juliuskunze / cwvae-jax Goto Github PK

I reproduced CWVAE on v100 GPU according to the parameters of the github code, but when evaluating, the result obtained under the condition of 100 samples is quite different from the paper. What is the number of samples corresponding to the results in the paper, or are there any other parameter setting details?（i used tensorflow and the batch_size is 50）Hope for your answer, thank you.

I get the following error while running train.py:

I get this error even though ffmpeg is already installed in the environment. Any idea how I can solve this issue?
Thanks.

for level in range(1, self.c.levels): for _ in range(self.c.enc_dense_layers - 1): x = nn.relu(nn.Dense(self.c.enc_dense_embed_size)(x)) if self.c.enc_dense_layers > 0: x = nn.Dense(feat_size)(x) layer = x

line 39 onwards in the cnn.py Encoder(), the depth of these layers increases with the level as the hidden variables is overwritten. At large n_levels and n_enc_dense_layers this will result in a very deep network mapping from the observation embedding to the latent space. Not sure it's intentional, doesn't seem to have a purpose, ie is there a reason the higher latent spaces need a deeper function to map from the embedding?

Same issue in the original tensorflow version vaibhavsaxena11/cwvae#2