Runway uses Docker to run models locally on your computer. If you are interested in running models locally, you'll have to install a version of Docker for operating system. Docker will need to be running at the same time that Runway is in order for Runway to interact with it to run models.

General Errors

If you are receiving docker-related errors, try the following:

  1. Close Runway
  2. Re-start Docker
  3. Open Runway

Adding Memory to Docker

Several models require a significant amount of memory in order to run properly. By default, Docker allocates a maximum of 2 GB of memory to run models. Some RunwayML models exceed that, which may lead to errors. Without enough memory models may misbehave or fail silently without a helpful error message. To prevent that, you need to increase Docker's memory limit. 

Open Docker's Preferences:

Go to the Advance tab:

Increase the Memory slider to 8.0 GB, then click Apply & Restart.

Find related instructions here for increasing the amount of memory that Docker allocates to models.

Running Docker without Sudo

Runway needs to be able to communicate with the Docker daemon without sudo  privileges. This should work correctly on MacOS and Windows, but may need to be manually configured on Linux. See the Running Docker without Sudo support article to configure this properly.

Server Error: Drive has not been shared 

When running models locally on Windows, your home drive (usually called the C drive) must be shared with Docker. Open the Dockers window app and select Shared Drive, then ensure that your home drive is shared with Docker. Once it's shared, you should no longer receive the 500 Server Error: Drive has not been shared error when running models locally.

Still Not Working?

If the above solutions don't fix the problem, you can chat our team directly with our team using the chat button in the app or by clicking the small button on the bottom right of the support.runwayml.com page. 

Alternatively, you don't need to install Docker to run models remotely, so you could try that instead!

Did this answer your question?