zlacker

[parent] [thread] 3 comments
1. coder5+(OP)[view] [source] 2023-12-21 00:40:06
The article shows (fine tuned) Mistral 7B outperforming GPT-4, never mind GPT-3.5.
replies(1): >>m3kw9+A6
2. m3kw9+A6[view] [source] 2023-12-21 01:43:12
>>coder5+(OP)
This model is not close to even 3.5 from when I used it. It first of all does not follow instructions properly and it just runs on and on
replies(1): >>coder5+m8
◧◩
3. coder5+m8[view] [source] [discussion] 2023-12-21 02:05:25
>>m3kw9+A6
What you're describing is the behavior you get from any base model that has not been instruction-tuned. The article is clear that this model is not for "direct use". It needs tuning for a specific application.
replies(1): >>m3kw9+sf
◧◩◪
4. m3kw9+sf[view] [source] [discussion] 2023-12-21 03:24:46
>>coder5+m8
how does one fine tune it to follow instructions? I would have thought they have open source training set for these instruction-follow fine tunes?
[go to top]