I’m trying to deploy the LaBSE model from Hugging Face (sentence-transformers/LaBSE · Hugging Face) to AWS SageMaker and use it for embeddings in Weaviate. I have configured Weaviate with the following multi_tenancy
settings:
{
"classes": [
{
"class": "BMP",
"multiTenancyConfig": {"enabled": True},
"description": "File uploaded",
"moduleConfig": {
"text2vec-aws": {
"name": "title_vector",
"region": "us-east-1",
"service": "sagemaker",
"endpoint": "https://runtime.sagemaker.us-east-1.amazonaws.com/endpoints/hf-labse-st-******/invocations"
}
},
"properties": [
{
"name": "fileName",
"dataType": ["text"],
"description": "Name of the file"
},
{
"name": "content",
"dataType": ["text"],
"description": "The content of the file"
},
{
"name": "source",
"dataType": ["text"],
"description": "The source of the uploaded file"
}
],
"vectorizer": "text2vec-aws"
}
]
}
I am encountering repeated errors in response, specifically:
{'error': [{'message': "unmarshal response body: invalid character '<' looking for beginning of value"}]}
This error appears to occur during the response parsing, possibly due to unexpected response content.
Server Setup Information
- Deployment Method: Docker
- Multi Node? Number of Running Nodes: Single Node
- Client Language and Version: Python, weaviate-client version 4.5.5
- Multitenancy?: Enabled
Any Additional Information
I have double-checked the SageMaker endpoint, and the configuration seems correct based on the Weaviate documentation.
I would appreciate any guidance on resolving this error, as it appears to be related to response parsing from SageMaker.