Dependencies

1. pytorch>=1.12
2. allennlp==2.10
3. complexPyTorch=0.4

Some notes about allennlp

1. --dry-run # load dataset but do not train the model
2. -f # force training, this command will override the save path
3. -s # save path
4. --include-package $path # personal work path including model, classifier and so on
5. config/xxx.jsonnet # config file with jsonnet format

Our train command

CQLSTM

allennlp train config/CQLSTM.jsonnet --include-package work -s ./result/cr_CQLSTM -f

Dataset

Please modify the variable "task_name" in model.jsonnet to change different datasets.

Other command

# delete training models
rm ./result/*/*_state_*.th -f