0.0.50

finish and release sparse convolutional causal attention, using pytor…

0.0.49

fix clip loss, thanks to @iceychris

0.0.48

fix for change to using <bos> instead of <eos>

Merge pull request #25 from lucidrains/switch-eos-over-to-bos

fix for change to using <bos> instead of <eos>

0.0.47

Merge pull request #25 from lucidrains/switch-eos-over-to-bos

fix for change to using <bos> instead of <eos>

0.0.46

Merge pull request #22 from lucidrains/switch-eos-over-to-bos

switch over to using <bos> instead of <eos>, assuming fixed length ge…

0.0.45

make sure sparse attention is turned off by default

0.0.44

omit the prefix sections of the sequence undergoing full attention fr…

0.0.43

omit the prefix sections of the sequence undergoing full attention fr…

0.0.42

Merge pull request #20 from lucidrains/prefix-full-attention

omit the prefix sections of the sequence undergoing full attention fr…

0.0.41

0.0.41