All modules are sourced from cog/ except for the chatbot functionality which is defined inside on_message event in bot.py
NOTE: with the dataset, the size of dependencies is somewhere north of 2GB so if you are low on bandwidth I suggest you comment following 3 lines in bot.py (to skip downloading the datasets). Chat functionality won't function without those, but you will save money.
tokenizer = AutoTokenizer.from_pretrained("microsoft/DialoGPT-medium")
model = AutoModelForCausalLM.from_pretrained("microsoft/DialoGPT-medium")
ranker = AutoModelForSequenceClassification.from_pretrained('microsoft/DialogRPT-human-vs-machine')
Alernatively, you can forego pretrained models and feed her your own, just be warned that it's very resource intensive.
Tested on python version 3.9
$ pip install -r requirements.txt
$ python bot.py
Also, there is possibility that i forgot to add some dependencies in requirements.txt, but python's debugger should warn if that's the case
All modules are sourced from cog/ except for the chatbot functionality which is defined inside on_message event in bot.py
NOTE: with the dataset, the size of dependencies is somewhere north of 2GB so if you are low on bandwidth I suggest you comment following 3 lines in bot.py (to skip downloading the datasets). Chat functionality won't function without those, but you will save money.
tokenizer = AutoTokenizer.from_pretrained("microsoft/DialoGPT-medium")
model = AutoModelForCausalLM.from_pretrained("microsoft/DialoGPT-medium")
ranker = AutoModelForSequenceClassification.from_pretrained('microsoft/DialogRPT-human-vs-machine')
Alernatively, you can forego pretrained models and feed her your own, just be warned that it's very resource intensive.
Tested on python version 3.9
$ pip install -r requirements.txt
$ python bot.py
Also, there is possibility that i forgot to add some dependencies in requirements.txt, but python's debugger should warn if that's the case