# Training details
Dataset : llama tokenizer 4000
distribution : 85% 15%
only for attributes 
number of epochs : 3
Progressive : no