Skip to content

Latest commit

 

History

History
64 lines (51 loc) · 1.79 KB

README.md

File metadata and controls

64 lines (51 loc) · 1.79 KB
Raw

Blender GPU Renders on Headless Servers

(made easy!)

Steps:

  1. Set up a cloud instance running Ubuntu with GPU drivers set up. I recommend Lambda Labs (https://cloud.lambdalabs.com) for pricing and the speed with which you can get going.

1.5. (Skip this step if you're already set up on your server with your .blend file) Get your KeyFile from your Cloud Provider. In the case of Lambda Labs this is under ssh keys on the left-side of the instances dashboard. Create a new key. A ".pem" file should download. Copy the IP address of your instance from the instances dashboardf or the following steps.

Apply correct permissions to the key file:

chmod 600 <YOUR_KEY>.pem

Upload your .blend file to the Ubuntu server with:

scp -i <YOUR_KEY>.pem <YOUR_BLENDER_FILE>.blend ubuntu@<YOUR_IP>:~/

Then ssh back into your server and verify it's there.

ssh -i <YOUR_KEY.pem> ubuntu@<YOUR_IP>
ls 
# you should see your .blend file listed

Remain logged into this session for the remaining steps

  1. Clone this script down with
git clone https://github.com/mpapili/BlenderHeadlessGPUScript.git

-and cd into the directory with:

cd BlenderHeadlessGPUScript/
  1. Run the setup.sh script:
./setup.sh

(optional) verify the install:

which blender
# should return /snap/bin/blender
  1. Render your animation by linking it to the "activate_gpus" script. Assuming your .blend file is in the above directory, you would run:
blender -b -P activate_gpu.py ./<YOUR_BLENDER_FILE>.blend -a
  1. (optional) On your host machine download the rendered files:

Exit from your Lambda Cloud Instance if you haven't already

exit

Use scp and the same key file to download the finished frames

scp -i <YOUR_KEY>.pem "ubuntu@<YOUR_IP>:/tmp/*png" "$(pwd)"

Enjoy!