Training parameters
#5
by add567743322 - opened
Could you document the training parameters used? I'd like to experiment with cleaning up the datasets used and adding some of my own.
of course. at the moment this model will be remade in v2. still trying new parameters out using the MoE. v2 of both this model and the MoE variant will be out by the end of this week with major improvements.
Not only were these models trained on some bad data (this was our mistake, effected datasets have been remade and reuploaded) but they didn't respond the way i'd hoped. Once the models are out we will attach the training pipeline in the model card