Post by account_disabled on Mar 7, 2024 3:47:01 GMT -5
a potential painter to cultivate. Then another question arises. Because everyone knows the superiority of the diffusion model. In addition to enI, there are many friends who are also doing diffusion models. Why does enI look more amazing? Because enI has such a thinking. I once worked on a large language model. I have achieved very good results and achieved such great success. Is it possible for me to use this experience to achieve a new success? The answer is yes. enI believes that its previous success in large language models is
due to the fact that en can be translated into Rich People Phone Number List tokens, tokens, and tokens, which will make it easier to understand some en. It can elegantly combine code, mathematics, and various natural languages. Unify and facilitate large-scale training. So they created the "he concept block" corresponding to en. If en is translated as word understanding, he may be translated by us as "the picture block is used to train the r video model. In fact, the reason why the application of en in large language models is so successful is also due to the rnfrer architecture, which is paired with en. Therefore, r, as a video generation diffusion model, is different from the mainstream video generation diffusion model in that it
adopts the rnfrer architecture. Mainstream video generation and diffusion models mostly use the U-Ne architecture, which means that enI wins in the choice of experience and technical route. But everyone knows that the "successful password" of the rnfrer architecture has become mainstream in text and image generation. Why do others use enI without thinking of using it for video generation? This comes from another problem: the full attention mechanism in the rnfrer architecture Memory requirements will increase quadratically with the length of the input sequence, so the computational cost will be very,