granite-3.0-2b-instruct-4bit

4-bit affine quantization of ibm-granite/granite-3.0-2b-instruct, produced with FFAI 0.1.0's ffai convert (mlx-affine format, group_size=64).

Conversion

ffai convert ibm-granite/granite-3.0-2b-instruct --bits 4 \
    --upload-repo ekryski/granite-3.0-2b-instruct-4bit

See also

• FFAI — fast Apple Silicon LLM inference. Model.load("ekryski/granite-3.0-2b-instruct-4bit") runs this checkpoint end-to-end.
• FFAI quickstart
• FFAI quantization docs