Usage

After you pull the image, you can simply use the openai compatible embedding endpoint:

curl -vvvv -H 'Content-Type: application/json' "http://localhost:11434/v1/embeddings" -d '{
  "input": ["passage: my text 1", "passage: my text 2"],
  "model": "jeffh/intfloat-e5-base-v2"
}'

Where model can also include the tag as necessary (such as jeffh/intfloat-e5-base-v2:f16)

Sizes

The following models quantizations are available:

• f32
• f16
• q8_0

Model Notes

You should prefix your text with query: or passage: for queries and document indexing.

Links

• Ollama Modes
• GitHub Repo
• Original Model