NVlabs / LSGM

The Official PyTorch Implementation of "LSGM: Score-based Generative Modeling in Latent Space" (NeurIPS 2021)

NVlabs/LSGM Issues

train_vada.py returns the error: " AttributeError: 'EMA' object has no attribute 'defaults' "
Updated 7 months ago
running evaluate_vada.py, RuntimeError: Address already in use
Updated a year ago1
Precomputing feature statistics: No module named 'fid'
Updated a year ago
Apex installation is successful, but training LSGM (best FID) faield
Updated a year ago
Can I use a simple vae?
Updated a year ago
can't run ‘train_vada.py' for best FID on cifar10
Closed 2 years ago3
Excellent Figure 1!
Closed a year ago5
NLL evaluation issues during training
Updated a year ago
train_vada.py with downloaded pre-trained NVAE checkpoint
Updated a year ago
CUDA out of memory
Updated 2 years ago
Questions about the latent dimension
Updated 2 years ago
Train on afhq dataset
Closed 2 years ago
“--time_eps 0.0" might be "0.01" in readme
Closed 2 years ago
error: unrecognized arguments: --arch_instance_dae res_ho_attn
Closed 2 years ago2
reconstruct image through diffusion
Closed 2 years ago5
Number of GPUs not set correctly during sampling
Updated 3 years ago
RuntimeError: Address already in use
Closed 3 years ago1
reconstruction loss in training_obj_disjoint.py
Closed 3 years ago1