https://github.com/patw/instructorvec

Create dense vectors using the instructor-large model, running on CPU in quantized mode (fast!)
https://github.com/patw/instructorvec

instructor-embeddings vector-database vectorization

Last synced: 29 days ago
JSON representation

Create dense vectors using the instructor-large model, running on CPU in quantized mode (fast!)

• Host: GitHub
• URL: https://github.com/patw/instructorvec
• Owner: patw
• License: mit
• Created: 2024-01-04T02:41:50.000Z (12 months ago)
• Default Branch: main
• Last Pushed: 2024-02-21T01:58:00.000Z (10 months ago)
• Last Synced: 2024-05-07T18:23:06.747Z (8 months ago)
• Topics: instructor-embeddings, vector-database, vectorization
• Language: Python
• Homepage:
• Size: 6.84 KB
• Stars: 0
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# InstructorVec

A small vector API service for generating quantized instructor-large vectors.  This is used in various other projects under my github repos

## Local Installation

```

pip install -r requirements.txt

```

or

```

pip3 install -r requirements.txt

```

## Local Running

```

uvicorn main:app --host 0.0.0.0 --port 3005

```

or using `python3`

```

python3 -m uvicorn main:app --host 0.0.0.0 --port 3005

```

**Warning**: The first run will be VERY slow to load

Visit `http://localhost:3005/docs` in a browser once it's loaded

Call it in python like this:

```

# Function to call the text embedder

def embed(text):

    response = requests.get(embedder["embedding_endpoint"], params={"text":text, "instruction": "Represent this text for retrieval:" }, headers={"accept": "application/json"})

    vector_embedding = response.json()

    return vector_embedding

```