prakharguptaz / Instructdial

Code for the paper Code for the paper InstructDial: Improving Zero and Few-shot Generalization in Dialogue through Instruction Tuning
Apache License 2.0
96 stars 13 forks source link

Task mixture for the hosted models #13

Closed cuthalionn closed 1 year ago

cuthalionn commented 1 year ago

Hi, thanks for sharing and maintaining this great code base!

I want to replicate the training of your main model hosted in huggingface but I am unsure which task mixture you used for the training. Could you share the list of tasks for both train and test?

Thanks!

prakharguptaz commented 1 year ago

Thanks for your interest. The config files for models we perform our main experiments on (Table 1 in the paper) are- train test

The models released on huggingface are trained on all possible tasks and there are no test tasks per say. The config file is alltasks-config