search

GRAPH-0 / GraphGDP

Implementation for the paper: GraphGDP: Generative Diffusion Processes for Permutation Invariant Graph Generation
MIT License
22 stars 6 forks source link

TypeError: propagate() got an unexpected keyword argument 'query' #4

Closed alvinliu97 closed 6 months ago

alvinliu97 commented 6 months ago

error encountered while runing python main.py --config configs/vp_com_small_pgsn.py --config.model.beta_max 5.0 --mode train --workdir ./test

GRAPH-0 commented 6 months ago

This looks like a problem with torch_geometric. Whether you are using another version of PyG, it is possible that PyG has updated related functions, causing an error.

alvinliu97 commented 6 months ago

Thanks for the prompt reply, yes, I have reinstalled the correct version of PyG and pytorch and it works now. Another question, may i ask that is it normal that if the code stuck here for very long time? i am using 3 NVIDIA RTX A5000, i can see the model gets loaded in to the gpu but the progress show 0%, wonder if it's stuck or just need to wait longer

python main.py --config configs/vp_com_small_pgsn.py --config.model.beta_max 5.0 --mode train --workdir ./x W0312 12:33:05.975298 140046314972992 utils.py:76] No checkpoint found at .x/checkpoints-meta/checkpoint.pth. Returned the same state as input I0312 12:33:06.107235 140046314972992 run_lib.py:118] Starting training loop at step 0.

GRAPH-0 commented 6 months ago

You can limit the use to a single gpu and see. It shouldn't be stuck for too long. In addition, you can adjust config.training.log_freq to see more loss output.

alvinliu97 commented 6 months ago

yes limit the use to a single gpu did the trick for me, thank you very much