HuggingFace模型存储位置

Hugging Face models, also known as transformers or pre-trained models, are primarily stored and accessed through their Model Hub website (<https://huggingface.co/models>) or the associated Python library called `transformers`. These models are hosted on Amazon Web Services (AWS) S3 storage by default. Users can download them for local use or load them directly into their applications using the `from_pretrained` function provided by the `transformers` library. You can access a model's specific location with the format `huggingface.co/<username>/<model_name>`, where `<username>` is the owner of the model and `<model_name>` is the name given to it during training.

To interact with these models in your code, you would typically do something like this:

from transformers import AutoModel, AutoTokenizer

model_name = "bert-base-chinese"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModel.from_pretrained(model_name)