0.0.38

allow for any modality to be separately trained

0.0.37

just give text its own forward for the AdaptiveWrapper as well

0.0.36

create a separate forward for language model pretraining

0.0.35

make sure prompting works, and force tests to start off with modality…

0.0.34

What's Changed

• test flex attn by @lucidrains in #10

New Contributors

• @lucidrains made their first contribution in #10

Full Changelog: 0.0.33...0.0.34

0.0.33

start applying unet skip connections in transformer settings more bro…

0.0.32

prepare for mixing different encoded shapes (audio, video) into the mix

0.0.31

get the overall logic of key value caching working, hide behind a fla…

0.0.30

auto move tensors within modality sample to the correct device

0.0.29

Full Changelog: 0.0.28...0.0.29