备注
Go to the end to download the full example code
RGCN-FB15K237-single-gpu || RGCN-FB15K237-single-gpu-wandb || RGCN-FB15K237-single-gpu-hpo
RGCN-FB15K237-single-gpu-wandb¶
备注
created by LuYF-Lemon-love <luyanfeng_nlp@qq.com> on May 7, 2023
备注
updated by LuYF-Lemon-love <luyanfeng_nlp@qq.com> on May 21, 2024
备注
last run by LuYF-Lemon-love <luyanfeng_nlp@qq.com> on May 21, 2024
这一部分介绍如何用一个 GPU 在 FB15K237 知识图谱上训练 R-GCN [SKB+18],使用 wandb 记录实验结果。
导入数据¶
pybind11-OpenKE 有一个工具用于导入数据: pybind11_ke.data.GraphDataLoader。
from pybind11_ke.utils import WandbLogger
from pybind11_ke.data import KGEDataLoader, RGCNSampler, RGCNTestSampler
from pybind11_ke.module.model import RGCN
from pybind11_ke.module.loss import RGCNLoss
from pybind11_ke.module.strategy import RGCNSampling
from pybind11_ke.config import Trainer, Tester
首先初始化 pybind11_ke.utils.WandbLogger 日志记录器,它是对 wandb 初始化操作的一层简单封装。
wandb_logger = WandbLogger(
project="pybind11-ke",
name="rgcn",
config=dict(
in_path = '../../benchmarks/FB15K237/',
batch_size = 60000,
neg_ent = 10,
test = True,
test_batch_size = 100,
num_workers = 16,
dim = 500,
num_layers = 2,
regularization = 1e-5,
use_tqdm = False,
use_gpu = True,
device = 'cuda:0',
epochs = 10000,
lr = 0.0001,
valid_interval = 500,
log_interval = 500,
save_interval = 500,
save_path = '../../checkpoint/rgcn.pth'
)
)
config = wandb_logger.config
pybind11-OpenKE 提供了很多数据集,它们很多都是 KGE 原论文发表时附带的数据集。
pybind11_ke.data.KGEDataLoader 包含 in_path 用于传递数据集目录。
dataloader = KGEDataLoader(
in_path = config.in_path,
batch_size = config.batch_size,
neg_ent = config.neg_ent,
test = config.test,
test_batch_size = config.test_batch_size,
num_workers = config.num_workers,
train_sampler = RGCNSampler,
test_sampler = RGCNTestSampler
)
导入模型¶
pybind11-OpenKE 提供了很多 KGE 模型,它们都是目前最常用的基线模型。我们下面将要导入
pybind11_ke.module.model.RGCN,它提出于 2017 年,是第一个图神经网络模型,
# define the model
rgcn = RGCN(
ent_tol = dataloader.get_ent_tol(),
rel_tol = dataloader.get_rel_tol(),
dim = config.dim,
num_layers = config.num_layers
)
损失函数¶
我们这里使用了 R-GCN [SKB+18] 原论文使用的损失函数:pybind11_ke.module.loss.RGCNLoss,
pybind11_ke.module.strategy.RGCNSampling 对
pybind11_ke.module.loss.RGCNLoss 进行了封装。
# define the loss function
model = RGCNSampling(
model = rgcn,
loss = RGCNLoss(model = rgcn, regularization = config.regularization)
)
训练模型¶
pybind11-OpenKE 将训练循环包装成了 pybind11_ke.config.Trainer,
可以运行它的 pybind11_ke.config.Trainer.run() 函数进行模型学习;
也可以通过传入 pybind11_ke.config.Tester,
使得训练器能够在训练过程中评估模型。
# test the model
tester = Tester(
model = rgcn, data_loader = dataloader, use_tqdm = config.use_tqdm,
use_gpu = config.use_gpu, device = config.device
)
# train the model
trainer = Trainer(model = model, data_loader = dataloader.train_dataloader(),
epochs = config.epochs, lr = config.lr, use_gpu = config.use_gpu, device = config.device,
tester = tester, test = config.test, valid_interval = config.valid_interval,
log_interval = config.log_interval, save_interval = config.save_interval,
save_path = config.save_path, use_wandb = True
)
trainer.run()
备注
上述代码的运行日志可以从 此处 下载。
备注
上述代码的运行报告可以从 此处 下载。
Total running time of the script: ( 0 minutes 0.000 seconds)