不支持大知识图谱的训练

[wjy3326]

我目前的知识图谱实体在1500万个左右，在运用transE进行训练时显示oom错误，即使batch size设置为1，也会出现oom错误，请问怎么解决呢？
Traceback (most recent call last):
File "train_transe_WN18_adv_sigmoidloss.py", line 52, in
trainer.run()
File "/ai-images/wjy/event_extraction/recommendation/OpenKE-PyTorch/openke/config/Trainer.py", line 93, in run
loss = self.train_one_step(data)
File "/ai-images/wjy/event_extraction/recommendation/OpenKE-PyTorch/openke/config/Trainer.py", line 52, in train_one_step
loss.backward()
File "/home/user/anaconda3/envs/image_text_match_tf_1.15/lib/python3.6/site-packages/torch/tensor.py", line 245, in backward
torch.autograd.backward(self, gradient, retain_graph, create_graph, inputs=inputs)
File "/home/user/anaconda3/envs/image_text_match_tf_1.15/lib/python3.6/site-packages/torch/autograd/init.py", line 147, in backward
allow_unreachable=True, accumulate_grad=True) # allow_unreachable flag
RuntimeError: CUDA out of memory. Tried to allocate 5.60 GiB (GPU 0; 23.70 GiB total capacity; 16.99 GiB already allocated; 1.43 GiB free; 17.00 GiB reserved in total by PyTorch)

[YBAgg]

你有找到其他支持大知识图谱的开源代码吗

[hopegithub]

cpu怎么样

[YijianLiu]

你有找到其他支持大知识图谱的开源代码吗

我觉得可以试试DGL-KE？不知道你有了解过没

[stupidoge]

你有找到其他支持大知识图谱的开源代码吗

我觉得可以试试DGL-KE？不知道你有了解过没

@YijianLiu DGL-KE感觉调试比这个OpenKE麻烦，而且那个不是源码，直接终端指令一键生成