Wiki C4 PIQA ARC-E ARC-C HellaSwag Wino Avg.
Unquantized 6.1 9.2 79.9 80.1 50.4 60.2 72.8 68.6
W4G64 6.11 9.38 79.33 79.79 49.74 59.22 73.95 68.41
W3G64 7.13 11.06 78.78 76.22 44.37 56.69 70.32 65.28

Revisions available in this repository:

  • main (W4G64, scales learned);
  • nfl_w3g64 (W3G64, scales learned);
  • nf_w4g64 (W4G64, scales not learned);
  • nf_w3g64 (W3G64, scales not learned);

Evaluations are provided for models with learned scales.
Benchmark scores (zero-shot) are computed with lm-evaluation-harness.

Downloads last month
15
Safetensors
Model size
2.9B params
Tensor type
FP16
·
F32
·
I16
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Collection including radi-cho/Meta-Llama-3-8B-FLUTE