metadata

library_name: open_clip
tags:
  - open_clip
  - vision
  - embeddings
inference: false

OpenCLIP ViT-g-14 Model

This is an OpenCLIP model using the ViT-g-14 architecture, pretrained on the LAION-2B dataset.

Usage

Run inference locally using the following example:

import open_clip
from PIL import Image
import torch

# Load the model and preprocessing pipeline
model, preprocess = open_clip.create_model_and_transforms(
    "hf-hub:NikkiZed/openclip-vit-g-14", 
    pretrained="open_clip_pytorch_model.bin"
)

# Load and preprocess an image
image = Image.open("path_to_image.jpg").convert("RGB")
input_tensor = preprocess(image).unsqueeze(0)

# Generate embeddings
with torch.no_grad():
    features = model.encode_image(input_tensor)

print("Image features:", features)