diff --git a/README.md b/README.md index 5a5cd380601c82d028c4cc1a3ccd25c63ded1994..4b2cd886801df8eea58a2a4efbe461ddeef0353f 100644 --- a/README.md +++ b/README.md @@ -1,11 +1,28 @@ The logo of Paddle Graph Learning (PGL) +[![PyPi Latest Release](https://img.shields.io/pypi/v/pgl.svg)](https://pypi.org/project/pgl/) +[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](./LICENSE) + [DOC](https://pgl.readthedocs.io/en/latest/) | [Quick Start](https://pgl.readthedocs.io/en/latest/quick_start/instruction.html) | [中文](./README.zh.md) +## Breaking News !! + +PGL v1.1 2020.4.29 + +- You can find **ERNIESage**, a novel model for modeling text and graph structures, and its introduction [here](./examples/erniesage/). + +- PGL for [Open Graph Benchmark](https://github.com/snap-stanford/ogb) examples can be find [here](./ogb_examples/). + +- We add newly graph level operators like **GraphPooling** and [**GraphNormalization**](https://arxiv.org/abs/2003.00982) for graph level predictions. + +- We relase a PGL-KE toolkit [here](./examples/pgl-ke) including classical knowledge graph embedding t algorithms like TransE, TransR, RotatE. + +------ + Paddle Graph Learning (PGL) is an efficient and flexible graph learning framework based on [PaddlePaddle](https://github.com/PaddlePaddle/Paddle). -The Framework of Paddle Graph Learning (PGL) +The Framework of Paddle Graph Learning (PGL) The newly released PGL supports heterogeneous graph learning on both walk based paradigm and message-passing based paradigm by providing MetaPath sampling and Message Passing mechanism on heterogeneous graph. Furthermor, The newly released PGL also support distributed graph storage and some distributed training algorithms, such as distributed deep walk and distributed graphsage. Combined with the PaddlePaddle deep learning framework, we are able to support both graph representation learning models and graph neural networks, and thus our framework has a wide range of graph-based applications. @@ -82,10 +99,11 @@ In most cases of large-scale graph learning, we need distributed graph storage a ## Model Zoo -The following are 13 graph learning models that have been implemented in the framework. See the details [here](https://pgl.readthedocs.io/en/latest/introduction.html#highlight-tons-of-models) +The following graph learning models have been implemented in the framework. You can find more [examples](./examples) and the [details](https://pgl.readthedocs.io/en/latest/introduction.html#highlight-tons-of-models) |Model | feature | |---|---| +| [**ERNIESage**](./examples/erniesage/) | ERNIE SAmple aggreGatE for Text and Graph | | GCN | Graph Convolutional Neural Networks | | GAT | Graph Attention Network | | GraphSage |Large-scale graph convolution network based on neighborhood sampling| diff --git a/README.zh.md b/README.zh.md index e7996398a42a08c1dc1a5022d515ef472eb4382e..8c2d7be4a988a99f8ea34c3ed1d43c8ed3bb321b 100644 --- a/README.zh.md +++ b/README.zh.md @@ -1,7 +1,24 @@ The logo of Paddle Graph Learning (PGL) +[![PyPi Latest Release](https://img.shields.io/pypi/v/pgl.svg)](https://pypi.org/project/pgl/) +[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](./LICENSE) + [文档](https://pgl.readthedocs.io/en/latest/) | [快速开始](https://pgl.readthedocs.io/en/latest/quick_start/instruction.html) | [English](./README.md) +## 最新消息 + +PGL v1.1 2020.4.29 + +- **ERNIESage**是PGL团队最新提出的模型，可以用于建模文本以及图结构信息。你可以在[这里](./examples/erniesage)看到详细的介绍。 + +- PGL现在提供[Open Graph Benchmark](https://github.com/snap-stanford/ogb)的一些例子，你可以在[这里](./ogb_examples)找到。 + +- 新增了图级别的算子包括**GraphPooling**以及[**GraphNormalization**](https://arxiv.org/abs/2003.00982)，这样你就能实现更多复杂的图级别分类模型。 + +- 新增PGL-KE工具包，里面包含许多经典知识图谱图嵌入算法，包括TransE, TransR, RotatE，详情可见[这里](./examples/pgl-ke) + +------ + Paddle Graph Learning (PGL)是一个基于[PaddlePaddle](https://github.com/PaddlePaddle/Paddle)的高效易用的图学习框架 The Framework of Paddle Graph Learning (PGL) @@ -77,10 +94,11 @@ Paddle Graph Learning (PGL)是一个基于[PaddlePaddle](https://github.com/Padd ## 丰富性——覆盖业界大部分图学习网络 -下列是框架中已经自带实现的十三种图网络学习模型。详情请参考[这里](https://pgl.readthedocs.io/en/latest/introduction.html#highlight-tons-of-models) +下列是框架中部分已经实现的图网络模型，更多的模型在[这里](./examples)可以找到。详情请参考[这里](https://pgl.readthedocs.io/en/latest/introduction.html#highlight-tons-of-models) | 模型 | 特点 | |---|---| +| [**ERNIESage**](./examples/erniesage/) | 能同时建模文本以及图结构的ERNIE SAmple aggreGatE | | GCN | 图卷积网络 | | GAT | 基于Attention的图卷积网络 | | GraphSage | 基于邻居采样的大规模图卷积网络 | diff --git a/docs/source/_static/framework_of_pgl.png b/docs/source/_static/framework_of_pgl.png index f8f6293a6821f7a13901f5b83d6d904a4f64ee0d..a054bd3438f60ca5e838d3e80ccdb567fddd23fd 100644 Binary files a/docs/source/_static/framework_of_pgl.png and b/docs/source/_static/framework_of_pgl.png differ diff --git a/docs/source/_static/framework_of_pgl_en.png b/docs/source/_static/framework_of_pgl_en.png new file mode 100644 index 0000000000000000000000000000000000000000..04482775663088455fdd7020d8de5f7db663a177 Binary files /dev/null and b/docs/source/_static/framework_of_pgl_en.png differ diff --git a/docs/source/_static/logo.png b/docs/source/_static/logo.png index fc602ea6c020ee08116c7cd86411834741b5377e..48d5966bfa12d537b92b010a0f9e0c525d3350ae 100644 Binary files a/docs/source/_static/logo.png and b/docs/source/_static/logo.png differ diff --git a/docs/source/quick_start/md/quick_start.md b/docs/source/quick_start/md/quick_start.md index 7df4f48a997e35c001ae462b1f9d3052434eb443..6c4fa5ddac32867a341f3a7d688434b7ca9f1eab 100644 --- a/docs/source/quick_start/md/quick_start.md +++ b/docs/source/quick_start/md/quick_start.md @@ -53,7 +53,6 @@ place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace() # use GraphWrapper as a container for graph data to construct a graph neural network gw = pgl.graph_wrapper.GraphWrapper(name='graph', - place = place, node_feat=g.node_feat_info()) ``` diff --git a/docs/source/quick_start/md/quick_start_for_heterGraph.md b/docs/source/quick_start/md/quick_start_for_heterGraph.md index 1be9cb228ce965cc8f0b6ae65ffc3738e08d0b2a..37ac0de4c362a097518f775da340907408f9a31e 100644 --- a/docs/source/quick_start/md/quick_start_for_heterGraph.md +++ b/docs/source/quick_start/md/quick_start_for_heterGraph.md @@ -77,7 +77,6 @@ place = fluid.CPUPlace() # create a GraphWrapper as a container for graph data gw = heter_graph_wrapper.HeterGraphWrapper(name='heter_graph', - place = place, edge_types = g.edge_types_info(), node_feat=g.node_feat_info(), edge_feat=g.edge_feat_info()) diff --git a/examples/GATNE/model.py b/examples/GATNE/model.py index b193849e14d31eec7822fc65379a74c7cb2be38d..18f83c89a31324256f20ae118372828fe8be955d 100644 --- a/examples/GATNE/model.py +++ b/examples/GATNE/model.py @@ -53,7 +53,6 @@ class GATNE(object): self.gw = heter_graph_wrapper.HeterGraphWrapper( name="heter_graph", - place=place, edge_types=self.graph.edge_types_info(), node_feat=self.graph.node_feat_info(), edge_feat=self.graph.edge_feat_info()) diff --git a/examples/dgi/train.py b/examples/dgi/train.py index 67742093486e9e391cea0d141d504580c2985df0..a23e4e7b8a154f706f9654a6aa69f131af922dde 100644 --- a/examples/dgi/train.py +++ b/examples/dgi/train.py @@ -65,7 +65,6 @@ def main(args): with fluid.program_guard(train_program, startup_program): gw = pgl.graph_wrapper.GraphWrapper( name="graph", - place=place, node_feat=dataset.graph.node_feat_info()) output = pgl.layers.gcn(gw, diff --git a/examples/distribute_graphsage/train.py b/examples/distribute_graphsage/train.py index fa52e3e002b52e14db5ea4e893377476eada41ef..4faafdd504415efd954f09d6d54dc7b38e6287c5 100644 --- a/examples/distribute_graphsage/train.py +++ b/examples/distribute_graphsage/train.py @@ -170,7 +170,7 @@ def main(args): with fluid.program_guard(train_program, startup_program): graph_wrapper = pgl.graph_wrapper.GraphWrapper( - "sub_graph", fluid.CPUPlace(), node_feat=[('feats', [None, 602], np.dtype('float32'))]) + "sub_graph", node_feat=[('feats', [None, 602], np.dtype('float32'))]) model_loss, model_acc = build_graph_model( graph_wrapper, num_class=data["num_class"], diff --git a/examples/gat/train.py b/examples/gat/train.py index 344948803539f260bbe7288a4ee16a423c5c5f8a..438565baad16780ad0c57ae1833db1301160500f 100644 --- a/examples/gat/train.py +++ b/examples/gat/train.py @@ -44,7 +44,6 @@ def main(args): with fluid.program_guard(train_program, startup_program): gw = pgl.graph_wrapper.GraphWrapper( name="graph", - place=place, node_feat=dataset.graph.node_feat_info()) output = pgl.layers.gat(gw, diff --git a/examples/graphsage/train.py b/examples/graphsage/train.py index da20f6e9643b25b800b87d3935feedaf3ec2a62b..463e0b6d4d6457d0307223e21dc79d246a2ef656 100644 --- a/examples/graphsage/train.py +++ b/examples/graphsage/train.py @@ -204,8 +204,8 @@ def main(args): graph_wrapper = pgl.graph_wrapper.GraphWrapper( "sub_graph", - fluid.CPUPlace(), node_feat=data['graph'].node_feat_info()) + model_loss, model_acc = build_graph_model( graph_wrapper, num_class=data["num_class"], diff --git a/examples/graphsage/train_multi.py b/examples/graphsage/train_multi.py index eda3a341c99e4f4b7123dc3eebf0f7f2f79617ad..1f8fe69c6496c2a6f4e933d3ea6aa7da01c50571 100644 --- a/examples/graphsage/train_multi.py +++ b/examples/graphsage/train_multi.py @@ -231,7 +231,6 @@ def main(args): with fluid.program_guard(train_program, startup_program): graph_wrapper = pgl.graph_wrapper.GraphWrapper( "sub_graph", - fluid.CPUPlace(), node_feat=data['graph'].node_feat_info()) model_loss, model_acc = build_graph_model( diff --git a/examples/graphsage/train_scale.py b/examples/graphsage/train_scale.py index f0625d0202b37ca4153b3077451b5032edaf0fbf..c6fce995246e0244badf51e6bc3583d9fc7be9c7 100644 --- a/examples/graphsage/train_scale.py +++ b/examples/graphsage/train_scale.py @@ -227,7 +227,6 @@ def main(args): with fluid.program_guard(train_program, startup_program): graph_wrapper = pgl.graph_wrapper.GraphWrapper( "sub_graph", - fluid.CPUPlace(), node_feat=data['graph'].node_feat_info()) model_loss, model_acc = build_graph_model( diff --git a/examples/stgcn/main.py b/examples/stgcn/main.py index 6be8df9991177daf2ba9fed1f39eebbb8e8ad83f..26adb6a4e6f3c81b4e4e2d35e3f049d0d80f03f5 100644 --- a/examples/stgcn/main.py +++ b/examples/stgcn/main.py @@ -49,7 +49,6 @@ def main(args): with fluid.program_guard(train_program, startup_program): gw = pgl.graph_wrapper.GraphWrapper( "gw", - place, node_feat=[('norm', [None, 1], "float32")], edge_feat=[('weights', [None, 1], "float32")]) diff --git a/examples/unsup_graphsage/train.py b/examples/unsup_graphsage/train.py index a53ffdcb042e74ed262b10a44e0b5c36fad3b1ef..cc7351bb58a05eed05b6f020fc87d41bf1ce6ef9 100644 --- a/examples/unsup_graphsage/train.py +++ b/examples/unsup_graphsage/train.py @@ -88,7 +88,7 @@ def build_graph_model(args): graph_wrappers.append( pgl.graph_wrapper.GraphWrapper( - "layer_0", fluid.CPUPlace(), node_feat=node_feature_info)) + "layer_0", node_feat=node_feature_info)) #edge_feat=[("f", [None, 1], "float32")])) num_embed = args.num_nodes diff --git a/pgl/graph_wrapper.py b/pgl/graph_wrapper.py index ddc19c204a87be0993fa3656b561ac1a6ad5ccc2..91dda8f78796aedd493b37e85a92ad9ecb1c6664 100644 --- a/pgl/graph_wrapper.py +++ b/pgl/graph_wrapper.py @@ -477,9 +477,6 @@ class GraphWrapper(BaseGraphWrapper): Args: name: The graph data prefix - place: fluid.CPUPlace or fluid.CUDAPlace(n) indicating the - device to hold the graph data. - node_feat: A list of tuples that decribe the details of node feature tenosr. Each tuple mush be (name, shape, dtype) and the first dimension of the shape must be set unknown @@ -518,7 +515,6 @@ class GraphWrapper(BaseGraphWrapper): }) graph_wrapper = GraphWrapper(name="graph", - place=place, node_feat=graph.node_feat_info(), edge_feat=graph.edge_feat_info()) @@ -533,12 +529,11 @@ class GraphWrapper(BaseGraphWrapper): ret = exe.run(fetch_list=[...], feed=feed_dict ) """ - def __init__(self, name, place, node_feat=[], edge_feat=[]): + def __init__(self, name, node_feat=[], edge_feat=[], **kwargs): super(GraphWrapper, self).__init__() # collect holders for PyReader self._data_name_prefix = name self._holder_list = [] - self._place = place self.__create_graph_attr_holders() for node_feat_name, node_feat_shape, node_feat_dtype in node_feat: self.__create_graph_node_feat_holders( diff --git a/pgl/heter_graph_wrapper.py b/pgl/heter_graph_wrapper.py index a56bc4e3151b1ab0a136e939e4b736cb8bc32742..bd786c7f459cae6a191a913904202a50e86b3e04 100644 --- a/pgl/heter_graph_wrapper.py +++ b/pgl/heter_graph_wrapper.py @@ -44,9 +44,6 @@ class HeterGraphWrapper(object): Args: name: The heterogeneous graph data prefix - place: fluid.CPUPlace or fluid.CUDAPlace(n) indicating the - device to hold the graph data. - node_feat: A dict of list of tuples that decribe the details of node feature tenosr. Each tuple mush be (name, shape, dtype) and the first dimension of the shape must be set unknown @@ -85,19 +82,15 @@ class HeterGraphWrapper(object): node_feat=node_feat, edge_feat=edges_feat) - place = fluid.CPUPlace() - gw = heter_graph_wrapper.HeterGraphWrapper( name='heter_graph', - place = place, edge_types = g.edge_types_info(), node_feat=g.node_feat_info(), edge_feat=g.edge_feat_info()) """ - def __init__(self, name, place, edge_types, node_feat={}, edge_feat={}): + def __init__(self, name, edge_types, node_feat={}, edge_feat={}, **kwargs): self.__data_name_prefix = name - self._place = place self._edge_types = edge_types self._multi_gw = {} for edge_type in self._edge_types: @@ -114,7 +107,6 @@ class HeterGraphWrapper(object): self._multi_gw[edge_type] = GraphWrapper( name=type_name, - place=self._place, node_feat=n_feat, edge_feat=e_feat)