Open FanqingM opened 3 weeks ago
We did not intentionally construct interleaved data during the training of the Emu3 series models. So our released post training model can not do interleaved understanding or generation well. You can try interleaved SFT on Emu3 by yourself.
Hello, Great work. Does emu3 can understand multiple images (interleaved image and text) or generation multiple images((interleaved image and text) ? If we can conduct interleaved images and text generation / understanding SFT on emu3?