About Pre-trained Checkpoints for ReaRev + LMSR

[mufeili]

Thanks for the great work!

When I tried to train a GNN retriever myself following the instructions here, I encountered the issue in the title. Could you release the checkpoints required for training the GNN retrievers on WebQSP and CWQ? Thanks.

[cmavro]

Thanks for the interest in our work!

You can use ReaRev_CWQ.ckpt for --load_experiment, which is provided here, or you can train from scratch, omitting --load_experiment argument.

[mufeili]

Thank you for the kind response!

python main.py ReaRev --entity_dim 50 --num_epoch 200 --batch_size 8 --eval_every 2 --data_folder data/webqsp/ --lm sbert --num_iter 3 --num_ins 2 --num_gnn 3 --relation_word_emb True --load_experiment ReaRev_webqsp.ckpt --is_eval --name webqsp

It seems that the released checkpoints are only for ReaRev + SBERT, which I did manage to load and evaluate. Could you also share the checkpoints for ReaRev + LMSR and the instructions to load and evaluate them?

[MINJIK01]

Hello, I also have the same question with mufeili. I think in your code, it only contains BERT and LSTM. Could I get the code for LM_SR as well? If I missed something then please let me know.

Thank you for your great work!

[hauntl]

@mufeili @MINJIK01

Hello, I also have the same question, could I ask if you solve this problem and how to solve that.

[cmavro]

Hello all,

Thanks for the interest and apologies for the late reply. LM-qr model and GNN are uploaded: here (acknowledgement: https://github.com/RUCKBReasoning/SubgraphRetrievalKBQA).

Let me know if you can use it or I can upload the GNN model again.