Would sparsegpt be available for Llama2?

[moonlightian]

No description provided.

[eleluong]

So i tested this with llama2 7B chat and ctranslate2, the results are comparable with the original and the inference times are usually lower.

[resorcap]

I tested this with llama2 13B, results are not good.

[Jimskns]

So i tested this with llama2 7B chat and ctranslate2, the results are comparable with the original and the inference times are usually lower.

Hi，what do you mean by lower? Faster or slower?