edward2

edward2 copied to clipboard

`TypeError: Dimension value must be integer or None or have an __index__ method, got value '' with type ''` ``` hidden_layer = tf.keras.layers.Dense(16) inputs = tf.keras.Input((224,224,3)) out = hidden_layer(inputs) output_layer...

keertika-11

Internal

copybara-service[bot]

JAX SpectralNormalization shape error with vanilla nn.Dense layer

1

Applying `edward2.jax.nn.SpectralNormalization` to a vanilla Flax `nn.Dense` layer fails with a shape error because the kernel is left-multiplied with the input instead of right-multiplied. On line 143 of `spectral_norm.py` the...

norabelrose

Add joint NLL from Epistemic Neural Networks (https://arxiv.org/abs/2107.08924) to imagenet/* scripts.

1

Add joint NLL from Epistemic Neural Networks (https://arxiv.org/abs/2107.08924) to imagenet/* scripts.

copybara-service[bot]

Add slight speed improvement by avoiding tiling and reshapes. We implicitly tile with the BE's first einsum op. Instead of reshape, we carry around the additional ensemble size axis, and...

copybara-service[bot]

The batch size of my program changes dynamically, but the function 'get_ dropout_ mask_ for_cell' in recurrent.py can only generate a fixed size of ​mask. After run the code like...

ross-Hr

Key error: "Trying to access `split['train']` but `splits` is empty. This likely indicate the dataset has not been generated yet. Thank you for your time.

geshuang307

Internal change

1

Internal change

copybara-service[bot]

Public: Adds util to compute binary predictive posterior variance

2

Public: Adds util to compute binary predictive posterior variance

copybara-service[bot]

changed keras

1

FrancescoPinto