Good exploration and experiment, i hope you continue with it.
#8
by
AnA202
- opened
There is a lot of problem that the model fails in math but in my opiniont the way it work and trying to solve it already looking good and with that i hope in future with maybe 7-12b parameters tune it could solving it as your tune looks good but need more improvment with more data and more training time.
Awesome model! I hope to see 0.5b and 1.5b SmallThinker models.