Hyunjun Ji

What is the recommended way of using webdataset with pytorch-lightning and ddp?

12

I am trying to build a `LightningDataModule` using webdataset, however I have encountered some difficulties implementing, since previously I have never used `IterableDataset`. The first thing I tried is [this...

Mimalloc support

2

This pull request adds the support for [mimalloc](https://github.com/microsoft/mimalloc), a better memory allocator. It has been claimed by authors that this outperforms other allocators including TCMalloc, but whether this applies to...

ONNXRuntime backend (WIP)

2

### Introduction This backend enables KataGo to run network files in .[onnx](https://onnx.ai/) format, which is an open standard for exchanging neural networks, with [ONNXRuntime](https://microsoft.github.io/onnxruntime/) backend. Currently vesion 8 networks can...

```python def plot_leagues_ratings( df: pl.DataFrame, filename: Union[Path, str], year: int, x_size: int, y_size: int ): title = f"Rating ({year})" df = df.with_columns(pl.lit(0).alias("dummy")) df = df.to_pandas() df["League"] = pd.Categorical( df["League"], categories=["LCK",...

It looks like [GAMI-Tree](https://arxiv.org/abs/2207.06950) is an extension/variant of EBM and GAMI-Net. GAMI-Tree utilizes model-based trees and boosting for describing main and interaction effects instead of neural networks, and there are...

I'm currently trying to rescore poses using the confidence model without running reverse diffusion. When I performed confidence model inference after setting time to zero (like in sampling()), it seemed...

OpenEye license check in OpenEyeToolkitWrapper

2

**Describe the bug** https://github.com/openforcefield/openff-toolkit/blob/fd5a20b6094ac0f64f66510f2791244e9e5ef9b4/openff/toolkit/utils/openeye_wrapper.py#L224 This method detects the presence of OpenEye toolkits licenses by checking if any of toolkit licenses is available. However, the `OpenEyeToolkitWrapper` requires `oechem`, `oeomega`, and `oequacpac`...

Support other clustering algorithms

1

**Is your feature request related to a problem? Please describe.** This is not a problem per se, but `AgglomerativeClustering` and `SpectralClustering` in `sklearn.cluster` is not always favorable especially for large...

Adding JSON I/O for molecules

1

Sometimes it is convenient to exchange molecules in JSON format rather than SMILES or `.sdf` files. Fortunately, `rdChem.Mol` JSON is already implemented in `rdkit.Chem.rdMolInterchange.MolsToJSON` and `rdkit.Chem.rdMolInterchange.JSONToMols`. This can be implemented...

When working with some types of data, I need to reshape arrays with zero elements. For example, ```python import numpy as np import einx import einops arr = np.zeros((0, 0))...