---
license: gemma
---

WARNING: `gemma-2-27b` models don't run well in `float16` precision.<br>
This FLUTE-quantized model is released in `bfloat16`.

|       | Wiki | C4    |
| ----- | ---- | ----- |
| W4G64 | 5.91 | 9.71  |
| W3G64 | TBD  | TBD   |


Evaluations are provided for models with learned scales.<br>
Check the base [`gemma-2-27b-FLUTE`](https://huggingface.co/radi-cho/gemma-2-27b-FLUTE) for `lm-eval-harness` benchmarks.