Closed xqymain closed 1 year ago
13b is available, I forgot to make it public. 30b is difficult as we are currently unable to train it. I will try to rewrite the training using HF accelerate, hoping this will allows more efficient parallelization but it will take a few days.
Will you make MedAlpaca 13b model publicly available, and the possible 30b?Its performance has me excited.