Load saved model in GenericLoraModel

[shrimonmuke0202]

Hi,
I want to load my finetuned model which I have fine-tuned using the GenericLoraModel class. But when I try to load the saved model using GenericLoraModel.load("local_path") I got the following error.

in <module> │ │ │ │ 1 from xturing.models.base import BaseModel │ │ 2 │ │ ❱ 3 model = BaseModel.load("saved_model/") │ │ 4 │ │ │ │ /usr/local/lib/python3.9/dist-packages/xturing/models/base.py:24 in load │ │ │ │ 21 │ │ │ return cls.load_from_local(weights_dir_or_model_name) │ │ 22 │ │ else: │ │ 23 │ │ │ print("Loading model from xTuring hub") │ │ ❱ 24 │ │ │ return cls.load_from_hub(weights_dir_or_model_name) │ │ 25 │ │ │ 26 │ @classmethod │ │ 27 │ def load_from_hub(cls, model_name): │ │ │ │ /usr/local/lib/python3.9/dist-packages/xturing/models/base.py:29 in load_from_hub │ │ │ │ 26 │ @classmethod │ │ 27 │ def load_from_hub(cls, model_name): │ │ 28 │ │ hub = ModelHub() │ │ ❱ 29 │ │ model_path = hub.load(model_name) │ │ 30 │ │ return cls.load_from_local(model_path) │ │ 31 │ │ │ 32 │ @classmethod │ │ │ │ /usr/local/lib/python3.9/dist-packages/xturing/utils/hub.py:27 in load │ │ │ │ 24 │ │ return item[len(self.prefix) :], self.static_path_map[item[len(self.prefix) :]] │ │ 25 │ │ │ 26 │ def load(self, path: str): │ │ ❱ 27 │ │ model_name, url = self[path] │ │ 28 │ │ model_dir = self.cache_path / model_name │ │ 29 │ │ │ │ 30 │ │ if not model_dir.exists(): │ │ │ │ /usr/local/lib/python3.9/dist-packages/xturing/utils/hub.py:21 in __getitem__ │ │ │ │ 18 │ │ │ 19 │ def __getitem__(self, item: str): │ │ 20 │ │ if item[: len(self.prefix)] != self.prefix: │ │ ❱ 21 │ │ │ raise ValueError( │ │ 22 │ │ │ │ f"Hub path {item} does not start with {self.prefix}, so it is not a vali │ │ 23 │ │ │ ) │ │ 24 │ │ return item[len(self.prefix) :], self.static_path_map[item[len(self.prefix) :]] │ ╰──────────────────────────────────────────────────────────────────────────────────────────────────╯ ValueError: Hub path saved_model/ does not start with x/, so it is not a valid hub path.

[tushar2407]

Hi @shrimonmuke0202, thanks for pointing this out. We have fixed this issue in our next release. Should be fixed in a day or 2.

[StochasticRomanAgeev]

Hi again, @shrimonmuke0202
We released new version with fix for this issue.
In some minutes also will be in pypy.