Tokenizer accuracy is good. Model is compressed excellently. However, inference is very BAD. Requires changes in BPE. Working on that. Will be dividing BPE file into a sub folder for external API's…

Sparse Matrix and Tensor Visualization Tool for Reorderings

An ultra-fast, GPU-based large graph embedding algorithm utilizing a novel coarsening algorithm requiring not more than a single GPU.

A General-Purpose Counting Filter: Counting Quotient Filter