sfleisch

GAN.py does not work when using tf.keras instead of the standalone keras

2

Get: InvalidArgumentError: You must feed a value for placeholder tensor 'discriminator_input' with dtype float and shape [?,28,28,1] [[{{node discriminator_input}}]] [[{{node loss_1/model_loss/broadcast_weights/assert_broadcastable/is_valid_shape/has_valid_nonscalar_shape/has_invalid_dims/concat}}]] I have not been able to trace this down...

Is mixed precision, mixed_float16, not supported for certain augmentations

4

I found that certain augmentations fail when the mixed precision policy is set to mixed_float16 but work with float32. So far I've found this to be true for RandomColorDegeneration and...