In 2022 I wrote a paper extending the idea of Character-Aware Neural Language Models to transformer and cross attention, it was rejected so I moved on. Glad to see it is now called Dynamic Tokenization and carried on by other researchers.
ctl
ctl
AI & ML interests
None yet
Recent Activity
commented on
an
article
about 2 months ago
There is no such thing as a tokenizer-free lunch
new activity
about 3 years ago
ctl/wav2vec2-large-xlsr-cantonese:Fix "YAML Metadata Error" by separating language and language_bcp47
updated
a model
about 3 years ago
ctl/wav2vec2-large-xlsr-cantonese
Organizations
None yet
commented on
There is no such thing as a tokenizer-free lunch
about 2 months ago
Fix "YAML Metadata Error" by separating language and language_bcp47
#1 opened about 3 years ago
by
aaronsl
updated
a
model
about 3 years ago