机器学习

数据库

• COD | Crystallography Open Database

• Materials Project

• JARVIS

1

pip install jarvis-tools

• ASE

1

pip install ase

• Pymatgen

  可与Materials Project对接，但结构转换不如ase准确。

1

pip install pymatgen

• atomate

• atomate2

• AiiDA

• Jobflow

结构搜索

• USPEX

  遗传算法。

• JAMIP/CALYPSO

  粒子群优化。

描述符

• Matminer

1

pip install matminer

• SISSO

• DScribe

1

pip install dscribe

• RDKit

  根据分子SMILES，获得TPSA、logP等描述符。

  在RDKit基础上开发的qed。

• XenonPy

• PF-OHIP

• maml

• QUIP

神经网络

• PyTorch

• TensorFlow

  Keras

• JAX

1

pip install --upgrade "jax[cuda12_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html # <= Python3.9`

安装GPU驱动

• 鼠标右键NVIDIA控制面板–帮助–系统信息–驱动程序版本： 217.00 (RTX 3080 Ti)，查看Table 3 CUDA Toolkit and Corresponding Driver Versions，CUDA 11.7 Update 1 >=515.48.07 >=516.31。或使用nvidia-smi查看。

• 从CUDA Toolkit Archive，下载CUDA Toolkit 11.7.1。

• 从cuDNN Archive，下载cuDNN v8.9.7 (2023 年 12 月 5 日), 适用于 CUDA 11.x。

  修改环境变量：右键此电脑–属性–高级系统设置–环境变量–Path–编辑–新建，

1
2
3
4

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.7\bin
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.7\include
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.7\lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.7\libnvvp

• 测试

1
2
3

cd C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.7\extras\demo_suite
deviceQuery.exe
bandwidthTest.exe

查看是否有Result = PASS输出。

GPU监控

• nvidia-smi

1

nvidia-smi

• gpustat

1
2

pip install gpustat
gpustat

安装TensorFlow、PyTorch、Keras

• PyTorch

1

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117

• TensorFlow

1
2
3
4

pip install tensorflow[and-cuda] # Linux
pip install tensorflow # Windows
# or
pip install tensorflow-gpu==2.10.1 # Windows Python <= 3.10, Anaconda3-2023.03-1*

测试TensorFlow、PyTorch、Keras

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

# -*-coding:utf-8 -*-

import os
os.environ['TF_ENABLE_ONEDNN_OPTS'] = '0'
import tensorflow as tf
import torch
import jax
from jax.lib import xla_bridge
import keras
import warnings
warnings.filterwarnings("ignore")
os.chdir(os.path.split(os.path.realpath(__file__))[0])
print('copyright by misaraty (misaraty@163.com)\n' + 'last update: 2024-04-23\n')

print("------------")
# TensorFlow GPU test
print("TensorFlow:")
print("TensorFlow version:", tf.__version__)
print("Num GPUs Available: ", len(tf.config.list_physical_devices('GPU')))
print("------------")

# PyTorch GPU test
print("PyTorch:")
print("PyTorch version:", torch.__version__)
print("Is CUDA available: ", torch.cuda.is_available())
print("Num GPUs Available:", torch.cuda.device_count())
print("GPU Name:", torch.cuda.get_device_name(0) if torch.cuda.is_available() else "No GPU")
print("------------")

# # JAX GPU test
# print("JAX:")
# print("JAX version:", jax.__version__)
# print("JAX backend:", xla_bridge.get_backend().platform)
# print("------------")

# Keras GPU test (uses TensorFlow backend)
print("Keras:")
print("Keras version:", keras.__version__)
print("Keras GPU available:", "Yes" if tf.config.list_physical_devices('GPU') else "No")
print("------------")

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

[root@master ~]# python
Python 3.11.7 (main, Dec 15 2023, 18:12:31) [GCC 11.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  ...
  File "/opt/ohpc/pub/apps/anaconda3/lib/python3.11/site-packages/google/protobuf/descriptor.py", line 47, in <module>
    from google.protobuf.pyext import _message
ImportError: /opt/ohpc/pub/compiler/gcc/8.3.0/lib64/libstdc++.so.6: version `GLIBCXX_3.4.29' not found (required by /opt/ohpc/pub/apps/anaconda3/lib/python3.11/site-packages/google/protobuf/pyext/_message.cpython-311-x86_64-linux-gnu.so)

1
2
3
4
5

cd /opt/ohpc/pub/apps/anaconda3/lib
strings libstdc++.so | grep GLIBCXX
# .bashrc
export LD_LIBRARY_PATH=/opt/ohpc/pub/apps/anaconda3/lib:$LD_LIBRARY_PATH

图神经网络

• MEGNet Chen C, Ye W, Zuo Y, et al. Graph networks as a universal machine learning framework for molecules and crystals[J]. Chemistry of Materials, 2019, 31(9): 3564-3572. 被引843

• CGCNN | Tutorial Park C W, Wolverton C. Developing an improved crystal graph convolutional neural network framework for accelerated materials discovery[J]. Physical Review Materials, 2020, 4(6): 063801. 被引228

  AugLiChem: CGCNN k-fold cross validation

  MatDGL

• ALIGNN Choudhary K, DeCost B. Atomistic line graph neural network for improved materials property predictions[J]. npj Computational Materials, 2021, 7(1): 185. 被引165

• M3GNet Chen C, Ong S P. A universal graph deep learning interatomic potential for the periodic table[J]. Nature Computational Science, 2022, 2(11): 718-728. 被引136

  MatGL

1

pip install matgl --no-deps

• DeeperGATGNN Louis S Y, Zhao Y, Nasiri A, et al. Graph convolutional neural networks with global attention for improved materials property prediction[J]. Physical Chemistry Chemical Physics, 2020, 22(32): 18141-18148. 被引130

• OGCNN Karamad M, Magar R, Shi Y, et al. Orbital graph convolutional neural network for material property prediction[J]. Physical Review Materials, 2020, 4(9): 093801. 被引94

• CHGNet Deng B, Zhong P, Jun K J, et al. CHGNet as a pretrained universal neural network potential for charge-informed atomistic modelling[J]. Nature Machine Intelligence, 2023, 5(9): 1031-1041. 被引29

• KGCNN Reiser P, Eberhard A, Friederich P. Graph neural networks in TensorFlow-Keras with RaggedTensor representation (kgcnn)[J]. Software Impacts, 2021, 9: 100095. 被引14

• SkipAtom Antunes L M, Grau-Crespo R, Butler K T. Distributed representations of atoms and materials for machine learning[J]. npj Computational Materials, 2022, 8(1): 44. 被引11

• CEGANN Banik S, Dhabal D, Chan H, et al. CEGANN: Crystal Edge Graph Attention Neural Network for multiscale classification of materials environment[J]. npj Computational Materials, 2023, 9(1): 23. 被引10

• ChargE3Net Koker T, Quigley K, Taw E, et al. Higher-order equivariant neural networks for charge density prediction in materials[J]. npj Computational Materials, 2024, 10(1): 161. 被引1

• Matformer

1
2
3

git clone https://github.com/YKQ98/Matformer.git
cd Matformer
python setup.py install

• Spektral

• CompStruct

• Graph Data - Keras

• MatBench

• PyG

  DIG

1

pip install torch-geometric

• DGL

1
2
3
4
5

# Linux:
pip install dgl==1.1.2 -f https://data.dgl.ai/wheels/cu117/repo.html

# Windows:
pip install dgl==1.1.2 -f https://data.dgl.ai/wheels/cu117/repo.html

• TensorFlow GNN

• DeepChem

• Jraph

  A library for graph neural networks in jax.

• graph-learn alibaba | graphlearn-for-pytorch

• tf2-gnn Microsoft | tf-gnn-samples

• tf_geometric

  Inspired by rusty1s/pytorch_geometric, we build a GNN library for TensorFlow.

• molgraph

• StellarGraph

  StellarGraph is built on TensorFlow 2 and its Keras high-level API, as well as Pandas and NumPy.

PyG与DGL对比

• PyTorch Geometric vs Deep Graph Library

• Performance Benchmarks

• Characterizing the Efficiency of Graph Neural Network Frameworks with a Magnifying Glass

原子间势

• NequIP Batzner S, Musaelian A, Sun L, et al. E (3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials[J]. Nature communications, 2022, 13(1): 2453. 被引637

• ChemicalMotifIdentifier Sheriff, K., Cao, Y. & Freitas, R. Chemical-motif characterization of short-range order with E(3)-equivariant graph neural networks. npj Comput Mater 10, 215 (2024).

• SchNetPack Schütt K T, Kessel P, Gastegger M, et al. SchNetPack: A deep learning toolbox for atomistic systems[J]. Journal of chemical theory and computation, 2018, 15(1): 448-455. 被引354

• Allegro Musaelian A, Batzner S, Johansson A, et al. Learning local equivariant representations for large-scale atomistic dynamics[J]. Nature Communications, 2023, 14(1): 579. 被引296

• equiformer

   equiformer_v2

• GPUMD

可视化

• Joblib

• shap

• dtreeviz

• Netron

• NN-SVG