file downloads on package import

[greenpau]

When first importing the package, I noticed a number of downloads happening.

from InstructorEmbedding import INSTRUCTOR
model = INSTRUCTOR('hkunlp/instructor-large')
sentence = "3D ActionSLAM: wearable person tracking in multi-floor environments"
instruction = "Represent the Science title:"
embeddings = model.encode([[instruction,sentence]])
print(embeddings)

I am deploying the package in the environment where I don't have access to Internet.

Is there a way to download all the required files ahead of time and tell the INSTRUCTOR to use the files instead of downloading them at runtime?

[hongjin-su]

Hi, You may try to cache the downloaded model in a local path before the test time, and load it later without Internet access.

from InstructorEmbedding import INSTRUCTOR
model = INSTRUCTOR('hkunlp/instructor-large', cache_folder="local_path")

The downloading only happens in the first time you run the script, and will not require Internet access from the second time.

Feel free to add any questions or comments!

[eshaanagarwal]

I want to download it before since the cache might get deleted

[hongjin-su]

Hi, you may specify any local path for the parameter cache_folder that will not be deleted.

[hiranya911]

Is there anyway to load the model from the git clone? For instance, with AutoModel or AutoTokenizer I can do this:

$ git clone https://huggingface.co/hkunlp/instructor-large

Then in the code:

from transformers import AutoTokenizer

AutoTokenizer.from_pretrained('./instructor-large')

But this doesn't seem to work with the INSTRUCTOR class. Any way to get this to work?

[hongjin-su]

Hi, would you like to share your codes? The following works for me

from InstructorEmbedding import INSTRUCTOR
model = INSTRUCTOR('./instructor-large')

[hiranya911]

Strange. That's what I tried too. But I get this error:

load INSTRUCTOR_Transformer
Traceback (most recent call last):
  File "/opt/homebrew/lib/python3.11/site-packages/transformers/modeling_utils.py", line 415, in load_state_dict
    return torch.load(checkpoint_file, map_location="cpu")
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.11/site-packages/torch/serialization.py", line 815, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.11/site-packages/torch/serialization.py", line 1033, in _legacy_load
    magic_number = pickle_module.load(f, **pickle_load_args)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
_pickle.UnpicklingError: invalid load key, 'v'.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/opt/homebrew/lib/python3.11/site-packages/sentence_transformers/SentenceTransformer.py", line 94, in __init__
    modules = self._load_sbert_model(model_path)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.11/site-packages/InstructorEmbedding/instructor.py", line 474, in _load_sbert_model
    module = module_class.load(os.path.join(model_path, module_config['path']))
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.11/site-packages/InstructorEmbedding/instructor.py", line 306, in load
    return INSTRUCTOR_Transformer(model_name_or_path=input_path, **config)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.11/site-packages/InstructorEmbedding/instructor.py", line 240, in __init__
    self._load_model(self.model_name_or_path, config, cache_dir, **model_args)
  File "/opt/homebrew/lib/python3.11/site-packages/sentence_transformers/models/Transformer.py", line 47, in _load_model
    self._load_t5_model(model_name_or_path, config, cache_dir)
  File "/opt/homebrew/lib/python3.11/site-packages/sentence_transformers/models/Transformer.py", line 55, in _load_t5_model
    self.auto_model = T5EncoderModel.from_pretrained(model_name_or_path, config=config, cache_dir=cache_dir)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.11/site-packages/transformers/modeling_utils.py", line 2429, in from_pretrained
    state_dict = load_state_dict(resolved_archive_file)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.11/site-packages/transformers/modeling_utils.py", line 420, in load_state_dict
    raise OSError(
OSError: You seem to have cloned a repository without having git-lfs installed. Please install git-lfs and run `git lfs install` followed by `git lfs pull` in the folder you cloned.

I'm on Mac btw.

[hongjin-su]

Have you installed git-lfs? You may also try to load the state dict and see whether there is error

[hiranya911]

I don't think I have git-lfs. Just have the usual git client. Is git-lfs a hard requirement for this use case?