RuntimeError: Need to set model argument to use create_agent

[vinnik-dmitry07]

RuntimeError: Need to set `model` argument to use create_agent.
Also I have KeyError: 'skip_generation' running DisplayModel.main.

You can reproduce it via the ParlAI Tutorial notebook.

[vinnik-dmitry07]

It seems to be my bad.

[bashiryounis]

hello
I have a model trained I want to convert to ParlAI! how can do it!

[klshuster]

where did you train it?

[bashiryounis]

pytorch using BertForSequenceClassification and dataset is Csv

[bashiryounis]

Enter text (q to quit): how are you
--- Text: how are you
--- Prediction: self.depression | Confidence: 0.4847826361656189
--- Confidence for each class
class confidence
0 self.Anxiety 0.179089
1 self.bipolar 0.158747
2 self.depression 0.484783
3 self.SuicideWatch 0.114597
4 self.offmychest
there is the output from my model I want to chat using parlai

[klshuster]

you'd need to update the state dict to match one of our existing models; most likely, you're looking for our transformer/classifier agent; specifically, modifying the weights to be a linear wrapper around a memnet model

[bashiryounis]

is there a tutorial in docs about that! I am stuck at this point I want clear detail!

[klshuster]

No, however, please see this script for an example of converting a fairseq checkpoint into a ParlAI model