Awọn ọna kika Serialization awoṣe

After training, a model is just numbers (weights) plus a description of its architecture. Serialization writes that state into a file. Different ecosystems use different formats. Python pickle can execute arbitrary code when loading untrusted data. Many PyTorch checkpoints use pickle-based serialization, but torch.load security behavior is version-sensitive: PyTorch 2.6 and later default to weights_only=True when no pickle module is specified. Check the installed documentation and avoid disabling this protection for untrusted files. ONNX (Open Neural Network Exchange) is a framework-neutral format that lets a model trained in PyTorch run in another runtime or language. TensorFlow SavedModel is a directory format used for serving. Keras 3 saves to the native .keras format by default; its model.export API can produce a SavedModel for TensorFlow serving. Check the version-specific API. For large language models, safetensors has become popular because it stores only tensor data in a simple, fast, memory-mappable layout with no code execution, making it both safer and quicker to load. GGUF is widely used for running quantized LLMs efficiently on local hardware.