(
self,
pretrained: Optional[Union[str, transformers.PreTrainedModel]] = "gpt2",
max_length: Optional[int] = None,
batch_size: Optional[Union[int, str]] = 1,
trust_remote_code: Optional[bool] = False,
**kwargs,
)
| 40 | class EvalHarnessAdaptor(HFLM): |
| 41 | |
| 42 | def __init__( |
| 43 | self, |
| 44 | pretrained: Optional[Union[str, transformers.PreTrainedModel]] = "gpt2", |
| 45 | max_length: Optional[int] = None, |
| 46 | batch_size: Optional[Union[int, str]] = 1, |
| 47 | trust_remote_code: Optional[bool] = False, |
| 48 | **kwargs, |
| 49 | ) -> None: |
| 50 | self.args = get_args() |
| 51 | build_tokenizer(self.args) |
| 52 | self.tokenizer = get_tokenizer() |
| 53 | self.is_main = torch.distributed.get_rank() == 0 |
| 54 | self.adaptive_seq_len = self.args.adaptive_seq_len |
| 55 | self.model_provider = kwargs['model_provider'] |
| 56 | |
| 57 | super().__init__(pretrained=pretrained, |
| 58 | batch_size=batch_size, |
| 59 | trust_remote_code=trust_remote_code, |
| 60 | max_length=max_length, |
| 61 | tokenizer=self.tokenizer) |
| 62 | |
| 63 | def _create_model( |
| 64 | self, |
nothing calls this directly
no test coverage detected