- Add support for multi-process and multi-threaded job starting and polling using vollo-rt
- Starting and polling jobs in vollo-rt is now faster
- Add support for programs containing multiple models
- Support for FP64 output (by compiling a .onnx model with FP64 leaves)
- Jobs might now be started upon calling
vollo_rt_add_job, instead of being saved up for a call to vollo_rt_poll
- Add support for the AMD V80 and V80LL accelerator cards
- Include
ami_tool and the AMI kernel driver for V80 boards
- Trees now use fp32 for inputs, branches and leaves
- Support for increased tree depth
- First release of decision tree accelerator