bug #2
Description
作者你好，非常感谢你的工作。
在复现代码时遇到了一个bug，您知道这是什么原因吗？
(pytorch) zx@pc-PowerEdge-T640:~/codes/P-INT$ python train.py --dataset "NELL-One" --few 5 --max_batches 200000
------HYPERPARAMETERS-------
dataset: NELL-One
embed_dim: 100
hidden_dim: 100
mode: test
few: 5
batch_size: 32
random_embed: False
lr: 0.0005
margin: 64.0
max_batches: 200000
dropout: 0.2
log_every: 100
test: False
grad_clip: 5.0
weight_decay: 0.0
embed_model: TransE
prefix: intial
seed: 88
epoch: 3
hop: 3
max_neighbor: 200
eval_every: 100
neighbor_limit: 300
topk: 100
kernel_num: 21
save_path: models/NELL-One5intial
2022-08-24 18:47:30 INFO: - START TRAINING...
5/200000Traceback (most recent call last):
File "/data/zx/codes/P-INT/train.py", line 48, in
main(args)
File "/data/zx/codes/P-INT/train.py", line 15, in main
trainer.train() # train
File "/data/zx/codes/P-INT/trainer.py", line 274, in train
loss = self.T_GRUA(support_pair, support_rel, support_path, query_head, query_tail, one_tomany_train, self.cos_rel_all, True, candidates_id)
File "/data/zx/anaconda3/envs/pytorch/lib/python3.9/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/data/zx/codes/P-INT/net.py", line 153, in forward
tree_q, tree_all_emb, parent_index, parent_node, aim_rel_all = self.T_GRUQ(support_tree_emb, support_rel, query_head, cos_rel_all, t_h, True)
File "/data/zx/anaconda3/envs/pytorch/lib/python3.9/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/data/zx/codes/P-INT/net.py", line 380, in forward
m = torch.distributions.multinomial.Multinomial(total_count=self.topk, logits=score_max)
File "/data/zx/anaconda3/envs/pytorch/lib/python3.9/site-packages/torch/distributions/multinomial.py", line 64, in init
super(Multinomial, self).init(batch_shape, event_shape, validate_args=validate_args)
File "/data/zx/anaconda3/envs/pytorch/lib/python3.9/site-packages/torch/distributions/distribution.py", line 53, in init
raise ValueError("The parameter {} has invalid values".format(param))
ValueError: The parameter probs has invalid values