Daily Standup July

[nissy-dev]

Summary

JAXChem

In July, I was mainly working for the example notebook and the documentation. Firstly, I made the Tox21 example notebook with Google Colab. The content of this notebook is how to install the JAXChem, how to use a model of the JAXChem, how to train a model and so on. Google colab doesn't require the hard work of building the environment, so I think this is really useful for newcomers. And then, I made the JAXChem documentation and deployed to readthedocs.org.

In addition to this, I tried to refactor the sparse pattern model using jax.ops.segment_sum method in order to fix the performance issue. The reason I thought segment_sum method may solve the performance issue is the jax.ops.segment_sum is used in the GNN model of flax which google develops as a neural network library for JAX. But, this modification didn't resolve the problem because jax.ops.segment_sum method use jax.ops.index_add internally. I will continue to work for resolving this problem.

What I did

• Make a notebook example
  • PRs : Add simple installation docs and notebook exmaple deepchem/jaxchem#9, Tox21 multi task deepchem/jaxchem#10, fix notebook for multitask classification deepchem/jaxchem#11
  • LInk : https://github.com/deepchem/jaxchem/blob/master/notebooks/gcn_property_prediction.ipynb
• Make a brief report during 1st eval period
  • Link : https://forum.deepchem.io/t/gsoc-brief-report-during-1st-evaluation-period/201
• Make a documentation
  • PRs : Add Documents deepchem/jaxchem#12
  • Link : https://jaxchem.readthedocs.io/en/latest/index.html
• Refactor sparse Pattern using segment_sum
  • PRs : Refactor sparse pattern models using the segment_sum deepchem/jaxchem#15

DeepChem

In July, I was mainly working for the nightly-build package and new dataclass for graphs. Firstly, I fixed the nightly-build installation guide and CI scripts. And then, after I improved the type annotation in deepchem.utils, deepchem.dock and deepchem.hyper, I implemented the new dataclass for graphs. Currently, DeepChem has specific data classes for each model (like ConvMol <- GraphConvModel), so I think this is not good situation considering about the maintainability or compatibility to PyG or DGL. (see the details deepchem/deepchem#1942) Now I'm working for the CGCNN support using the new dataclass for graphs.

What I did

• Review PRs
  • PRs : Overhauling the Metrics deepchem/deepchem#1928, [WIP] MoleculeNet contribution template deepchem/deepchem#1938
• Fix pip publish
  • PRs : Fix pip publish deepchem/deepchem#1986, Update requirements docs deepchem/deepchem#1988
• Fix docs build
  • PRs : Fix warnings for docs deepchem/deepchem#1983, Fix doctest error and add docs build status in README deepchem/deepchem#2000, Fix doctest error deepchem/deepchem#2001
• Improve type annotation
  • PRs : add flake8 and improve type annotation in dc.hyper and dc.dock deepchem/deepchem#2027, Improve type of some utils function deepchem/deepchem#2031
• Add new graph data class
  • PRs : Add new molecule graph data deepchem/deepchem#2012
• Add CGCNN featurizer
  • PRs : Modify MoleculeGraphData and add CGCNN featurizer deepchem/deepchem#2045

Next plan (during final evaluation period)

In August, I will continue to work for implementing the CGCNN model and the PyG or DGL support for DeepChem. This plan is different from the original plan, but I think this is a really useful for many newcomers which want to try ML with chemistry. And then, I will restart to work for JAXChem after finishing this crystal support.

• Implement the CGCNN featurizer (finish Modify MoleculeGraphData and add CGCNN featurizer deepchem/deepchem#2045) (7/27 - 8/5)
• Implement CGCNN model using DGL or PyG (8/3 - 8/9)
• Add examples script for CGCNN (8/10 - 8/16)
• Add other examples using DeepChem with DGL or PyG (8/17 - 8/23)
• Prepare for the final submit (8/24 - 8/30)

[nissy-dev]

7/2

Yesterday

I write evaluations and issues. deepchem#8, deepchem/deepchem#1941, deepchem/deepchem#1942

Today

I will make a notebook example

blockers

I think it is no bockers.

[nissy-dev]

7/4

Yesterday

I mad the notebook example

Today

I will write the blog post based on this comment #1 (comment).

blockers

I think it is no bockers.

[nissy-dev]

7/7

Yesterday

I posted the blog!
https://forum.deepchem.io/t/gsoc-brief-report-during-1st-evaluation-period/201

Today

I will build the documents using sphinx.

blockers

I think it is no bockers.

[nissy-dev]

7/8

Yesterday

I built the document using sphinx.

Today

I'll host the document using the https://readthedocs.org/.
And, I'll start looking into the Material Project Dataset support.

blockers

I think it is no bockers.

[nissy-dev]

7/9

Yesterday

I hosted the document using the https://readthedocs.org/.
Link : https://jaxchem.readthedocs.io/en/latest/

Today

I'll fix the deepchem installation docs and review some PRs.
And, I'll start looking into the Material Project Dataset support.

blockers

I think it is no bockers.

[nissy-dev]

7/13

Yesterday

I didn't work in 7/11, 7/12

Today

I will refactor the sparse pattern GCN. Currently, JAXChem uses the jax.ops.index_add for aggregating the neighbor nodes and this leads to a bad performance.

But, two day ago, I found the flax has the example of sparse GCN model. (it includes MPNN!) The flax uses the jax.ops.segment_sum . jax.ops.segment_sum is a almost same as the torch-scatter, I think it is better to use the jax.ops.segment_sum. (torch-scatter is used in PyG) And, I will try to implement the MPNN.

Flax GCN example : https://github.com/google/flax/blob/9cc9b042c52f394464a2f292f6764e1e2836a01a/examples/graph/models.py

About the Material Project Dataset support, I seem that Nathan is working hard deepchem/deepchem#1999. So, after Nathan added the new materials datasets to molnet, I will try to implement the CGCNN.

blockers

I think it is no bockers.

[nissy-dev]

7/16

Yesterday

I refactor the sparse pattern GCN using jax.ops.segment_sum The PR is deepchem#15
This modification didn't solve the issue because jax.ops.segment_sum also uses jax.ops.index_add internally...

Today

I implement the new molecule graph data and refactor some featurizers and ci scripts.
After merging these PRs, I will make the new graph model featurizers using them

blockers

I think it is no bockers.

[nissy-dev]

7/18

Yesterday

In a few days, I improved the type annotation and keep codes more healthy. But, these works are not related to jaxchem... So, I am regretting my actions....

Today

In this week, I will focus on the CGCNN support. I will implement the CGCNN featurizer and molnet function for Materials Project Data.

My GSoC goal is supporting inorganic crystal data for DeepChem and fixing the sparse pattern models.

blockers

I think it is no bockers.

[nissy-dev]

7/23

Yesterday

In a few days, I couldn't work for GSoC because of my research work....

Today

Today, I added the CGCNN featurizer using new graph data class.

My GSoC goal is supporting inorganic crystal data for DeepChem and fixing the sparse pattern models.

blockers

I think it is no bockers.