DiGress
DiGress copied to clipboard
cvignac
code for the paper "DiGress: Discrete Denoising diffusion for graph generation"
Hi Vignac, Thanks for your code! I notice that ConGress normalizes the data for traning, and unnormalizes it for val_nll and sample. Besides, from continuous.yaml, only X is actually normalized....
Hi, I met a question about the regressor model. I think the model does not correctly performed the conditional generation. The error for this command python3 src/guidance/train_qm9_regressor.py +experiment=regressor_model.yaml happens at...
When I use the default setting (qm9 dataset) for model training. During the data preprocessing stage I encounted a problem that the compute node core dumped caused by insufficient memory....
Thank you for your work, this project provides great results and inspiration! This work seems to have chosen to base its processing on discrete graph structures on the QM9 dataset,...
Hi @cvignac , I'm unclear about the sampling part in the code base. Like the `add_noise` step [here](https://github.com/cvignac/DiGress/blob/e88372a837adcac88d24387f3c01d04ae2d36012/src/diffusion_model_discrete.py#L110) I'm wondering if there is a `prev_sample` while performing reverse process? ```...
Hi, I am having some trouble loading the checkpoint file (e.g. checkpoint_planar.ckpt) I cloned the latest version of your code and did: from diffusion_model_discrete import DiscreteDenoisingDiffusion model =DiscreteDenoisingDiffusion.load_from_checkpoint('/.../checkpoint_planar.ckpt',map_location='cuda:0') Then I...
Hello: The work described in this article is wonderful. My research direction is to generate graphs from text. After training the model, how to use text conditions to control the...
Error message: File "/home/patrickli/working/DiGressGuidance/src/diffusion/noise_schedule.py", line 79, in get_alpha_bar return self.alphas_bar[t_int.long()] RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu) Solution: replace return...
in guidance branch, in /DiGressGuidance/src/guidance/main_guidance.py, train_smiles = qm9_dataset.get_train_smiles(cfg, datamodule, dataset_infos) will raise datamodule has no len() Error. **SOLUTION:** replace datamodule with datamodule.train_dataloader()