Embedding Format Mismatch with HuggingFaceInferenceAPIE...

At a glance

A community member is using the HuggingFaceInferenceAPIEmbedding module from llamaindex and is encountering an issue where the embedding returned by the model is not in the expected format. They are using the CODE-BERT model for sentence embedding. The community member is seeking guidance on this issue.

In the comments, another community member suggests that the issue might be related to the response type not being handled properly, and that a pull request (PR) would be needed to fix it. Another community member asks if they should raise an issue on GitHub for the specific CODE-BERT model being used.

There is no explicitly marked answer in the provided information.

Useful resources

SSniper

Has anyone used HuggingFaceInferenceAPIEmbedding module from llamaindex?
I am running into issue where embedding returned by model is in not in the expecteed format what llamaindex is expecting.
I am using CODE-BERT model and using task as sentence embedding
Would love to get some guindance if possible.
CODE:

embed_model = HuggingFaceInferenceAPIEmbedding( model_name='https://*******.us-east-1.aws.endpoints.huggingface.cloud', token='*******' ) text_embedding = embed_model.get_text_embedding("Your text here") print(text_embedding)

emebding format what model is returning
response: b'{"embeddings":[0.02895120158791542,0.4713986814022064,0.41316112875938416,0.2680184543132782,0.4315677583217621 ,...........]
Error:

TypeError: float() argument must be a string or a real number, not 'dict'

3 comments

LLogan M

seems like its not handling the response type properly here? Not sure if huggingface changed the response type or if its specific to this model

LLogan M

would need a PR to fix

SSniper

@Logan M I am using codeBert Model.
do you want me raise a issue on gitHub
https://huggingface.co/microsoft/codebert-base

Add a reply

Find answers from the community

Embedding Format Mismatch with HuggingFaceInferenceAPIEmbedding