Text-to-Image AND Image-to-Text: CM3leon explained

[@All_u_need_channel]

Good work, I appreciate your hard work and beneficial contents.

[@borismeinardus]

I really appreciate that! 😊

[@im._abhi]

can u guide on the what is reverse image search?

[@carsongutierrez7072]

Wow! ML bro~

[@borismeinardus]

🤩🧡

[@al-aminibrahim1394]

awesome, so the image tokenizer they use is VQGAN ?. according to the paper they said they used image tokenizer from Gafni et al. (2022a) MUSE paper, wich according to the MUSE too they said they used VQGAN. so does that mean CM3Leon is using VQGAN as image tokenizer?. thanks

[@borismeinardus]

Hey! Yes, it seems like they are using a modified version of the VQ-GAN approach as their tokenizer. At least the encoder part (generator) of a VQ-GAN trained generator and discriminator pair.