search

INK-USC / RE-Net

Recurrent Event Network: Autoregressive Structure Inference over Temporal Knowledge Graphs (EMNLP 2020)
http://inklab.usc.edu/renet/
436 stars 95 forks source link

KeyError: 'id' is reported when I run the pretrain.py #55

Open GrayChan04 opened 2 years ago

GrayChan04 commented 2 years ago

excuse me，does anyone know how to deal? batched_graph.ndata['h'] = ent_embeds[batched_graph.ndata['id']].view(-1, ent_embeds.shape[1]) File "/home/xxxx/anaconda3/envs/renet/lib/python3.6/site-packages/dgl/view.py", line 60, in __getitem__ return self._graph.get_n_repr(self._nodes)[key] KeyError: 'id' it seems like something wrong with code batched_graph.ndata['h'] = ent_embeds[batched_graph.ndata['id']].view(-1, ent_embeds.shape[1]) but I don‘t know how to deal with the problem

GrayChan04 commented 2 years ago

I change GDELT to ICEWS14, however still the same problem... according to readme.md, I configuration the environment, but i still CANNOT run the code. do I miss something?

$ python3 pretrain.py -d ICEWS14 --gpu 0 --dropout 0.5 --n-hidden 200 --lr 1e-3 --max-epochs 20 --batch-size 1024 Using backend: pytorch Namespace(batch_size=1024, dataset='ICEWS14', dropout=0.5, gpu=0, grad_norm=1.0, lr=0.001, max_epochs=20, maxpool=1, model=3, n_hidden=200, num_k=10, rnn_layers=1, seq_len=10) start training... /home/xxxx/RE-Net-cp/Aggregator.py:32: UserWarning: This overload of nonzero is deprecated: nonzero(Tensor input, , Tensor out) Consider using one of the following signatures instead: nonzero(Tensor input, , bool as_tuple) (Triggered internally at /pytorch/torch/csrc/utils/python_arg_parser.cpp:766.) num_non_zero = len(torch.nonzero(t_list)) /home/xxxx/RE-Net-cp/utils.py:290: UserWarning: Implicit dimension choice for log_softmax has been deprecated. Change the call to include dim=X as an argument. return torch.mean(torch.sum(- soft_targets * logsoftmax(pred), 1)) Traceback (most recent call last): File "pretrain.py", line 139, in train(args) File "pretrain.py", line 92, in train model.global_emb = model.get_global_emb(train_times_origin, graph_dict) File "/home/xxxx/RE-Net-cp/global_model.py", line 67, in get_globalemb emb, , _ = self.predict(t, graph_dict) File "/home/xxxx/RE-Net-cp/global_model.py", line 88, in predict rnn_inp = self.aggregator.predict(t, self.ent_embeds, graph_dict, reverse=reverse) File "/home/xxxx/RE-Net-cp/Aggregator.py", line 97, in predict batched_graph.ndata['h'] = ent_embeds[batched_graph.ndata['id']].view(-1, ent_embeds.shape[1]) File "/home/xxxx/anaconda3/envs/renet/lib/python3.6/site-packages/dgl/view.py", line 60, in getitem return self._graph.get_n_repr(self._nodes)[key] KeyError: 'id'

Zhai-Huichen commented 2 years ago

batched_graph.ndata['h'] = ent_embeds[batched_graph.ndata['id']].view(-1, ent_embeds.shape[1]) 应该是batched_graph在cpu上，用gpu上的东西给它赋值出错了，在出错那一行之前把batched_graph转到gpu里

GrayChan04 commented 2 years ago

已经转到gpu了 但是还是报同样的错

Zhai-Huichen commented 2 years ago

你的get_history_graph.py里，方法get_data_witht有改过吗？ x = data[np.where(data[3] == tim)].copy()改成x = data[np.where(data[**:,_** 3] == tim)].copy() 我用的ICEWS18跑的，14好像有问题

已经转到gpu了 但是还是报同样的错

Zhai-Huichen commented 2 years ago

你的get_history_graph.py里，方法get_data_witht有改过吗？ x = data[np.where(data[3] == tim)].copy()改成x = data[np.where(data[**:,_** 3] == tim)].copy() 我用的ICEWS18跑的，14好像有问题

GrayChan04 commented 2 years ago

你的get_history_graph.py里，方法get_data_witht有改过吗？ x = data[np.where(data[3] == tim)].copy()改成x = data[np.where(data[**:,_** 3] == tim)].copy() 我用的ICEWS18跑的，14好像有问题

我没有改 这里的代码我觉得没有问题

WWWindrunner commented 2 years ago

你的get_history_graph.py里，方法get_data_witht有改过吗？ x = data[np.where(data[3] == tim)].copy()改成x = data[np.where(data[**:,_** 3] == tim)].copy() 我用的ICEWS18跑的，14好像有问题

我没有改 这里的代码我觉得没有问题

这个方法你打印过中间结果吗，我之前遇到['id']报错，就是因为这里执行不对，导致batched_graph为空。 我改为了 triples = [[quad[0], quad[1], quad[2]] for quad in data if quad[3] == tim] 就ok了

MrLiuCC commented 2 years ago

你的get_history_graph.py里，方法get_data_witht有改过吗？ x = data[np.where(data[3] == tim)].copy()改成x = data[np.where(data[**:,_** 3] == tim)].copy() 我用的ICEWS18跑的，14好像有问题

ICEWS14里面没有valid.txt

MrLiuCC commented 2 years ago

图片 请问这部分是这样子改的吗？

WWWindrunner commented 2 years ago

嗯

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2022年3月31日(星期四) 下午4:28 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [INK-USC/RE-Net] KeyError: 'id' is reported when I run the pretrain.py (Issue #55)

请问这部分是这样子改的吗？

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

s821220 commented 2 years ago

请问最后解决了吗？？解决的方案是什么

Young0222 commented 2 years ago

我有解决这个问题，请参考： batched_graph.ndata['h'] = ent_embeds[batched_graph.ndata['id']].view(-1, ent_embeds.shape[1]).to('cpu') batched_graph = batched_graph.to('cuda:0') 另外，删除move_dgl_to_cuda(batched_graph)这一行