Which file sould be saved when finetuning Bert a classifcation task?

🤗Transformers
1

Hello , I finetune Bert on a classification task and I want to know which files in the checkpoint of the finetune model should be keep and which I can delete. I obtain the following file ::

                                                                      config.json
-rw-r----- 1 umg16uw kpf 584K Dec 14 11:37 merges.txt
-rw-r----- 1 umg16uw kpf 2.1G Dec 14 11:37 optimizer.pt
-rw-r----- 1 umg16uw kpf 1.1G Dec 14 11:37 pytorch_model.bin
-rw-r----- 1 umg16uw kpf  18K Dec 14 11:37 rng_state.pth
-rw-r----- 1 umg16uw kpf  623 Dec 14 11:37 scheduler.pt
-rw-r----- 1 umg16uw kpf  305 Dec 14 11:37 special_tokens_map.json
-rw-r----- 1 umg16uw kpf  587 Dec 14 11:37 tokenizer_config.json
-rw-r----- 1 umg16uw kpf  14K Dec 14 11:37 trainer_state.json
-rw-r----- 1 umg16uw kpf 2.7K Dec 14 11:37 training_args.bin
-rw-r----- 1 umg16uw kpf 1.3M Dec 14 11:37 vocab.json

Are they all necessary when reloading the model for inference or testing on new non annotated data ? Because the repertory is quite heavy and I cannot download it on my computer ?