torchdrug / Git / Diff of /README.md

Models:

Amanda-D/

torchdrug

Downloads: 1

Diff of /README.md [36b44b] .. [3eb847]

Switch to unified view










[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1Tbnr1Fog_YjkqU1MOhcVLuxqZ4DC-c8-#forceEdit=true&sandboxMode=true)
[![Contributions](https://img.shields.io/badge/contributions-welcome-blue)](https://github.com/DeepGraphLearning/torchdrug/blob/master/CONTRIBUTING.md)
[![License Apache-2.0](https://img.shields.io/github/license/DeepGraphLearning/torchdrug?color=blue)](https://github.com/DeepGraphLearning/torchdrug/blob/master/LICENSE)
[![PyPI downloads](https://static.pepy.tech/personalized-badge/torchdrug?period=total&units=international_system&left_color=grey&right_color=blue&left_text=downloads)](https://pypi.org/project/torchdrug/)
[![TorchDrug Twitter](https://img.shields.io/twitter/url?label=TorchDrug&style=social&url=https%3A%2F%2Ftwitter.com%2FDrugTorch)](https://twitter.com/DrugTorch)

[Docs] | [Tutorials] | [Benchmarks] | [Papers Implemented]

[Docs]: https://deepgraphlearning.github.io/torchdrug-site/docs
[Tutorials]: https://deepgraphlearning.github.io/torchdrug-site/docs/tutorials
[Benchmarks]: https://deepgraphlearning.github.io/torchdrug-site/docs/benchmark
[Papers Implemented]: https://deepgraphlearning.github.io/torchdrug-site/docs/paper

TorchDrug is a [PyTorch]-based machine learning toolbox designed for several purposes.

- Easy implementation of graph operations in a PyTorchic style with GPU support
- Being friendly to practitioners with minimal knowledge about drug discovery
- Rapid prototyping of machine learning research

[PyTorch]: https://pytorch.org/

Installation
------------

TorchDrug can be installed on either Linux, Windows or macOS. It is compatible with
3.7 <= Python <= 3.10 and PyTorch >= 1.8.0.

### From Conda ###

```bash
conda install torchdrug -c milagraph -c conda-forge -c pytorch -c pyg
```

### From Pip ###

```bash
pip install torch==1.9.0
pip install torch-scatter torch-cluster -f https://pytorch-geometric.com/whl/torch-1.9.0+cu102.html
pip install torchdrug
```

To install `torch-scatter` for other PyTorch or CUDA versions, please see the
instructions in https://github.com/rusty1s/pytorch_scatter

### From Source ###

```bash
git clone https://github.com/DeepGraphLearning/torchdrug
cd torchdrug
pip install -r requirements.txt
python setup.py install
```

### Windows (PowerShell) ###

We need to first install the build tools for Visual Studio. We then install the
following modules in PowerShell.

```powershell
Install-Module Pscx -AllowClobber
Install-Module VSSetup
```

Initialize Visual Studio in PowerShell with the following commands. We may setup
this for all PowerShell sessions by writing it to the PowerShell profile. Change
the library path according to your own case.

```powershell
Import-VisualStudioVars -Architecture x64
$env:LIB += ";C:\Program Files\Python37\libs"
```

### Apple Silicon (M1/M2 Chips) ###

We need PyTorch >= 1.13 to run TorchDrug on Apple silicon. For `torch-scatter` and
`torch-cluster`, they can be compiled from their sources. Note TorchDrug doesn't
support `mps` devices.

```bash
pip install torch==1.13.0
pip install git+https://github.com/rusty1s/pytorch_scatter.git
pip install git+https://github.com/rusty1s/pytorch_cluster.git
pip install torchdrug
```

Quick Start
-----------

TorchDrug is designed for humans and focused on graph structured data.
It enables easy implementation of graph operations in machine learning models.
All the operations in TorchDrug are backed by [PyTorch] framework, and support GPU
acceleration and auto differentiation.

```python
from torchdrug import data

edge_list = [[0, 1], [1, 2], [2, 3], [3, 4], [4, 5], [5, 0]]
graph = data.Graph(edge_list, num_node=6)
graph = graph.cuda()
# the subgraph induced by nodes 2, 3 & 4
subgraph = graph.subgraph([2, 3, 4])
```

Molecules are also supported in TorchDrug. You can get the desired molecule
properties without any domain knowledge.

```python
mol = data.Molecule.from_smiles("CCOC(=O)N", atom_feature="default", bond_feature="default")
print(mol.node_feature)
print(mol.atom_type)
print(mol.to_scaffold())
```

You may also register custom node, edge or graph attributes. They will be
automatically processed during indexing operations.

```python
with mol.edge():
    mol.is_CC_bond = (mol.edge_list[:, :2] == td.CARBON).all(dim=-1)
sub_mol = mol.subgraph(mol.atom_type != td.NITROGEN)
print(sub_mol.is_CC_bond)
```

TorchDrug provides a wide range of common datasets and building blocks for drug
discovery. With minimal code, you can apply standard models to solve your own
problem.

```python
import torch
from torchdrug import datasets

dataset = datasets.Tox21()
dataset[0].visualize()
lengths = [int(0.8 * len(dataset)), int(0.1 * len(dataset))]
lengths += [len(dataset) - sum(lengths)]
train_set, valid_set, test_set = torch.utils.data.random_split(dataset, lengths)
```

```python
from torchdrug import models, tasks

model = models.GIN(dataset.node_feature_dim, hidden_dims=[256, 256, 256, 256])
task = tasks.PropertyPrediction(model, task=dataset.tasks)
```

Training and inference are accelerated by multiple CPUs or GPUs.
This can be seamlessly switched in TorchDrug by just a line of code.
```python
from torchdrug import core

# Single CPU / Multiple CPUs / Distributed CPUs
solver = core.Engine(task, train_set, valid_set, test_set, optimizer)
# Single GPU
solver = core.Engine(task, train_set, valid_set, test_set, optimizer, gpus=[0])
# Multiple GPUs
solver = core.Engine(task, train_set, valid_set, test_set, optimizer, gpus=[0, 1, 2, 3])
# Distributed GPUs
solver = core.Engine(task, train_set, valid_set, test_set, optimizer, gpus=[0, 1, 2, 3, 0, 1, 2, 3])
```

Experiments can be easily tracked and managed through [Weights & Biases platform].
```python
solver = core.Engine(task, train_set, valid_set, test_set, optimizer, logger="wandb")
```

[Weights & Biases platform]: https://wandb.ai/

Contributing
------------

Everyone is welcome to contribute to the development of TorchDrug.
Please refer to [contributing guidelines](CONTRIBUTING.md) for more details.

License
-------

TorchDrug is released under [Apache-2.0 License](LICENSE).

	a/README.md		b/README.md
1	[![TorchDrug](asset/torchdrug_logo_full.svg)](https://torchdrug.ai/)	1
2	<h1 align="center">
3	with
4	<a href="https://torchprotein.ai/">
5	<img src="asset/torchprotein_logo_tight.svg" alt="TorchProtein" style="height:26px" />
6	</a>
7	</h1>
8
9	[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1Tbnr1Fog_YjkqU1MOhcVLuxqZ4DC-c8-#forceEdit=true&sandboxMode=true)	2	[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1Tbnr1Fog_YjkqU1MOhcVLuxqZ4DC-c8-#forceEdit=true&sandboxMode=true)
10	[![Contributions](https://img.shields.io/badge/contributions-welcome-blue)](https://github.com/DeepGraphLearning/torchdrug/blob/master/CONTRIBUTING.md)	3	[![Contributions](https://img.shields.io/badge/contributions-welcome-blue)](https://github.com/DeepGraphLearning/torchdrug/blob/master/CONTRIBUTING.md)
11	[![License Apache-2.0](https://img.shields.io/github/license/DeepGraphLearning/torchdrug?color=blue)](https://github.com/DeepGraphLearning/torchdrug/blob/master/LICENSE)	4	[![License Apache-2.0](https://img.shields.io/github/license/DeepGraphLearning/torchdrug?color=blue)](https://github.com/DeepGraphLearning/torchdrug/blob/master/LICENSE)
12	[![PyPI downloads](https://static.pepy.tech/personalized-badge/torchdrug?period=total&units=international_system&left_color=grey&right_color=blue&left_text=downloads)](https://pypi.org/project/torchdrug/)	5	[![PyPI downloads](https://static.pepy.tech/personalized-badge/torchdrug?period=total&units=international_system&left_color=grey&right_color=blue&left_text=downloads)](https://pypi.org/project/torchdrug/)
13	[![TorchDrug Twitter](https://img.shields.io/twitter/url?label=TorchDrug&style=social&url=https%3A%2F%2Ftwitter.com%2FDrugTorch)](https://twitter.com/DrugTorch)	6	[![TorchDrug Twitter](https://img.shields.io/twitter/url?label=TorchDrug&style=social&url=https%3A%2F%2Ftwitter.com%2FDrugTorch)](https://twitter.com/DrugTorch)
14		7
15	[Docs] \| [Tutorials] \| [Benchmarks] \| [Papers Implemented]	8	[Docs] \| [Tutorials] \| [Benchmarks] \| [Papers Implemented]
16		9
17	[Docs]: https://deepgraphlearning.github.io/torchdrug-site/docs	10	[Docs]: https://deepgraphlearning.github.io/torchdrug-site/docs
18	[Tutorials]: https://deepgraphlearning.github.io/torchdrug-site/docs/tutorials	11	[Tutorials]: https://deepgraphlearning.github.io/torchdrug-site/docs/tutorials
19	[Benchmarks]: https://deepgraphlearning.github.io/torchdrug-site/docs/benchmark	12	[Benchmarks]: https://deepgraphlearning.github.io/torchdrug-site/docs/benchmark
20	[Papers Implemented]: https://deepgraphlearning.github.io/torchdrug-site/docs/paper	13	[Papers Implemented]: https://deepgraphlearning.github.io/torchdrug-site/docs/paper
21		14
22	TorchDrug is a [PyTorch]-based machine learning toolbox designed for several purposes.	15	TorchDrug is a [PyTorch]-based machine learning toolbox designed for several purposes.
23		16
24	- Easy implementation of graph operations in a PyTorchic style with GPU support	17	- Easy implementation of graph operations in a PyTorchic style with GPU support
25	- Being friendly to practitioners with minimal knowledge about drug discovery	18	- Being friendly to practitioners with minimal knowledge about drug discovery
26	- Rapid prototyping of machine learning research	19	- Rapid prototyping of machine learning research
27		20
28	[PyTorch]: https://pytorch.org/	21	[PyTorch]: https://pytorch.org/
29		22
30	Installation	23	Installation
31	------------	24	------------
32		25
33	TorchDrug can be installed on either Linux, Windows or macOS. It is compatible with	26	TorchDrug can be installed on either Linux, Windows or macOS. It is compatible with
34	3.7 <= Python <= 3.10 and PyTorch >= 1.8.0.	27	3.7 <= Python <= 3.10 and PyTorch >= 1.8.0.
35		28
36	### From Conda ###	29	### From Conda ###
37		30
38	```bash	31	```bash
39	conda install torchdrug -c milagraph -c conda-forge -c pytorch -c pyg	32	conda install torchdrug -c milagraph -c conda-forge -c pytorch -c pyg
40	```	33	```
41		34
42	### From Pip ###	35	### From Pip ###
43		36
44	```bash	37	```bash
45	pip install torch==1.9.0	38	pip install torch==1.9.0
46	pip install torch-scatter torch-cluster -f https://pytorch-geometric.com/whl/torch-1.9.0+cu102.html	39	pip install torch-scatter torch-cluster -f https://pytorch-geometric.com/whl/torch-1.9.0+cu102.html
47	pip install torchdrug	40	pip install torchdrug
48	```	41	```
49		42
50	To install `torch-scatter` for other PyTorch or CUDA versions, please see the	43	To install `torch-scatter` for other PyTorch or CUDA versions, please see the
51	instructions in https://github.com/rusty1s/pytorch_scatter	44	instructions in https://github.com/rusty1s/pytorch_scatter
52		45
53	### From Source ###	46	### From Source ###
54		47
55	```bash	48	```bash
56	git clone https://github.com/DeepGraphLearning/torchdrug	49	git clone https://github.com/DeepGraphLearning/torchdrug
57	cd torchdrug	50	cd torchdrug
58	pip install -r requirements.txt	51	pip install -r requirements.txt
59	python setup.py install	52	python setup.py install
60	```	53	```
61		54
62	### Windows (PowerShell) ###	55	### Windows (PowerShell) ###
63		56
64	We need to first install the build tools for Visual Studio. We then install the	57	We need to first install the build tools for Visual Studio. We then install the
65	following modules in PowerShell.	58	following modules in PowerShell.
66		59
67	```powershell	60	```powershell
68	Install-Module Pscx -AllowClobber	61	Install-Module Pscx -AllowClobber
69	Install-Module VSSetup	62	Install-Module VSSetup
70	```	63	```
71		64
72	Initialize Visual Studio in PowerShell with the following commands. We may setup	65	Initialize Visual Studio in PowerShell with the following commands. We may setup
73	this for all PowerShell sessions by writing it to the PowerShell profile. Change	66	this for all PowerShell sessions by writing it to the PowerShell profile. Change
74	the library path according to your own case.	67	the library path according to your own case.
75		68
76	```powershell	69	```powershell
77	Import-VisualStudioVars -Architecture x64	70	Import-VisualStudioVars -Architecture x64
78	$env:LIB += ";C:\Program Files\Python37\libs"	71	$env:LIB += ";C:\Program Files\Python37\libs"
79	```	72	```
80		73
81	### Apple Silicon (M1/M2 Chips) ###	74	### Apple Silicon (M1/M2 Chips) ###
82		75
83	We need PyTorch >= 1.13 to run TorchDrug on Apple silicon. For `torch-scatter` and	76	We need PyTorch >= 1.13 to run TorchDrug on Apple silicon. For `torch-scatter` and
84	`torch-cluster`, they can be compiled from their sources. Note TorchDrug doesn't	77	`torch-cluster`, they can be compiled from their sources. Note TorchDrug doesn't
85	support `mps` devices.	78	support `mps` devices.
86		79
87	```bash	80	```bash
88	pip install torch==1.13.0	81	pip install torch==1.13.0
89	pip install git+https://github.com/rusty1s/pytorch_scatter.git	82	pip install git+https://github.com/rusty1s/pytorch_scatter.git
90	pip install git+https://github.com/rusty1s/pytorch_cluster.git	83	pip install git+https://github.com/rusty1s/pytorch_cluster.git
91	pip install torchdrug	84	pip install torchdrug
92	```	85	```
93		86
94	Quick Start	87	Quick Start
95	-----------	88	-----------
96		89
97	TorchDrug is designed for humans and focused on graph structured data.	90	TorchDrug is designed for humans and focused on graph structured data.
98	It enables easy implementation of graph operations in machine learning models.	91	It enables easy implementation of graph operations in machine learning models.
99	All the operations in TorchDrug are backed by [PyTorch] framework, and support GPU	92	All the operations in TorchDrug are backed by [PyTorch] framework, and support GPU
100	acceleration and auto differentiation.	93	acceleration and auto differentiation.
101		94
102	```python	95	```python
103	from torchdrug import data	96	from torchdrug import data
104		97
105	edge_list = [[0, 1], [1, 2], [2, 3], [3, 4], [4, 5], [5, 0]]	98	edge_list = [[0, 1], [1, 2], [2, 3], [3, 4], [4, 5], [5, 0]]
106	graph = data.Graph(edge_list, num_node=6)	99	graph = data.Graph(edge_list, num_node=6)
107	graph = graph.cuda()	100	graph = graph.cuda()
108	# the subgraph induced by nodes 2, 3 & 4	101	# the subgraph induced by nodes 2, 3 & 4
109	subgraph = graph.subgraph([2, 3, 4])	102	subgraph = graph.subgraph([2, 3, 4])
110	```	103	```
111		104
112	Molecules are also supported in TorchDrug. You can get the desired molecule	105	Molecules are also supported in TorchDrug. You can get the desired molecule
113	properties without any domain knowledge.	106	properties without any domain knowledge.
114		107
115	```python	108	```python
116	mol = data.Molecule.from_smiles("CCOC(=O)N", atom_feature="default", bond_feature="default")	109	mol = data.Molecule.from_smiles("CCOC(=O)N", atom_feature="default", bond_feature="default")
117	print(mol.node_feature)	110	print(mol.node_feature)
118	print(mol.atom_type)	111	print(mol.atom_type)
119	print(mol.to_scaffold())	112	print(mol.to_scaffold())
120	```	113	```
121		114
122	You may also register custom node, edge or graph attributes. They will be	115	You may also register custom node, edge or graph attributes. They will be
123	automatically processed during indexing operations.	116	automatically processed during indexing operations.
124		117
125	```python	118	```python
126	with mol.edge():	119	with mol.edge():
127	mol.is_CC_bond = (mol.edge_list[:, :2] == td.CARBON).all(dim=-1)	120	mol.is_CC_bond = (mol.edge_list[:, :2] == td.CARBON).all(dim=-1)
128	sub_mol = mol.subgraph(mol.atom_type != td.NITROGEN)	121	sub_mol = mol.subgraph(mol.atom_type != td.NITROGEN)
129	print(sub_mol.is_CC_bond)	122	print(sub_mol.is_CC_bond)
130	```	123	```
131		124
132	TorchDrug provides a wide range of common datasets and building blocks for drug	125	TorchDrug provides a wide range of common datasets and building blocks for drug
133	discovery. With minimal code, you can apply standard models to solve your own	126	discovery. With minimal code, you can apply standard models to solve your own
134	problem.	127	problem.
135		128
136	```python	129	```python
137	import torch	130	import torch
138	from torchdrug import datasets	131	from torchdrug import datasets
139		132
140	dataset = datasets.Tox21()	133	dataset = datasets.Tox21()
141	dataset[0].visualize()	134	dataset[0].visualize()
142	lengths = [int(0.8 * len(dataset)), int(0.1 * len(dataset))]	135	lengths = [int(0.8 * len(dataset)), int(0.1 * len(dataset))]
143	lengths += [len(dataset) - sum(lengths)]	136	lengths += [len(dataset) - sum(lengths)]
144	train_set, valid_set, test_set = torch.utils.data.random_split(dataset, lengths)	137	train_set, valid_set, test_set = torch.utils.data.random_split(dataset, lengths)
145	```	138	```
146		139
147	```python	140	```python
148	from torchdrug import models, tasks	141	from torchdrug import models, tasks
149		142
150	model = models.GIN(dataset.node_feature_dim, hidden_dims=[256, 256, 256, 256])	143	model = models.GIN(dataset.node_feature_dim, hidden_dims=[256, 256, 256, 256])
151	task = tasks.PropertyPrediction(model, task=dataset.tasks)	144	task = tasks.PropertyPrediction(model, task=dataset.tasks)
152	```	145	```
153		146
154	Training and inference are accelerated by multiple CPUs or GPUs.	147	Training and inference are accelerated by multiple CPUs or GPUs.
155	This can be seamlessly switched in TorchDrug by just a line of code.	148	This can be seamlessly switched in TorchDrug by just a line of code.
156	```python	149	```python
157	from torchdrug import core	150	from torchdrug import core
158		151
159	# Single CPU / Multiple CPUs / Distributed CPUs	152	# Single CPU / Multiple CPUs / Distributed CPUs
160	solver = core.Engine(task, train_set, valid_set, test_set, optimizer)	153	solver = core.Engine(task, train_set, valid_set, test_set, optimizer)
161	# Single GPU	154	# Single GPU
162	solver = core.Engine(task, train_set, valid_set, test_set, optimizer, gpus=[0])	155	solver = core.Engine(task, train_set, valid_set, test_set, optimizer, gpus=[0])
163	# Multiple GPUs	156	# Multiple GPUs
164	solver = core.Engine(task, train_set, valid_set, test_set, optimizer, gpus=[0, 1, 2, 3])	157	solver = core.Engine(task, train_set, valid_set, test_set, optimizer, gpus=[0, 1, 2, 3])
165	# Distributed GPUs	158	# Distributed GPUs
166	solver = core.Engine(task, train_set, valid_set, test_set, optimizer, gpus=[0, 1, 2, 3, 0, 1, 2, 3])	159	solver = core.Engine(task, train_set, valid_set, test_set, optimizer, gpus=[0, 1, 2, 3, 0, 1, 2, 3])
167	```	160	```
168		161
169	Experiments can be easily tracked and managed through [Weights & Biases platform].	162	Experiments can be easily tracked and managed through [Weights & Biases platform].
170	```python	163	```python
171	solver = core.Engine(task, train_set, valid_set, test_set, optimizer, logger="wandb")	164	solver = core.Engine(task, train_set, valid_set, test_set, optimizer, logger="wandb")
172	```	165	```
173		166
174	[Weights & Biases platform]: https://wandb.ai/	167	[Weights & Biases platform]: https://wandb.ai/
175		168
176	Contributing	169	Contributing
177	------------	170	------------
178		171
179	Everyone is welcome to contribute to the development of TorchDrug.	172	Everyone is welcome to contribute to the development of TorchDrug.
180	Please refer to [contributing guidelines](CONTRIBUTING.md) for more details.	173	Please refer to [contributing guidelines](CONTRIBUTING.md) for more details.
181		174
182	License	175	License
183	-------	176	-------
184		177
185	TorchDrug is released under [Apache-2.0 License](LICENSE).	178	TorchDrug is released under [Apache-2.0 License](LICENSE).