GPT-NeoX-20B was trained with fp16, thus it is recommended to initialize the model as follows: | |
python | |
model = GPTNeoXForCausalLM.from_pretrained("EleutherAI/gpt-neox-20b").half().cuda() | |
GPT-NeoX-20B also has a different tokenizer from the one used in GPT-J-6B and GPT-Neo. |