FlashAttentionPrecision
Enums · AiDotNet.NeuralNetworks.Attention
Precision modes for Flash Attention computation.
Fields
| Field | Summary |
|---|---|
FlashAttentionPrecision Float16 |
Use 16-bit floating point (half precision). |
FlashAttentionPrecision Float32 |
Use 32-bit floating point (single precision). |
FlashAttentionPrecision Mixed |
Use mixed precision (FP16 for matmul, FP32 for softmax). |
int value__ |