This is a Llama 2 architecture model series trained on the TinyStories dataset, intended for use in the llama2.c project by Andrej Karpathy.
Trained on a single v100 32GB GPU for 3 epochs, we achieve an inference speed of ~72 tokens/sec on the same.
Achieved tok/s: 161.819538 on 12th Gen Intel(R) Core(TM) i9-12900HK
Learn more on how to run inference in pure C using llama2.c