Instruction Tuning Availability?
#4
by
ArihantSheth
- opened
Hi, I wanted to know if this model is instruction tuned and capable of following multi-line instructions? Or is it tuned on the base (non instruction-tuned) version of Mistral 7B?
I want to use the mathematical logic to solve mathematical problems using their language descriptions. So I wanted to know how well is this model instruction-tuned? I tried running inference but haven't been able to get the best responses so far.
We use the MetaMath (https://github.com/meta-math/MetaMath) to do the instruction tuning on this base model, the ckpt is here (https://huggingface.co/TencentARC/MetaMath-Mistral-Pro).