forked from mrq/DL-Art-School
c0f61a2e15
It might make more sense to have top tokens, then bottom tokens with top tokens having different discretized values. |
||
---|---|---|
.. | ||
__init__.py | ||
kmeans_mask_producer.py | ||
scaled_weight_conv.py | ||
vqvae_3_hardswitch.py | ||
vqvae_3_separated_coupler.py | ||
vqvae_3.py | ||
vqvae_audio_xformer.py | ||
vqvae_no_conv_transpose_hardswitched_lambda.py | ||
vqvae_no_conv_transpose_switched_lambda.py | ||
vqvae_no_conv_transpose.py | ||
vqvae.py |