Quite suprised I did not upload the gpt2. I must have gotten it confused with another one. It was the same size I think as smpanaro's gpt2 and it was like why? If its two different methods of converting? Anyways here is the code repo: https://github.com/antmikinka/StatefulGPT2CoreML
Model tree for anthonymikinka/gpt2_coreml_kv_cache_try1
Base model
openai-community/gpt2