Transformers
chemistry
materials
molecules
crystals
diffusion
transformer
latent-diffusion
all-atom
Eval Results (legacy)
Instructions to use chaitjo/all-atom-diffusion-transformer with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use chaitjo/all-atom-diffusion-transformer with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("chaitjo/all-atom-diffusion-transformer", dtype="auto") - Notebooks
- Google Colab
- Kaggle
Checkpoints for QM9 only model?
#1
by ameya98 - opened
Hi Chaitanya, thanks for this implementation! Do you have checkpoints for a model trained on QM9 only?
I'm looking for it on my HPC. I do have wandb logs but can't locate the checkpoint at the moment. It takes 1 day on 2x A100s to train a decent model with validity hitting 96%+ (and 1 day more to get best results) so you could try to retrain yourself.
If its not urgent, I can also train the model for you in about 2 weeks time as I have some PhD-related deadlines.
PS. we also released 10K sampled molecules.