[BUG] Inaccurate results using remote model

[petrospgithub]

What is the bug?
I followed this tutorial: "https://opensearch.org/docs/2.15/ml-commons-plugin/remote-models/index/" to integrate OpenSearch with an external model via REST calls. I used model https://huggingface.co/nomic-ai/nomic-embed-text-v1.5. Model is deployed successfully and the embeddings created as expected.
Also, I have tested the model locally with some data to be ensure about the quality of the results.

The problem occurs when I try to query the data using neural search. The score is almost zero (e.g., 0.0033772818, 0.0032656149) but it should have been greater than 60 based on my local testing.

What is the expected behavior?
Neural search score is inaccurate when using external models.

What is your host/environment?

• OS: Ubuntu 22.04
• Version: 2.15
• Plugins: "connectors"

[petrospgithub]

It was models problem. The model https://huggingface.co/nomic-ai/nomic-embed-text-v1 works as expected