metadata
license: apache-2.0
base_model:
- all-hands/openhands-lm-32b-v0.1
pipeline_tag: text-generation
tags:
- mlx
rdsm/openhands-lm-32b-v0.1-mlx-mixed-3_6bit
MLX Mixed Quantization 3/6bits, using --quant-predicate mixed_3_6 ~4bpw
The Model rdsm/openhands-lm-32b-v0.1-mlx-mixed-3_6bit was converted to MLX format from all-hands/openhands-lm-32b-v0.1 using mlx-lm version 0.22.2.
Note: Qwen 2.5 0.5b / 1.5b Instruct seems to work fine as draft models.