fanfanda / m3ddm Goto Github PK

Can you show me the details of the inference process, because I see that you concatenate corrupted raw video latents, random masked video latent, and masks before feeding them into the 3D UNet network in the training stage, but in the inference, thee should do not exist corrupted raw video latents, and the input dimension should change for the 3D UNet network, Does the network structure of the training and reasoning process change，how can this work?

Meanwhile, if you used the pre-trained diffusion model, if the input dimension change after concatenate, how can you feed it to the
3D UNet network, is there exist a layer for dimensional transformation

nice job! will you release the project code?

Is this the only channel that can be used for trial now? However, this merchant account seems to require a company to register. Can I try it out if I am a student?