Convert ONNX models into MLX callables for accelerating inference on Apple Silicon.

Open a terminal and run the following command:

# Install onnx2mlx
$ pip install --upgrade onnx2mlx

Use the onnx2mlx function to create a callable that uses MLX to run the model:

import mlx.core as mx
import onnx
from onnx2mlx import onnx2mlx

# Load an ONNX model
model = onnx.load("model.onnx")

# Convert to MLX
model_mlx = onnx2mlx(onnx_model)

# Run the MLX model
outputs = model_mlx(mx.array(...))

• Join our Slack community.
• Check out the docs.
• Read our blog.
• Reach out to us at [email protected].