THU-KEG / MetaKGR

Source codes and datasets for EMNLP 2019 paper "Adapting Meta Knowledge Graph Information for Multi-Hop Reasoning over Few-Shot Relations"
115 stars 21 forks source link

when i run "./experiment-rs.sh configs/fb15k-237-rs.sh --train 0 --few_shot" #6

Open wj1108114106 opened 4 years ago

wj1108114106 commented 4 years ago

I change the model to point.rs.conve. In fact, there doesn't exist error, but a userwanrning like follows: **"‘UserWarning: RNN module weights are not part of single contiguous chunk of memory. This means they need to be compacted at every call, possibly reately increasing memory usage. To compact weights again call flattenparameters()"*_ I found the code: self.path = [self.path_encoder(init_action_embedding, (init_h, init_c))[1]] self.path_encoder = nn.LSTM() found the class LSTM, and added the flatten_parameters, it doesn''t still work. and is it related to the version of anaconda(v_4.8.3) could you give some suggestions?

summerone123 commented 2 years ago

I have the same question,Have you solved it?or could you give some suggestions?

I change the model to point.rs.conve. In fact, there doesn't exist error, but a userwanrning like follows: **"‘UserWarning: RNN module weights are not part of single contiguous chunk of memory. This means they need to be compacted at every call, possibly reately increasing memory usage. To compact weights again call flattenparameters()"*_ I found the code: self.path = [self.path_encoder(init_action_embedding, (init_h, init_c))[1]] self.path_encoder = nn.LSTM() found the class LSTM, and added the flatten_parameters, it doesn''t still work. and is it related to the version of anaconda(v_4.8.3) could you give some suggestions?