PyPI page
Home page
Author:
Asher Feldman
Summary:
KL-divergence scoring of quantized students (MLX safetensors or GGUF) against a full-precision teacher, with a self-managing disk cache of teacher logits
Latest version:
0.1.1
Required dependencies:
datasets
|
huggingface-hub
|
mlx
|
mlx-lm
|
numpy
|
safetensors
|
tqdm
Optional dependencies:
gmlx
|
mlx-kquant
|
pytest
|
ruff
Downloads last day:
0
Downloads last week:
15
Downloads last month:
17
7-day smoothing