你好，我想请教下代码的一些细节问题

[songzeceng]

1、max-r-length和max-c-length表示什么意思呢

2、是不是train_data.jsonl语料中的一些电影没有在id2entity.pkl收录呢，我打断点调试到dataset.py文件里detect_movice()函数中，发现entity = self.id2entity[int(movie)]有时返回为空...

[songzeceng]

而且调试到utils.py中_create_embeddings()函数时，发现也没有word2vec_redial.npy文件..

[songzeceng]

另外我在代码里找不到公式3、4、9中提到的互信息最大化损失对应的位置..

[tanyue2019]

你好，我想麻烦问一下您，您是用作者要求的pytorch==1.3.0,torch_geometric==1.3.2进行训练的吗？现在这两个版本已经没有了，我不太知道如何能够配好正确的环境。期待您的回复，谢谢！

[songzeceng]

你好，我想麻烦问一下您，您是用作者要求的pytorch==1.3.0,torch_geometric==1.3.2进行训练的吗？现在这两个版本已经没有了，我不太知道如何能够配好正确的环境。期待您的回复，谢谢！

我的环境是1.6.1的torch-geometric和1.6.0+cu101的pytorch

[tanyue2019]

哦哦，那您能成功运行是吗？

[songzeceng]

哦哦，那您能成功运行是吗？

可以训练并计算损失函数，还没整完：
3%|█▎ | 50/1616 [00:31<16:13, 1.61it/s]info db loss is 8.705469
6%|██▍ | 100/1616 [01:02<15:41, 1.61it/s]info db loss is 1.966950
9%|███▋ | 150/1616 [01:34<15:11, 1.61it/s]info db loss is 1.705429
12%|████▉ | 200/1616 [02:05<14:40, 1.61it/s]info db loss is 1.738045
15%|██████▏ | 250/1616 [02:36<14:08, 1.61it/s]info db loss is 1.594990
19%|███████▍ | 300/1616 [03:07<13:37, 1.61it/s]info db loss is 1.417708
22%|████████▋ | 350/1616 [03:38<13:06, 1.61it/s]info db loss is 1.397896
25%|█████████▉ | 400/1616 [04:09<12:35, 1.61it/s]info db loss is 1.399305
28%|███████████▏ | 450/1616 [04:40<12:04, 1.61it/s]info db loss is 1.504028
31%|████████████▍ | 500/1616 [05:11<11:33, 1.61it/s]info db loss is 1.447741
34%|█████████████▌ | 550/1616 [05:42<11:02, 1.61it/s]info db loss is 1.490247

[tanyue2019]

好的，太感谢啦！

[songzeceng]

好的，太感谢啦！

不客气

[tanyue2019]

您好，您是已经训练了多久还没整完？我看您九月份就开始运行代码了

[songzeceng]

您好，您是已经训练了多久还没整完？我看您九月份就开始运行代码了

估计得一下午。我是从九月份阅读论文复现实验的，但是没有仔细研究，因为十月份我换了方向了

[tanyue2019]

好我刚刚关注了您的博客 我是河北人 哈哈离得很近呢！谢谢

[songzeceng]

好我刚刚关注了您的博客 我是河北人 哈哈离得很近呢！谢谢

没事儿

[tanyue2019]

{'recall@1': 0.03322259136212625, 'recall@10': 0.20066445182724252, 'recall@50': 0.412624584717608, 'loss': tensor(6.9143, device='cuda:0'), 'gate': 0.0, 'count': 1.0, 'gate_count': 0.0}
Traceback (most recent call last):
File "run.py", line 687, in
loop.train()
File "run.py", line 217, in train
self.model.save_model()
File "/home/ubuntu/KGSF-master/KGSF-master/model.py", line 540, in save_model
torch.save(self.state_dict(), 'saved_model/net_parameter1.pkl')
File "/home/ubuntu/anaconda3/envs/tan18/lib/python3.8/site-packages/torch/serialization.py", line 369, in save
with _open_file_like(f, 'wb') as opened_file:
File "/home/ubuntu/anaconda3/envs/tan18/lib/python3.8/site-packages/torch/serialization.py", line 230, in _open_file_like
return _open_file(name_or_buffer, mode)
File "/home/ubuntu/anaconda3/envs/tan18/lib/python3.8/site-packages/torch/serialization.py", line 211, in init
super(_open_file, self).init(open(name, mode))
FileNotFoundError: [Errno 2] No such file or directory: 'saved_model/net_parameter1.pkl'

你好~请问您第一阶段训练完后是这种结果吗？有没有出现这个找不到文件的错误呢

[songzeceng]

{'recall@1': 0.03322259136212625, 'recall@10': 0.20066445182724252, 'recall@50': 0.412624584717608, 'loss': tensor(6.9143, device='cuda:0'), 'gate': 0.0, 'count': 1.0, 'gate_count': 0.0}
Traceback (most recent call last):
File "run.py", line 687, in
loop.train()
File "run.py", line 217, in train
self.model.save_model()
File "/home/ubuntu/KGSF-master/KGSF-master/model.py", line 540, in save_model
torch.save(self.state_dict(), 'saved_model/net_parameter1.pkl')
File "/home/ubuntu/anaconda3/envs/tan18/lib/python3.8/site-packages/torch/serialization.py", line 369, in save
with _open_file_like(f, 'wb') as opened_file:
File "/home/ubuntu/anaconda3/envs/tan18/lib/python3.8/site-packages/torch/serialization.py", line 230, in _open_file_like
return _open_file(name_or_buffer, mode)
File "/home/ubuntu/anaconda3/envs/tan18/lib/python3.8/site-packages/torch/serialization.py", line 211, in init
super(_open_file, self).init(open(name, mode))
FileNotFoundError: [Errno 2] No such file or directory: 'saved_model/net_parameter1.pkl'

你好~请问您第一阶段训练完后是这种结果吗？有没有出现这个找不到文件的错误呢
会报这种错误，所以我把is_finetune参数默认设为false(run.py第70行左右)，具体原因你可以查一下saved_model/net_parameter1.pkl在哪儿出现，这个文件被加载时的调用栈是如何的

[tanyue2019]

train.add_argument("-is_finetune","--is_finetune",type=bool,default=False)

谢谢！您改完之后就好了吗？可是我看源码里本来就是false？