added tokenizer and tokenizer config from MNTP as well

Added config from MNTP in attempt to try and do further PEFT finetuning on SimCSE with new dataset (DK>Scandi)

Re-uploaded SimCSE from step1000 initial peft

Rm 3000 step SimCSE, worse performance than 1000 steps

Uploaded 3000 step model from long_train branch

Deleted old safetensor, to replae w. 3000 steps

Del. only needed on mntp

Added config w. link to LLM2Vec

restructured files

Copy of checkpoint model files from simcse final checkpoint

initial commit