adda1221 opened this issue 10 months ago · 1 comments
It depends how you train the model. If you train the whole model, it would has about 8.2B parameters. If you train vision-related parameters only, it would be about 1B parameters.