QuantizationType<T>
Enums · AiDotNet.LoRA.Adapters
Specifies the type of 4-bit quantization to use for base layer weights.
How It Works
Same quantization types as QLoRA. The alternating optimization works with both.
Fields
| Field | Summary |
|---|---|
QuantizationType<T> INT4 |
4-bit integer quantization with uniform spacing (-8 to 7). |
QuantizationType<T> NF4 |
4-bit Normal Float quantization optimized for normally distributed weights. |
int value__ |