as this was my very first Embedding, the quality of training images was too diverse.
Completely retrained from scratch, with smaller learning rate and more steps.
This will probably be the last and best version.
NOTE: The trigger word has changed as so the original token will not be evoked
as this was my very first Embedding, the quality of training images was too diverse.
Completely retrained from scratch, with smaller learning rate and more steps.
This will probably be the last and best version.
NOTE: The trigger word has changed as so the original token will not be evoked