Problem with qW.sample()


#1

When I try to evaluate a sample from the distribution of my neural network weights, I get this error.

w = qW.sample()
w.eval()

NodeDef expected inputs 'float, float' do not match 0 inputs specified; Op<name=MatMul; signature=a:T, b:T -> product:T; attr=transpose_a:bool,default=false; attr=transpose_b:bool,default=false; attr=T:type,allowed=[DT_HALF, DT_FLOAT, DT_DOUBLE, DT_INT32, DT_COMPLEX64, DT_COMPLEX128]>; NodeDef: inference/sample/model/MatMul_715 = MatMul[T=DT_FLOAT, transpose_a=false, transpose_b=false]()

Does anyone know why?

This is qW.sample()

qW1.sample()

<tf.Tensor ‘model_2/Normal_14/sample_8/Reshape:0’ shape=(2, 1) dtype=float32>