john-bradshaw/synthesis-dags

Code for our paper "Barking up the right tree: an approach to search over molecule synthesis DAGs"

Hi I'm John! 👋

Most recently I was a postdoc in the Coley Research Group at MIT. Before that, I was a PhD student in Machine Learning at the University of Cambridge/Max Planck Institute for Intelligent Systems. On my GitHub you can find the code for my papers as well as my implementations for some general ML algorithms. Some highlights include:

📄 Code for Papers

Synthesis DAGs is code for our paper "Barking up the right tree: an approach to search over molecule synthesis DAGs".
Molecule Chef is code for our paper "A Model to Search for Synthesizable Molecules".
Electro is code for our paper "A Generative Model For Electron Paths".
GPDNN (and here) is a gist demonstrating some of the key concepts of our paper "Adversarial Examples, Uncertainty, and Transfer Testing Robustness in Gaussian Process Hybrid Deep Networks".

🧺 Misc

ML for Molecules is a notebook I wrote for a tutorial about ML for molecules.
GNN is a library for graph neural networks in PyTorch.
kriggie is a playground for exploring Gaussian processes.

(Note some of this work has been done in collaboration or been heavily inspired by work elsewhere, which I have tried to highlight on the repos' READMEs).

synthesis-dags's Issues

Molecular Transformer failed to run

Hi John, thanks for updating the Readme. I really appreciate it!

I have encountered a new issue and I'm wondering whether you could help me with it. When I'm running the functional_tests/reaction_predictor_server_checker.py, there is an error occurred. The return value of the request is: {'error': 'Runtime Error: cublas runtime error : the GPU program failed to execute at /opt/conda/conda-bld/pytorch_1533672544752/work/aten/src/THC/THCBlas.cu:249', 'status': 'error'}
The error is: Traceback (most recent call last):
File "functional_tests/reaction_predictor_server_checker.py", line 48, in
main()
File "functional_tests/reaction_predictor_server_checker.py", line 42, in main
print(rp(reactants))
File "/rds/user/yo279/hpc-work/project/synthesis-dags/syn_dags/model/reaction_predictors.py", line 55, in call
self._run_list_of_reactant_sets(reactant_sets_new_needed)))
File "/rds/user/yo279/hpc-work/project/synthesis-dags/syn_dags/model/reaction_predictors.py", line 134, in _run_list_of_reactant_sets
raise ex
File "/rds/user/yo279/hpc-work/project/synthesis-dags/syn_dags/model/reaction_predictors.py", line 131, in _run_list_of_reactant_sets
op_back = return_list[0]
KeyError: 0
Besides, the training code for synthesis-dags can run successfully, while the sampling code also has the error above.
Thanks for your assistance!

Google Drive for Molecule Transformer doesn't exist

Hi John, the link of Google Drive for the pertrained Molecule Transformer is invalid. Could you please update the link?
Thanks a lot.

Breaking changes to protobuf

Hi, I just installed this on a new machine and found one issue regarding the conda_dogae_gpu.yml environment setup. Namely, it seems like Google recently introduced backwards-incompatible changes to protobuf:

https://developers.google.com/protocol-buffers/docs/news/2022-05-06#python-updates

I'm not exactly clear on what changed, but the suggested option of downgrading to protobuf==3.20 (via pip) worked for me. There may be a better option (… regenerating some _pb2 files?), but otherwise would suggest updating the environment to pin version on protobuf.