The Tema Q development team, team zenei, has developed a new importance matrix method called TaQuants (Tensor-aware Adaptive Quantization) . This model is a TaQuants version of temaq-org/Tema Q-R-4B created with TaQuants v2.0.
--- base_model: - temaq-org/Tema_Q-R-4B pipeline_tag: text-generation tags: - TaQuants - uncensored - non-censored - unfiltered --- # Tema_Q-R-4B TaQuants [The Repository](https://github.com/ek15072809/TaQuants) [Technical Report](https://github.com/ek15072809/TaQuants/blob/main/docs/TaQuants_Technical_Report.pdf) The Tema_Q development team, team zenei, has developed a new importance matrix method called **TaQuants (Tensor-aware Adaptive Quantization)**. This model is a TaQuants version of [temaq-org/Tema_Q-R-4B](https://huggingface.co/temaq-org/Tema_Q-R-4B) created with TaQuants v2.0. The model size and performance are as follows: TaIQ2_M is 0.01GB compressed and shows a 0.96% improvement in PPL compared to IQ2_M. TaIQ3_S has a file size increase of 0.16GB compared to IQ3_S. On the other hand, it shows a 3.43% improvement in PPL compared to Q4_K_M, which is 0.35GB larger.
The Tema Q development team, team zenei, has developed a new importance matrix method called TaQuants (Tensor-aware Adaptive Quantization) . This model is a TaQuants version of temaq-org/Tema Q-R-4B created with TaQuants v2.0.
