Instructions to use Emma02/vqvae_ckpts with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Emma02/vqvae_ckpts with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Emma02/vqvae_ckpts", dtype="auto") - Notebooks
- Google Colab
- Kaggle
These VQGAN checkpoints have the same hash as the taming transformers VQGAN trained on ImageNet
#1
by prachigarg23 - opened
Hi,
I wanted to use the LAION 7B pretrained VQGAN as mentioned in the LVM paper and code repository. As I was comparing the VQGAN trained on ImageNet (from OpenMUSE) with this checkpoint, I realised all the indices are exactly the same in the codebook and at a closer look realised the hash of the model parameters of both checkpoints is the same. I wonder if anyone else faced this issue and the authors could cross-check if they uploaded the incorrect checkpoint?
Thanks!