1 write to _scaling
Microsoft.ML.TorchSharp (1)
NasBert\Modules\MultiHeadAttention.cs (1)
74_scaling = Math.Pow(_headDim, -0.5);
1 reference to _scaling
Microsoft.ML.TorchSharp (1)
NasBert\Modules\MultiHeadAttention.cs (1)
200q.mul_(_scaling);