in grok/data.py [0:0]
def encode(self, obj: Union[str, List]) -> Tensor:
"""
Convert a string of text into a rank-1 tensor of token ids
or convert a list of strings of text into a rank-2 tensor of token ids
:param obj: the string or list of strings to convert
:returns: a tensor of the token ids
"""
if isinstance(obj, str):
return self._encode(obj)
elif isinstance(obj, list):
return torch.stack([self._encode(s) for s in obj], dim=0)
else:
raise NotImplementedError