-
When I start a retrosynthesis task using the API, predict_automatic_retrosynthesis, can I abort this task after it has started? Sometimes, the task takes hours to run, which affects my subsequent task…
-
Hello. I tried the example using an API key but I get this error. May I know if the API endpoint is still supported. Thanks.
![image](https://user-images.githubusercontent.com/1525257/196396312-7ee2f…
-
Hi, this work is very interesting and I am trying to incorporate the sub-graph vocabulary construction into my work. Here I am wondering about the specific mechanism of the algorithm dealing with the …
-
I tried to build a very simple policy network 'model' which contains only one reaction template, tried to run this using the target from the template and stock containing reactants from the template. …
-
Implementing the G2G model in dataset USPTO-50k you have functions _get_difference and _get_reaction_center and in the second one, you define edge_labels which you use for the training center identifi…
-
Dear Authors,
I'm impressed by the performance on USPTO-50K dataset. However, the top-1 accuracy of USPT-MIT is just 54.1, which is much lower than other papers (i.e., 80 or even 90 accuracy). May …
-
In the [retrosynthesis tutorial](https://torchdrug.ai/docs/tutorials/retrosynthesis.html#)
this code of [synthon completion](https://torchdrug.ai/docs/tutorials/retrosynthesis.html#synthon-completi…
-
Hello, when I tried to predict a retrosynthesis pathway, I got an error `'status': 429, 'error': 'Too Many Requests'`. `get_predict_reaction_results` and `paragraph_to_actions` are able to get correct…
-
It seems like in the source code for G2G center identification (https://torchdrug.ai/docs/_modules/torchdrug/tasks/retrosynthesis.html#CenterIdentification) log softmax and cross-entropy function are …
-
Thanks for releasing `torchdrug` 0.1.3!
I updated to the new version and see improved behavior in many places. However, unfortunately some functionalities that were stable in `0.1.2` are failing n…