This will be the last release for this forked version of cuFHE. This is because we encountered difficulties in extending cuFHE's NTT codes to other dimensions due to numerous hard-coded constants. We are also aware that multiple new GPU implementations of TFHE exist, like TFHE-rs's CUDA backend and HEonGPU.
We are currently planning to develop a successor to cuFHE with an almost identical API, but utilizing different NTT/FFT algorithms for evaluating polynomial multiplications.