A general-purpose continuing pre-training based on LLAMA3-8B.
Note that this version is only allowed to be loaded with =bfloat16=.
You can use =AutoClass= of huggingface to load this model.
