search

hengyicai / Adaptive_Multi-curricula_Learning_for_Dialog

The codebase for "Learning from Easy to Complex: Adaptive Multi-curricula Learning for Neural Dialogue Generation" (Cai et al., AAAI 2020)
MIT License
19 stars 3 forks source link

several files of building up data are missing #3

Open frankdarkluo opened 2 years ago

frankdarkluo commented 2 years ago

When I run the command bash projects/adaptive_learning/shell/build_data.sh, there is an error as follows:

python: can't open file 'projects/adaptive_learning/scripts/build_single_turn_fb_dialog.py': [Errno 2] No such file or directory

There is no 'build_single_turn_fb_dialog.py' in the corresponding directory and I cannot make the data.

In addition, 'format_fb_dialog.py' and 'correct_train_data.py' are missing

I intend to use other datasets and generate files to fit into the model.

I appreciate it if the author can provide the scripts, or is there an alternative script to build up the data?

hengyicai commented 2 years ago

Sorry for the inconvenience. These simple scripts are used for double-checking the dataset is in the right format, and are placed into the projects/adaptive_learning/scripts/ now.

frankdarkluo commented 2 years ago

Sorry for the inconvenience. These simple scripts are used for double-checking the dataset is in the right format, and are placed into the projects/adaptive_learning/scripts/ now.

Thank you for your reply and your updated scripts, they are very useful.

In addition, however, in the 'build_data.sh', I also find that there is no 'model.opt' provided, though you list them from line 7 to 35 in the shell file 'build_data.sh', could you please also share those files, or there may be some other paths to the 'model.opt'? thank you.

hengyicai commented 2 years ago

When using ParlAI to train a specific model, e.g., transformer, ParlAI saves the model parameters to the given checkpoint dir: model' for saving the parameters andmodel.opt' for saving the model configuration, as well as the vocabulary file. These ``model.opt'' files can be simply obtained by training specific models using ParlAI. Hope it helps.