data source to link to. mount specifies the path to mount it at. You can use this parameter multiple times. See data section for more details.
--mode [jupyter|serve] command Specify the mode you want to run the project. The default behavior executes the command you specify. See jupyter and serve sections for more info on them.
--no-open You can disable the CLI from opening the jupyter notebook url. It will print the URL instead.
--env [tensorflow:py3|tensorflow:py2|...] keras:py3 Specify the environment you want to use for your project. See environments for the full list.
--message <message_str> Attach a message to the specific run of the project.
--tensorboard Starts tensorboard in the environment. Tensorboard URL can be found in the dashboard.
command Command to execute when running your project on Floyd.


This command syncs the code tracked by the CLI to the Floyd servers and executes your command. You can see the progress with status command. To view the logs from your code use logs command.


$ floyd run --env tensorflow --gpu "python mnist_cnn.py"
Syncing code ...

$ floyd logs floydhub/projects/lung-cancer/2


Floyd runs standard Docker images for various deep learning frameworks.(See environments for details). If your code requires additional Python dependencies you can specify them in a floyd_requirements.txt file and place it at the root directory of your project. These dependencies will be installed before running your code.


$ cat floyd_requirements.txt
$ floyd run "python train_tf.py -lr 0.01 -output /output/model.bin"

Jupyter notebook

Floyd supports running Jupyter/iPython notebooks on the server. Make sure that the notebook (.ipynb) files are present in the current directory. Use --mode jupyter and you will be presented with a URL to view your Jupyter environment. You do not need to specify a command in this mode. See jupyter page for more details.


$ floyd run --mode jupyter
Path to jupyter notebook: https://www.floydhub.com/notebooks/g8uGRZFQz85meArJGToEcs

Attaching multiple datasets

You can attach upto 5 datasets when you run a project using the run command. You can specify both datasets you uploaded and output datasets of your previous runs. You can specify the mount point also when you specify the data id to mount.


$ floyd run --data udacity/datasets/celeba/1:training --data udacity/datasets/mnist/1:testing "python script.py"
The above datasets will be mounted at /training and /testing respectively.


Floyd can be used to host the model you generated as a REST api. This api can be used to evaluate your model over HTTP. Use --mode serve and you will be presented with a URL to access your API. Floyd currently supports only Flask apps. It runs app.py file and expects the service to run on port 5000. You do not need to specify a command in this mode. See serve page for more details.


$ floyd run --mode serve
URL to service endpoint: https://www.floydlabs.com/expose/mkxjJa46aJBdwP4AEdKxfU

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.