Closed PawelLipski closed 4 years ago
Non-idiomatic stuff to change:
Should:
def astype[U](newType: NumPyType[U]): NDArray[U]
def float32: NumPyType[Float]
model
accuracy
sigmoid
Reader
Union
_enter_()
tensorflow.*
def AdamOptimizer(lr: Double): Optimizer
def AdamOptimizer(lr: Double): AdamOptimizer
s/Example/GradientDescentOptimizerExample
Could:
apply_gradients
Variable
NDArray
~the examples are almost identical for Scala 2.x and 3.x, possibly their code could be shared on some level~ moved to a separate issue
Non-idiomatic stuff to change:
Should:
def astype[U](newType: NumPyType[U]): NDArray[U]
+def float32: NumPyType[Float]
model
,accuracy
,sigmoid
- make enums instead of stringsReader
instance forUnion
s - auto derivation (?)_enter_()
- with contextstensorflow.*
)def AdamOptimizer(lr: Double): Optimizer
=>def AdamOptimizer(lr: Double): AdamOptimizer
etc.s/Example/GradientDescentOptimizerExample
for consistencyCould:
apply_gradients
(and maybe other operations) are mutable - they changeVariable
sNDArray
- make more type safe? not clear how this can be done