All pipeline is llama.cpp?
#3
by albex123 - opened
Do you have .safetensors for this model?
And why 3B score is better than 7B
We spent less resources/time training the 7B version than the 3B version. We decided to focus in models up to 4b in size, so our 7B model is not completelly trained with our methodology. In the near future a more capable 4b version will be delivered. We do not plan to disclose .safetensors for the time being untill our research isnt finished.