-
I think we can decouple `Metric` from `LossFunction`. The forward method is already different because it does use task weights to reduce the loss. It also should have it's own `__init__` function beca…
-
**Describe the bug**
There are shape mismatch problems with both masking and calculating losses/metrics when using mve or evidential regression networks (MveFFN, EvidentialFFN). This issue is present…
-
Hello guys! I have a question and possibly an issue, What exactly are V_d and how do I implement it into my trainer? I have generated LogP (lipophilicity) and I believe it is an extra atom descriptors…
-
Hello,
In the paper, the reported ∆t value for the 2D dam break dataset is 0.03. In the `metadata.json` file, we can see:
```
"case": "DAM",
"solver": "SPH",
"density_evolution": tr…
iSach updated
1 month ago
-
I want to be able to encode smiles string like [Al].[OH2+]c1ccccc1 which can have any number of molecules or fragments in the smile string.
Can chemprop encode this information giving the informati…
-
per #968, the actual definition of the Noam scheduler does not match up with the v1 implementation. It looks like it isn't necessary to know the total number of steps because it's originally defined a…
-
I'm getting the following RuntimeError when trying to run 'python structure_embeddings/preprocess.py'.
`Traceback (most recent call last):
File "structure_embeddings/preprocess.py", line 316, in…
-
All methods from v1 other than conformal
-
Hello,
I am doing hyperparameter tuning using chemprop_hyperopt for a couple of hyperparameteres. How can I make sure the hyperparameter optimization is converged. I picked number of trials 40 for th…
-
When using attentive aggregation in a given model and then using a trainer that specifies "'''bf16-mixed" precision the following error is given:
```
[127](https://vscode-remote+wsl-002bub…