GPU Overlay for Tensorflow

If you have a NVidia GPU on Windows or Linux, you may be able to use it to accelerate neural-net processing using Tensorflow. The GPU will need to have 4+ GB of RAM and suitable compute capabilities. (On macOS, CoreML will use the CPU, GPU, or ANE accelerator built into an M1 chip, at its discretion.)

NOTE: This overlay affects neural-net processing only. The default SynthEyes install already has code to use a suitable GPU for reading ARRI, BRAW, or RED files, if enabled in the preferences.

Download

Please select the appropriate version to download:

• SynthEyes 2024.1 for Windows
• SynthEyes 2024.1 for Linux. Recommend not using Chrome to download, it has been known to mangle .gz files!
• Files are subject to Tensorflow and NVidia licenses, see the enclosed license texts.

Windows Installation

• Start with SynthEyes 2024.1 already installed. You must redo this procedure if you reinstall SynthEyes or install a later version.
• Exit any running SynthEyes.
• Download the ZIP file above (approximately 770 MB).
• Open the ZIP file with a Windows Explorer window. You don't need to extract the files!
• Open a second Windows Explorer window to:

C:\Program Files\BorisFX\SynthEyes 2024.1

• Drag the files from inside the ZIP file to this SynthEyes folder.
• Accept the UAC security prompt (you'll need to be an administrator for the machine) to permit the files to be copied there.
• There is already a (CPU) tensorflow.dll in the SynthEyes folder; tell Windows to replace the existing file with the (much larger) one from the winGPU.zip file.

Linux Installation

• Start with SynthEyes 2024.1 already installed. You must redo this procedure if you reinstall SynthEyes or install a later version.
• Exit any running SynthEyes.
• Download the .tar.gz file above (approximately 630 MB). Don't use Chrome! (It tries to handle .gz itself, without removing the extension.)
• Extract the files from the .tar.gz file to a temporary folder.
• Open a terminal window, and change directory (cd) to the temporary folder.
• Be sure to cd into any linuxGPU wrapper folder, so that the directory list (ls) shows the full list of a dozen or so files.
• Identify where SynthEyes is installed, typically /opt/BorisFX/SynthEyes2024.1.
• Run the following command, adjusting as needed if SynthEyes is installed somewhere else. You must have root privileges.

sudo cp -a * /opt/BorisFX/SynthEyes2024.1