-
Hi,
I am an undergraduate student currently experimenting with your models to improve my understanding of LDMs.
Running the following command as suggested by @Warvito to generate sample brain imag…
-
Currently, I guess the inferer is not very usefull to perform the forward pass of the VQ-VAE+transformer during training, since it does not return the target sequence.
For example, in
https://githu…
-
Upon thorough examination of the repository, we believe it could be enriched by the introduction of additional features. These enhancements aim to augment the repository's functionality and extend the…
-
Add code to verify if the number of attentions levels is the same of the number elements in num_channels. Currently, the DiffusionModelUNet is raising the following error message
```
Traceback (mo…
-
I think GEGLU and FeedForward components from diffusion_model_unet.py
https://github.com/Project-MONAI/GenerativeModels/blob/7dcdc9a84e96e25e57325ce9b503b8b39c9801af/generative/networks/nets/diffusi…
-
As mentioned in https://github.com/pytorch/pytorch/issues/86679, F.interpolate is not compatible with bfloat.
It is necessary to convert to float32 its usage in
https://github.com/Project-MONAI…
-
One possible way to the LDM training fails is due to the inappropriate range values that the latent representation of the autoencoder might create. This issue is more crucial in the upscalers, but it …
-
It will be useful to be able to easily compute the likelihood (or some bound on the likelihood, where the exactly likelihood is not available) from all our models, and a prerequisite for some issues, …
-
Hi,
Thanks for providing this amazing tool! I have tried to fit a cognitive model with Softmax as the choice rule in the end. That is, P(option i) = exp(beta * Ui) / [sum(exp(beta * Uj)) for all j]; …
-
Currently, if we chose to use one of the medicalnet pretrained networks and set spatial_dims as 2D, it do not raise any error during the instantiation of the loss object. We should add this to avoid a…