Supported Models and Quantization

This section outlines the supported model types for code generation and the quantization configurations available in the DEEPCRAFT™ Model Converter.

Supported Models

You can generate code for TensorFlow, Keras or PyTorch models with the following specifications:

TensorFlow Lite or TensorFlow Keras

You can use TensorFlow models built with Keras 2 saved in (.h5) format. The models exported with Keras 3 are not supported. If you use TensorFlow 2.16+ to build a model: install the tf-keras python package, force Keras 2 with os.environ["TF_USE_LEGACY_KERAS"] = "1" before the statement importing Tensorflow or Keras into your script.

import os
os.environ["TF_USE_LEGACY_KERAS"] = "1"
import tensorflow as tf
from tensorflow import keras

PyTorch

You must use torch==2.6.0, when exporting models to the .pt2 format. Models exported with other PyTorch versions are not supported. For information on exporting pytorch models to .pt2, refer to torch.export  and serialization  section in the PyTorch documentation.

Supported Quantization

The table below lists the supported quantization configurations in Model Converter:

You can quantize Keras (.h5) models to INT8×8 or INT16×8 by providing representative input samples for calibration. For PyTorch models exported as (.pt2), you can quantize the model to int8x8 quantization only.