Siglip2 So400m Patch16 384

from mlx_embeddings import load, generate
import mlx.core as mx

model, tokenizer = load("mmmorks/siglip2-so400m-patch16-384")

# For image-text embeddings
images = [
    "./images/cats.jpg",  # cats
]
texts = ["a photo of cats", "a photo of a desktop setup", "a photo of a person"]

# Process all image-text pairs
outputs = generate(model, processor, texts, images=images)
logits_per_image = outputs.logits_per_image
probs = mx.sigmoid(logits_per_image) # probabilities for this image
for i, image in enumerate(images):
    print(f"Image {i+1}:")
    for j, text in enumerate(texts):
        print(f"  {probs[i][j]:.1%} match with '{text}'")
    print()

--- license: apache-2.0 tags: - vision - mlx widget: - src: https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg candidate_labels: bee in the sky, bee on the flower example_title: Bee library_name: transformers pipeline_tag: zero-shot-image-classification --- # mmmorks/siglip2-so400m-patch16-384 The Model [mmmorks/siglip2-so400m-patch16-384](https://huggingface.co/mmmorks/siglip2-so400m-patch16-384) was converted to MLX format from [google/siglip2-so400m-patch16-384](https://huggingface.co/google/siglip2-so400m-patch16-384) using mlx-lm version **0.1.0**. ## Use with mlx ```bash pip install mlx-embeddings ``` ```python from mlx_embeddings import load, generate import mlx.core as mx model, tokenizer = load("mmmorks/siglip2-so400m-patch16-384") # For image-text embeddings images = [ "./images/cats.jpg", # cats ] texts = ["a photo of cats", "a photo of a desktop setup", "a photo of a person"] # Process all image-text pairs outputs = generate(model, processor, texts, images=images) logits_per_image = outputs.logits_per_image probs = mx.sigmoid(logits_per_image) # probabilities for this image for i, image in enumerate(images): print(f"Image {i+1}:") for j, text in enumerate(texts): print(f" {probs[i][j]:.1%} match with '{text}'") print() ```