0.0.24

use the kv cache logic from x-transformers library

0.0.23

even simpler

0.0.22

complete eos logic given 3 vertices with D quantized sequence

0.0.21

complete eos logic given 3 vertices with D quantized sequence

0.0.20

logic was not correct for can_eos

0.0.19

make sure commitment loss does not account for null vertex token. pla…

0.0.18

recon loss must account for variable lengthed faces

0.0.17

already checked for in vq library

0.0.16

works like a charm

0.0.15

able to derive face mask directly from codes based on padding id, for…