Used Data Set and Training / Fine Tuning Approach with German Language?
Thank you for providing your model here! Compared to another LLama 2 fine tunned on German, this one is quite better!
I can not find any information regarding the data set you used for fine tuning or any information regarding your training / fine tuning approach with german language? If you could elaborate on that, it would be very interesting for me :-)
Thanks in advance!
Hi, I plan do do a writeup once the model performs better; this was just a first experiment.
Basically I used some German conversation data to finetune this model (which I can't share at this point unfortunately but working on it). I would be interested in your results/feedback for different kind of prompts to improve the model!
Thanks for your reply!
Your work is really interesting. Can you provide some rough information about the training in advance: such as some metrics about your data set (size)? Was the fine tunning on general German conversation data or just specific conversation data? How an example look like? How long your training take and what gpu you used?
That would be really valuable to know!