Drop <Float> type parameters which can be inferred

Author: shadaj

Examples/LeNet-MNIST/main.swift

@@ -23,8 +23,8 @@ let dataset = MNIST(batchSize: batchSize)
 // The LeNet-5 model, equivalent to `LeNet` in `ImageClassificationModels`.
 
 let denseDef = AutoDenseDefinition<Float>(outputSize: 120, activation: relu)
-  .then(AutoDenseDefinition<Float>(outputSize: 84, activation: relu))
-  .then(AutoDenseDefinition<Float>(outputSize: 10))
+  .then(AutoDenseDefinition(outputSize: 84, activation: relu))
+  .then(AutoDenseDefinition(outputSize: 10))
 
 var classifier = Sequential {
     Conv2D<Float>(filterShape: (5, 5, 1, 6), padding: .same, activation: relu)