备注
Go to the end to download the full example code
TransH-FB15K237-single-gpu || TransH-FB15K237-single-gpu-wandb || TransH-FB15K237-single-gpu-hpo || TransH-FB15K237-accelerate || TransH-FB15K237-accelerate-wandb
TransH-FB15K237-single-gpu¶
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 11, 2024
备注
last run by LuYF-Lemon-love <luyanfeng_nlp@qq.com> on May 11, 2024
这一部分介绍如何用一个 GPU 在 FB15K237 知识图谱上训练 TransH [WZFC14]。
导入数据¶
pybind11-OpenKE 有 1 个工具用于导入数据: pybind11_ke.data.KGEDataLoader。
from pybind11_ke.data import KGEDataLoader, BernSampler, TradTestSampler
from pybind11_ke.module.model import TransH
from pybind11_ke.module.loss import MarginLoss
from pybind11_ke.module.strategy import NegativeSampling
from pybind11_ke.config import Trainer, Tester
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 for training
dataloader = KGEDataLoader(
in_path = "../../benchmarks/FB15K237/",
batch_size = 4096,
neg_ent = 25,
test = True,
test_batch_size = 30,
num_workers = 16,
train_sampler = BernSampler,
test_sampler = TradTestSampler
)
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.TransH,它提出于 2014 年,是第二个平移模型,
将关系建模为超平面上的平移操作。
# define the model
transh = TransH(
ent_tol = dataloader.get_ent_tol(),
rel_tol = dataloader.get_rel_tol(),
dim = 200,
p_norm = 1,
norm_flag = True)
损失函数¶
我们这里使用了 TransE [BUGD+13] 原论文使用的损失函数:pybind11_ke.module.loss.MarginLoss,
pybind11_ke.module.strategy.NegativeSampling 对
pybind11_ke.module.loss.MarginLoss 进行了封装,加入权重衰减等额外项。
# define the loss function
model = NegativeSampling(
model = transh,
loss = MarginLoss(margin = 4.0),
# regul_rate = 0.01
)
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 = transh, data_loader = dataloader, use_tqdm = False, use_gpu = True, device = 'cuda:0')
tester.set_hits([1, 3, 10, 30, 100, 200])
# train the model
trainer = Trainer(model = model, data_loader = dataloader.train_dataloader(),
epochs = 1000, lr = 0.5, use_gpu = True, device = 'cuda:0',
tester = tester, test = True, valid_interval = 100,
log_interval = 100, save_interval = 100, save_path = '../../checkpoint/transh.pth',
delta = 0.01)
trainer.run()
备注
上述代码的运行日志可以从 此处 下载。
Total running time of the script: ( 0 minutes 0.000 seconds)