Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
| 72 | |
| 73 | @dataclass |
| 74 | class ModelArguments: |
| 75 | """ |
| 76 | Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch. |
| 77 | """ |
| 78 | |
| 79 | model_name_or_path: Optional[str] = field( |
| 80 | default=None, |
| 81 | metadata={ |
| 82 | "help": "The model checkpoint for weights initialization." |
| 83 | "Don't set if you want to train a model from scratch." |
| 84 | }, |
| 85 | ) |
| 86 | model_type: Optional[str] = field( |
| 87 | default=None, |
| 88 | metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)}, |
| 89 | ) |
| 90 | config_overrides: Optional[str] = field( |
| 91 | default=None, |
| 92 | metadata={ |
| 93 | "help": "Override some existing default config settings when a model is trained from scratch. Example: " |
| 94 | "n_embd=10,resid_pdrop=0.2,scale_attn_weights=false,summary_type=cls_index" |
| 95 | }, |
| 96 | ) |
| 97 | config_name: Optional[str] = field( |
| 98 | default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"} |
| 99 | ) |
| 100 | tokenizer_name: Optional[str] = field( |
| 101 | default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"} |
| 102 | ) |
| 103 | cache_dir: Optional[str] = field( |
| 104 | default=None, |
| 105 | metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co"}, |
| 106 | ) |
| 107 | use_fast_tokenizer: bool = field( |
| 108 | default=True, |
| 109 | metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."}, |
| 110 | ) |
| 111 | model_revision: str = field( |
| 112 | default="main", |
| 113 | metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."}, |
| 114 | ) |
| 115 | use_auth_token: bool = field( |
| 116 | default=False, |
| 117 | metadata={ |
| 118 | "help": "Will use the token generated when running `transformers-cli login` (necessary to use this script " |
| 119 | "with private models)." |
| 120 | }, |
| 121 | ) |
| 122 | from_scratch: bool = field( |
| 123 | default=False, |
| 124 | metadata={ |
| 125 | "help": "Train the model from scratch, starting with a new config, model and tokenizer." |
| 126 | }, |
| 127 | ) |
| 128 | group_name: str = field(default="default") |
| 129 | |
| 130 | def __post_init__(self): |
| 131 | if self.config_overrides is not None and (self.config_name is not None or self.model_name_or_path is not None): |
nothing calls this directly
no outgoing calls
no test coverage detected