…d local hugging face model and finetune loaded model with hugging face dataset

Added features to download models from hugging face model hub/load local hugging face model and finetune loaded model with hugging face dataset. Model loading and fine-tuning can happen both at the initialization stage and after the agent has been initialized (see README in `agentscope/examples/load_finetune_huggingface_model` for details). Major changes to the repo include creating the example script `load_finetune_huggingface_model`, adding a new model wrapper `HuggingFaceWrapper`, and creating a new agent type Finetune_DialogAgent. All changes are done in a new example directory `agentscope/examples/load_finetune_huggingface_model`.

made customized hyperparameters specification available from `model_configs` for fine-tuning at initialization, or through `fine_tune_config` in `Finetune_DialogAgent`'s `fine_tune` method after initialization

… from local.

fixed issue related to `format` method

… branch for it)

updated the dependencies needed

Updated the way to read token from .env file, so that it can work in any example directory.

…tuned_model

optimized the behavior of `device_map` when loading a huggingface model. Now if `device` is not given by the user, `device_map="auto"` by default; otherwise  `device_map` is set to the user-specified `device`.

…o disable ).

…o disable ).

now the user can choose to do full-parameter finetuning by not passing `lora_config`

now the user can choose to do full-parameter finetuning by not passing `lora_config` and `bnb_config`

…g and

…etuned_model' into conversation_with_agent_with_finetuned_model

updated according to the latest comments

…ing to comments

…tuned_model

…ug for continual finetuning.

… `PeftModel` before convert it to `PeftModel`

… parameters

…tuned_model

…Agent