train.py

# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import tqdm
import pgl
import paddle
import paddle.nn as nn
from pgl.utils.logger import log
import numpy as np
import time
import argparse
from paddle.optimizer import Adam


class GCN(nn.Layer):
    """Implement of GCN
    """

    def __init__(self,
                 input_size,
                 num_class,
                 num_layers=1,
                 hidden_size=64,
                 dropout=0.5):
        super(GCN, self).__init__()
        self.num_class = num_class
        self.num_layers = num_layers
        self.hidden_size = hidden_size
        self.dropout = dropout
        self.gcns = nn.LayerList()
        for i in range(self.num_layers):
            if i == 0:
                self.gcns.append(
                    pgl.nn.GCNConv(
                        input_size,
                        self.hidden_size,
                        activation="relu",
                        norm=True))
            else:
                self.gcns.append(
                    pgl.nn.GCNConv(
                        self.hidden_size,
                        self.hidden_size,
                        activation="relu",
                        norm=True))
            self.gcns.append(nn.Dropout(self.dropout))
        self.gcns.append(pgl.nn.GCNConv(self.hidden_size, self.num_class))

    def forward(self, graph, feature):
        for m in self.gcns:
            if isinstance(m, nn.Dropout):
                feature = m(feature)
            else:
                feature = m(graph, feature)
        return feature


def normalize(feat):
    return feat / np.maximum(np.sum(feat, -1, keepdims=True), 1)


def load(name, normalized_feature=True):
    if name == 'cora':
        dataset = pgl.dataset.CoraDataset()
    elif name == "pubmed":
        dataset = pgl.dataset.CitationDataset("pubmed", symmetry_edges=True)
    elif name == "citeseer":
        dataset = pgl.dataset.CitationDataset("citeseer", symmetry_edges=True)
    else:
        raise ValueError(name + " dataset doesn't exists")

    indegree = dataset.graph.indegree()
    dataset.graph.node_feat["words"] = normalize(dataset.graph.node_feat[
        "words"])

    dataset.graph.tensor()
    train_index = dataset.train_index
    dataset.train_label = paddle.to_tensor(
        np.expand_dims(dataset.y[train_index], -1))
    dataset.train_index = paddle.to_tensor(np.expand_dims(train_index, -1))

    val_index = dataset.val_index
    dataset.val_label = paddle.to_tensor(
        np.expand_dims(dataset.y[val_index], -1))
    dataset.val_index = paddle.to_tensor(np.expand_dims(val_index, -1))

    test_index = dataset.test_index
    dataset.test_label = paddle.to_tensor(
        np.expand_dims(dataset.y[test_index], -1))
    dataset.test_index = paddle.to_tensor(np.expand_dims(test_index, -1))

    return dataset


def train(node_index, node_label, gnn_model, graph, criterion, optim):
    gnn_model.train()
    pred = gnn_model(graph, graph.node_feat["words"])
    pred = paddle.gather(pred, node_index)
    loss = criterion(pred, node_label)
    loss.backward()
    acc = paddle.metric.accuracy(input=pred, label=node_label, k=1)
    optim.step()
    optim.clear_grad()
    return loss, acc


@paddle.no_grad()
def eval(node_index, node_label, gnn_model, graph, criterion):
    gnn_model.eval()
    pred = gnn_model(graph, graph.node_feat["words"])
    pred = paddle.gather(pred, node_index)
    loss = criterion(pred, node_label)
    acc = paddle.metric.accuracy(input=pred, label=node_label, k=1)
    return loss, acc


def set_seed(seed):
    paddle.seed(seed)
    np.random.seed(seed)


def main(args):
    dataset = load(args.dataset, args.feature_pre_normalize)

    graph = dataset.graph
    train_index = dataset.train_index
    train_label = dataset.train_label

    val_index = dataset.val_index
    val_label = dataset.val_label

    test_index = dataset.test_index
    test_label = dataset.test_label
    criterion = paddle.nn.loss.CrossEntropyLoss()

    dur = []

    best_test = []

    for run in range(args.runs):
        cal_val_acc = []
        cal_test_acc = []
        cal_val_loss = []
        cal_test_loss = []
        gnn_model = GCN(input_size=graph.node_feat["words"].shape[1],
                        num_class=dataset.num_classes,
                        num_layers=1,
                        dropout=0.5,
                        hidden_size=16)

        optim = Adam(
            learning_rate=0.01,
            parameters=gnn_model.parameters(),
            weight_decay=0.0005)

        for epoch in tqdm.tqdm(range(200)):
            if epoch >= 3:
                start = time.time()
            train_loss, train_acc = train(train_index, train_label, gnn_model,
                                          graph, criterion, optim)
            if epoch >= 3:
                end = time.time()
                dur.append(end - start)
            val_loss, val_acc = eval(val_index, val_label, gnn_model, graph,
                                     criterion)
            cal_val_acc.append(val_acc.numpy())
            cal_val_loss.append(val_loss.numpy())

            test_loss, test_acc = eval(test_index, test_label, gnn_model,
                                       graph, criterion)
            cal_test_acc.append(test_acc.numpy())
            cal_test_loss.append(test_loss.numpy())

        log.info("Runs %s: Model: GCN Best Test Accuracy: %f" %
                 (run, cal_test_acc[np.argmin(cal_val_loss)]))

        best_test.append(cal_test_acc[np.argmin(cal_val_loss)])

    log.info("Average Speed %s sec/ epoch" % (np.mean(dur)))
    log.info("Dataset: %s Best Test Accuracy: %f ( stddev: %f )" %
             (args.dataset, np.mean(best_test), np.std(best_test)))


if __name__ == '__main__':
    parser = argparse.ArgumentParser(
        description='Benchmarking Citation Network')
    parser.add_argument(
        "--dataset", type=str, default="cora", help="dataset (cora, pubmed)")
    parser.add_argument("--epoch", type=int, default=200, help="Epoch")
    parser.add_argument("--runs", type=int, default=10, help="runs")
    parser.add_argument(
        "--feature_pre_normalize",
        type=bool,
        default=True,
        help="pre_normalize feature")
    args = parser.parse_args()
    log.info(args)
    main(args)