how could fastembed load the local model?

[helxsz]

since hf connection is not stable, I pre-download all the models needed in the local environment, how could I use the api to load these models ?

embedding_model = TextEmbedding(model_name='/root/.cache/modelscope/hub/models/BAAI/bge-small-en-v1.5')

but got an error saying

Traceback (most recent call last):
  File "/usr/opt/nky-ai/chat-agent/test.py", line 11, in <module>
    embedding_model = TextEmbedding(model_name='/root/.cache/modelscope/hub/models/BAAI/bge-small-en-v1.5') 
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/lib/python3.12/site-packages/fastembed/text/text_embedding.py", line 126, in __init__
    raise ValueError(
ValueError: Model /root/.cache/modelscope/hub/models/BAAI/bge-small-en-v1.5 is not supported in TextEmbedding. Please check the supported models using `TextEmbedding.list_supported_models()`

any idea ?

[joein]

Hey @helxsz

There are several options

1. If hf connection is not stable, but still works, then if you can cache the model the usual way:

TextEmbedding(model_name='BAAI/bge-small-en-v1.5')

You can then launch it with local_files_only option set to true (this param is just propagated to the underlying hf hub library):

TextEmbedding(model_name='BAAI/bge-small-en-v1.5', local_files_only=True)

2. If you downloaded raw model files like model.onnx, etc., then you can do the following:

TextEmbedding(model_name='BAAI/bge-small-en-v1.5', specific_model_path='<path to the model dir>')

[helxsz]

thanks, i just tried

embedding_model = TextEmbedding(model_name='BAAI/bge-small-en-v1.5',specific_model_path='/root/.cache/modelscope/hub/models/BAAI/bge-small-en-v1.5/onnx')

it throws out an error saying

FAIL : Load model from /root/.cache/modelscope/hub/models/BAAI/bge-small-en-v1.5/onnx/model_optimized.onnx failed:system error number 13

but there is only model.onnx not model_optimized.onnx, even i renamed model.onnx to model_optimized.onnx, still got the error.

finally changed the path of ./bge-small-en-v1.5/onnx/model_optimized.onnx to the path of ./bge-small-en-v1.5/model_optimized.onnx.

and it worked.

but loading the model locally in this way is so wasting time