This repository provides a build system for the TensorFlow C++ API, enabling its use outside the TensorFlow source tree and without Bazel. It targets C++ developers needing to integrate TensorFlow's machine learning capabilities into their applications, offering a simplified build and installation process.

How It Works

The project utilizes CMake to download, build, and install the TensorFlow C++ API. It manages dependencies and configurations, abstracting away the complexities of TensorFlow's native Bazel build system. This approach simplifies the build process and allows for easier integration into existing C++ projects.

Quick Start & Requirements

• Docker: Pre-built images are available on Docker Hub for CPU (floopcz/tensorflow_cc:ubuntu) and GPU (floopcz/tensorflow_cc:ubuntu-cuda).
  • CPU: docker run -it floopcz/tensorflow_cc:ubuntu /bin/bash
  • GPU: docker run --gpus all -it floopcz/tensorflow_cc:ubuntu-cuda /bin/bash
• Manual Build (Ubuntu): Requires cmake, curl, g++-7, git, python3-dev, python3-numpy, wget, and bazel. GPU support requires NVIDIA CUDA Toolkit (>=11.1), drivers, and cuDNN.
• Manual Build (Arch Linux): Requires base-devel, bazel, cmake, git, python, python-numpy, wget. GPU support requires cuda, cudnn, nvidia.
• Build Steps: Clone the repository, create a build directory, run cmake .., make, and sudo make install.
• Resources: Building TensorFlow can be resource-intensive; consider CMake options like -DLOCAL_RAM_RESOURCES and -DLOCAL_CPU_RESOURCES for low-memory environments.
• Documentation: Usage examples and Dockerfiles are provided.

Highlighted Details

• Eliminates the need for Bazel for TensorFlow C++ API builds.
• Provides pre-built Docker images for quick setup and GPU acceleration.
• CMake-based build system simplifies integration into C++ projects.
• Supports both CPU and NVIDIA GPU configurations.

Maintenance & Community

No specific contributors, sponsorships, or community links (Discord/Slack) are mentioned in the README.

Licensing & Compatibility

The repository's license is not explicitly stated in the README. Compatibility for commercial use or closed-source linking would depend on the underlying TensorFlow license and any additional licensing applied to this wrapper project.

Limitations & Caveats

Newer TensorFlow versions may fail to build with the latest Bazel; older Bazel versions might be necessary. Protobuf linkage issues can occur, potentially requiring the -DINSTALL_PROTOBUF=ON CMake option. CPU optimizations are enabled by default for Haswell and newer, which may cause incompatibility with older CPUs if CC_OPT_FLAGS="-march=native" is used.