ML Timewarp, FluidMorph & Reference Warp — PyBox Release
This is the first release of ML Timewarp, FluidMorph and Reference Warp, packaged as a PyBox for direct use in Flame Batch on both fairly old setups and current systems, have additional controls for initial scale and bi-directional processing as well as model confidence output helpful for spotting.potential artefacts
It can be used on older systems as well as most modern ones.
Installation
Download all archive parts for the CUDA build appropriate for your system, then unpack them with:
cat talosh.twml.v001.cu126.tar.gz.part-* | tar -xzf -Replace cu126 with the CUDA build you downloaded.
Available builds:
- cu118 — for older systems and older NVIDIA driver/CUDA configurations.
- cu126 — recommended for Pascal GPUs such as the Quadro P6000.
- cu128 — for newer GPUs; this build does not support Pascal.
- cu130 — intended for Blackwell-generation GPUs.
Cuda has minor version compatibility so for example 12.2 should be fine with both cu126 and cu128
Make sure all parts belonging to the same build have been downloaded before unpacking.
ML Timewarp
See the attached images for the Batch setup.
Flame's native Timewarp should be set to Timing mode. Duplicate the source and feed the second branch through a Mux node shifted by one frame.
The Timing channel from Flame's Timewarp node should then be linked to the corresponding input on the ML Timewarp PyBox.
Reference Warp
Reference Warp is currently under-trained, but may still be useful in its present state. Updated weights will be released as training progresses.
macOS
Mac versions are coming soon.
Example of batch setup for Timewarp:
Do not forget to link channels:
