Releases: lucidrains/magvit2-pytorch
Releases · lucidrains/magvit2-pytorch
0.0.38
fully wire up accelerate
0.0.37
preliminary training loop
0.0.36
just copy over a optimizer factory fn
0.0.35
apply transform to each frame of video
0.0.34
bring in dataset and dataloader
0.0.33
saves the video tokenizer config in the .pt file
0.0.31
save version
0.0.30
add linear attention blocks to discriminator
0.0.28
assert that decoding from code indices works correctly and is the sam…
0.0.27
add linear space attention layer, using a specific one from meta ai