why two models here are same

[RQiand]

[sd983527]

it is a temporary change in the converting script, that we haven't tuned paras for tl kernels. In fact, for such small LLMs, i2_s should be faster than TL1 and TL2 so that we recommend just downloading the i2_s gguf model file https://huggingface.co/microsoft/bitnet-b1.58-2B-4T-gguf