Cool! So you ran your first job and got the output of nvidia-smi. Now let's run a real neural network. We are going to use a GPU to transform a video into a stylized video.
Create your project
- In a terminal run
git clone https://github.com/Paperspace/fast-style-transfer && \ cd fast-style-transfer
- You should now be in the fast-style-transfer directory. Run: paperspace project init
- This project expects a small movie file that will be converted into a "style transfered" movie through the magic of GPU-powered neural networks!
- The project comes with the fox movie shown above but if you want to use your own, you will need to download a movie and add it to the folder './examples/content/'. You will also need to update line 20 in the run.sh file to point to your movie file instead of the default.
Run the job
- In a terminal run:
paperspace jobs create --container Test-Container \ --machineType P5000 --command "/paperspace/run.sh"
- You will start seeing logs in your terminal. To stop the logs you can send a CTRL+C key combination and it will stop tailing the logs or you can just wait for the job to complete and the log output will stop at that point.
- You can also visit www.paperspace.com/console/jobs to view the status of your job.
Get your output
- Once the job has completed it will have put a movie file in to the /artifacts directory of the container. Anything in this folder gets uploaded and you are given a private link to the file.
- You can tell that a job has an output artifact by looking for the yellow gift box icon.
- Scroll down to the File Output (Artifacts) section and you should see your outputted movie file.