[ot][spam][crazy][data] transformer model 'attention' improvement

[Undiscussed Horrific Abuse, Victim & Survivor of]

11:24 now the attentions match :)

now probably will open a pull request with memory-efficient-attention.
plan to descfribe it as a draft looking for comments

commit 0e48ea180a50d201f107c8c69f9493db78cf813b (HEAD ->
memory-efficient-attention, xloem/memory-efficient-attention)
Author: xloem <0xloem at gmail.com>
Date:   Sun Jan 30 11:26:31 2022 +0000

    fix for incorrect mask shape