r""" Instantiate a pretrained LTP model from a pre-trained model configuration from huggingface-hub. The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated). To train the model, you should first set
(
cls,
pretrained_model_name_or_path: str,
force_download: bool = False,
resume_download: bool = False,
proxies: Dict = None,
use_auth_token: Optional[str] = None,
cache_dir: Optional[str] = None,
local_files_only: bool = False,
**model_kwargs,
)
| 28 | |
| 29 | @classmethod |
| 30 | def from_pretrained( |
| 31 | cls, |
| 32 | pretrained_model_name_or_path: str, |
| 33 | force_download: bool = False, |
| 34 | resume_download: bool = False, |
| 35 | proxies: Dict = None, |
| 36 | use_auth_token: Optional[str] = None, |
| 37 | cache_dir: Optional[str] = None, |
| 38 | local_files_only: bool = False, |
| 39 | **model_kwargs, |
| 40 | ): |
| 41 | r""" |
| 42 | Instantiate a pretrained LTP model from a pre-trained model |
| 43 | configuration from huggingface-hub. The model is set in |
| 44 | evaluation mode by default using `model.eval()` (Dropout modules |
| 45 | are deactivated). To train the model, you should first set it |
| 46 | back in training mode with `model.train()`. |
| 47 | |
| 48 | Parameters: |
| 49 | pretrained_model_name_or_path (`str` or `os.PathLike`): |
| 50 | Can be either: |
| 51 | - A string, the `model id` of a pretrained model |
| 52 | hosted inside a model repo on huggingface.co. |
| 53 | Valid model ids are [`LTP/tiny`, `LTP/small`, |
| 54 | `LTP/base`, `LTP/base1`, `LTP/base1`, `LTP/legacy` |
| 55 | ], the legacy model only support cws, pos and ner, |
| 56 | but more fast. |
| 57 | - You can add `revision` by appending `@` at the end |
| 58 | of model_id simply like this: |
| 59 | `dbmdz/bert-base-german-cased@main` Revision is |
| 60 | the specific model version to use. It can be a |
| 61 | branch name, a tag name, or a commit id, since we |
| 62 | use a git-based system for storing models and |
| 63 | other artifacts on huggingface.co, so `revision` |
| 64 | can be any identifier allowed by git. |
| 65 | - A path to a `directory` containing model weights |
| 66 | saved using |
| 67 | [`~transformers.PreTrainedModel.save_pretrained`], |
| 68 | e.g., `./my_model_directory/`. |
| 69 | - `None` if you are both providing the configuration |
| 70 | and state dictionary (resp. with keyword arguments |
| 71 | `config` and `state_dict`). |
| 72 | force_download (`bool`, *optional*, defaults to `False`): |
| 73 | Whether to force the (re-)download of the model weights |
| 74 | and configuration files, overriding the cached versions |
| 75 | if they exist. |
| 76 | resume_download (`bool`, *optional*, defaults to `False`): |
| 77 | Whether to delete incompletely received files. Will |
| 78 | attempt to resume the download if such a file exists. |
| 79 | proxies (`Dict[str, str]`, *optional*): |
| 80 | A dictionary of proxy servers to use by protocol or |
| 81 | endpoint, e.g., `{'http': 'foo.bar:3128', |
| 82 | 'http://hostname': 'foo.bar:4012'}`. The proxies are |
| 83 | used on each request. |
| 84 | use_auth_token (`str` or `bool`, *optional*): |
| 85 | The token to use as HTTP bearer authorization for remote |
| 86 | files. If `True`, will use the token generated when |
| 87 | running `transformers-cli login` (stored in |