VisualInference's picture
Inkling 2 Mini for MLX, 3bit: weights, card, validation results
e0b70af verified
Raw History Blame Contribute Delete
391 Bytes
{
"recipe": {
"experts": {
"default": "gptq:3:64"
},
"dense": {
"attn": [
8,
64
],
"embed": [
8,
64
],
"head": [
8,
64
],
"vision": [
8,
64
]
}
},
"calibration_tokens": 49465,
"method": "GPTQ (error-compensated rounding), affine, group size 64; scales and offsets optimised at float16 precision and stored as bfloat16"
}