>
    Strata Copilot
    Supported Model Formats
    Focus
    Download PDF
    Focus
    1. Home
    2. Prisma AIRS
    3. Secure Your AI Models with AI Model Security
    4. Supported Model Formats
    Download PDF
    Prisma AIRS

    Supported Model Formats

    Table of Contents

    Supported Model Formats

    Prisma AIRS AI Model Security supports scanning and security checks across various AI and ML model formats.
    Where Can I Use This?What Do I Need?
    • Prisma AIRS (Model Security)
    • Prisma AIRS Model Security License
    AI Model Security checks are supported on the following formats:
    Model NameModel Description
    CNTK ModelsModels saved in Microsoft Cognitive Toolkit format.
    Flax ModelsModels created with Flax, a neural network library for JAX.
    GGUF ModelsGeneral-purpose model format using GGUF.
    Keras Models
    • Legacy Keras Models—Older Keras models, often saved with HDF5.
    • Keras 3 Models—Newer Keras models using the latest Keras version.
    • Keras Pickle Models—Keras models saved with Python pickle.
    • Keras H5 Models—Models in HDF5 format, compatible with legacy and newer versions.
    • Keras Weights—Separate files storing only model weights.
    • Keras JSON—Models saved in JSON format for architecture storage.
    • Keras Metadata—Auxiliary files that store metadata for Keras models.
    KModelKModel files specific to Keras.
    LightGBM ModelsGradient boosting models using LightGBM.
    MS Lite ModelsMicrosoft Lite format for lightweight models.
    MXNet ModelsModels saved in Apache MXNet format.
    Numpy Models
    • Numpy Array Files—Arrays saved in .npy format.
    • Numpy Zip Files—Arrays compressed in .npz format.
    • Numpy Pickle Files—Arrays serialized with pickle.
    OM ModelsModels in Huawei Ascend's OM format.
    ONNX ModelsModels saved in Open Neural Network Exchange format.
    OpenVINO Models
    • OpenVINO Binary Files—Compiled binary files for OpenVINO.
    • OpenVINO XML Files—XML files storing OpenVINO model metadata.
    Pickle FilesModels serialized using Python's pickle.
    PyTorch Models
    • Various PyTorch Versions—Models saved with different PyTorch versions.
    • TorchScript—PyTorch's format for serializing models.
    • PyTorch Archives—Archived files containing serialized models.
    RKNN ModelsModels saved in Rockchip Neural Network (RKNN) format.
    Safetensors
    • Safetensors Models—Models saved using safetensors format for secure tensor storage.
    • Safetensors Index—Index files for safetensors.
    SKLearn ModelsScikit-learn models serialized for deployment.
    TensorRT ModelsNVIDIA's TensorRT models optimized for inference.
    TensorFlow Models
    • SavedModel—TensorFlow's standard saved model format.
    • TFHub—Models from TensorFlow Hub.
    • MetaGraph—TensorFlow's MetaGraph format for exporting graphs.
    • TFLite—Lightweight format for mobile and embedded devices.
    • TFJS—TensorFlow.js format for models running in the browser.
    Torch ModelsGeneral format for PyTorch models.
    JSON FilesJSON-based configurations or model descriptions.

    © 2025 Palo Alto Networks, Inc. All rights reserved.