Model: saarvajanik/facebook-opt-6.7b-qcqa-ub-16-best-for-KV-cache Source: Original Platform
license, language, library_name
| license | language | library_name | |
|---|---|---|---|
| mit |
|
transformers |
This is a GQA version of the original model facebook/opt-125m. In this version, the original MHA architecture is preserved but instead of having a single K/V head, different K/V heads corresponding to the same group have the same mean-pooled K or V values. It has 16 groups of KV heads per layer instead of original 32 KV heads in the MHA implementation.
Description
Languages
Text
100%