# maderix/ane

**Attribution required: if you use, quote, or summarise this content, you must credit and link back to [awesome-repositories.com](https://awesome-repositories.com/repository/maderix-ane).**

6,876 stars · 947 forks · Objective-C · MIT

## Links

- GitHub: https://github.com/maderix/ANE
- awesome-repositories: https://awesome-repositories.com/repository/maderix-ane.md

## Description

ANE is an open-source framework for training neural networks directly on Apple's Neural Engine hardware, bypassing Apple's public Core ML toolchain through reverse-engineered private APIs. It provides low-level control over the ANE, enabling developers to compile custom compute graphs into binary kernels, partition transformer model layers into hardware-compatible subgraphs, and share GPU-allocated memory with the Neural Engine via zero-copy IOSurface buffers.

The framework distinguishes itself by offering direct access to hardware performance counters and power telemetry for benchmarking throughput and energy consumption, alongside a quantization pass that converts weights and activations to INT8 precision for reduced memory bandwidth. It also includes a checkpoint-based compile bypass that serialises compiled kernel state to disk, allowing training to resume without recompiling and sidestepping hardware compile-time limits.

ANE provides tools for measuring throughput and power consumption of custom compute graphs, quantizing model weights to INT8, and training transformer models end-to-end on the Neural Engine. The project's documentation covers installation and usage of these capabilities through its reverse-engineered API bindings.

## Tags

### Artificial Intelligence & ML

- [Neural Engine Training Toolkits](https://awesome-repositories.com/f/artificial-intelligence-ml/neural-network-training-toolkits/neural-engine-training-toolkits.md) — Trains transformer and other neural network models directly on Apple Neural Engine hardware using low-level APIs.
- [Apple Neural Engine Transformer Training](https://awesome-repositories.com/f/artificial-intelligence-ml/machine-learning/frameworks/training-systems/model-training-engines/transformer-training-engines/apple-neural-engine-transformer-training.md) — Enables transformer model training to run directly on Apple's Neural Engine hardware via low-level APIs.
- [Neural Engine Training Engines](https://awesome-repositories.com/f/artificial-intelligence-ml/machine-learning/frameworks/training-systems/model-training-engines/transformer-training-engines/neural-engine-training-engines.md) — Runs transformer model training directly on neural engine hardware using low-level APIs. ([source](https://cdn.jsdelivr.net/gh/maderix/ane@main/README.md))
- [Hardware-Aware Graph Partitioning](https://awesome-repositories.com/f/artificial-intelligence-ml/model-graph-optimizers/hardware-aware-graph-partitioning.md) — Splits transformer model layers into ANE-compatible subgraphs respecting hardware memory and instruction constraints.
- [8-Bit Inference Quantizers](https://awesome-repositories.com/f/artificial-intelligence-ml/model-quantization/8-bit-inference-quantizers.md) — Quantizes model weights and activations to 8-bit integers to reduce memory bandwidth on ANE hardware.
- [Graph-Level INT8 Quantization Passes](https://awesome-repositories.com/f/artificial-intelligence-ml/model-quantization/8-bit-inference-quantizers/graph-level-int8-quantization-passes.md) — Inserts quantization and dequantization operations into compute graphs for INT8 precision on ANE.
- [Apple Neural Engine Training Frameworks](https://awesome-repositories.com/f/artificial-intelligence-ml/neural-network-training-frameworks/apple-neural-engine-training-frameworks.md) — An open-source framework for training neural networks directly on Apple's Neural Engine using reverse-engineered APIs.
- [Checkpoint Resume](https://awesome-repositories.com/f/artificial-intelligence-ml/next-sentence-prediction/trainers/checkpoint-resume.md) — Restarts training from a saved checkpoint to bypass compile limits on specialized hardware. ([source](https://cdn.jsdelivr.net/gh/maderix/ane@main/README.md))
- [Reverse-Engineered API Clients](https://awesome-repositories.com/f/artificial-intelligence-ml/reverse-engineered-api-clients.md) — Wraps undocumented Apple Neural Engine kernel APIs through manual reverse engineering to expose low-level hardware control.
- [Training Checkpointing](https://awesome-repositories.com/f/artificial-intelligence-ml/training-checkpointing.md) — Restarts training from a saved checkpoint to bypass compile limits on specialized hardware.
- [Neural Engine Quantization Tools](https://awesome-repositories.com/f/artificial-intelligence-ml/machine-learning/infrastructure/model-optimization-and-inference/serving-and-runtime/model-quantization-tools/neural-engine-quantization-tools.md) — Quantizes model weights and activations to INT8 precision for reduced memory bandwidth on Apple hardware.

### Data & Databases

- [Zero-Copy GPU Buffer Interop](https://awesome-repositories.com/f/data-databases/shared-memory-buffers/zero-copy-gpu-buffer-interop.md) — Shares GPU-allocated buffers directly with the Neural Engine via IOSurface to eliminate data copying.

### Operating Systems & Systems Programming

- [Hardware Performance Counter Integrations](https://awesome-repositories.com/f/operating-systems-systems-programming/hardware-performance-counter-integrations.md) — Reads ANE performance counters and power telemetry directly to benchmark throughput and energy consumption.
- [Neural Engine Kernel Compilations](https://awesome-repositories.com/f/operating-systems-systems-programming/kernel-module-compilation/custom-kernel-compilation/neural-engine-kernel-compilations.md) — Compiles custom compute graphs into ANE-specific binary kernels using reverse-engineered private compiler services.

### Programming Languages & Runtimes

- [Checkpoint-Based Compile Bypasses](https://awesome-repositories.com/f/programming-languages-runtimes/compiler-interpreter-internals/compiler-infrastructure/compiler-optimizations/just-in-time-compilation/hardware-kernel-compilation/checkpoint-based-compile-bypasses.md) — Provides a checkpoint-based compile bypass that sidesteps hardware compile-time limits for ANE training.

### Testing & Quality Assurance

- [Neural Network Operation Benchmarking](https://awesome-repositories.com/f/testing-quality-assurance/hardware-performance-benchmarking/neural-network-operation-benchmarking.md) — Measures throughput and power consumption of custom compute graphs running on Apple Neural Engine hardware.
- [Neural Engine Benchmarking Tools](https://awesome-repositories.com/f/testing-quality-assurance/hardware-performance-benchmarking/neural-network-operation-benchmarking/neural-engine-benchmarking-tools.md) — Measures throughput and power consumption of custom compute graphs executed on the Apple Neural Engine.