search

MeetKai / functionary

Chat language model that can use tools and interpret the results
MIT License
1.37k stars 107 forks source link

Decision process for fn calls #245

Closed xdevfaheem closed 1 month ago

xdevfaheem commented 1 month ago

I'm very curious about, how functionary model knows when to call a fn or when to chat normally. Before i came to know about functionary, I thought in order do so for my PA agent, i has to incorporate a classifier for every forward pass throught the module.

Is there a aligning process or it's basically thought to do through SFT or how? It would be very helpful if you could share some info about this @khai-meetkai, TQSM.

jeffrey-fong commented 1 month ago

Typically, we fine-tune the models using SFT on our own dataset. Here are some details about training, although we haven't update it for the newer base models like Llama3/3.1 yet.

We do not have plans to open-source our dataset and it's creation process currently. Thank you for your understanding.

xdevfaheem commented 1 month ago

So it's basically trained, when or when not to and how to respond. Great!