Releases: jiahuei/sparse-image-captioning
Releases · jiahuei/sparse-image-captioning
v0.6.0
v0.4.0
Main changes:
- Added Radix tokenizer
v0.3.0
Main changes:
- simplify
forward
method of all models - remove
vocab_size + 1
and change default vocab_size from10000
to10001
- cleanup
train_*.py
- update Config
compat()
function - refactor cached transformer base code
- update transformer
argparse
argument names - remove
--tokenizer_vocab_size
argument - cleanup