filter_shape = (h_filter, w_filter, channels_in, channels_out)
filter_values = np.random.normal(size=filter_shape).astype(np.int32)
inp = Int32Tensor(tf.constant(input_conv))
out = inp.conv2d(Int32Tensor(tf.constant(filter_values)), strides)
with tf.Session() as sess:
actual = sess.run(out.to_native())
After Change
filter_values = np.random.normal(size=filter_shape).astype(np.int32)
inp = int32factory.tensor(input_conv)
out = inp.conv2d(int32factory.tensor(filter_values), strides)
with tf.Session() as sess:
actual = sess.run(out.to_native())
