We implement a bigram character-level language model that can generate names based on a given txt file. This model predicts the next letter based on the previous 2 letters.
This model is trained on the names of the 1000 most common baby names for boys and girls taken from ssa.gov.
The code is inspired by the open source code from Andrej Karpathy's makemore. This code includes a detailed explanation of the model and the training process in leyman terms.