# Software Tools

```{toctree}
:hidden:

NeutronConverter
NeutronRunner
TFLiteExtractor
TFLiteOptimizer
TFLiteProfiler
TFLiteQuantizer
```

This documentation describes a cohesive **tools ecosystem designed to prepare, analyze, and optimize machine‑learning
models for efficient deployment**, with a strong focus on TensorFlow Lite–based workflows and Neutron integration. Each
tool in the ecosystem addresses a specific stage of the model lifecycle—from model inspection and transformation to
optimization, quantization, and deployment—while remaining interoperable and easy to compose into automated pipelines.

The guides in this ecosystem are intended to be read both **independently** and **as part of an end‑to‑end workflow**,
allowing developers to adopt only the components they need or to follow a complete, production‑ready path from model
input to optimized output.

* * *

## Tools in This Ecosystem

### **Neutron Converter**

The **{doc}`Neutron Converter <NeutronConverter>`** enables conversion of supported model formats into a
Neutron‑compatible representation. It focuses on correctness, operator coverage, and compatibility with downstream
tooling, forming a critical bridge between model training environments and Neutron-based execution targets.


* * *

### **TFLite Extractor**

The **{doc}`TF Lite Extractor <TFLiteExtractor>`** provides visibility into TensorFlow Lite models by extracting
structural, operator, and metadata information. It is typically used as a first inspection step to understand model
composition and identify optimization opportunities.


* * *

### **TFLite Optimizer**

The **{doc}`TF Lite Optimizer <TFLiteOptimizer>`** applies graph‑level and operator‑level optimizations to improve
performance and reduce resource usage. These transformations preserve model correctness while targeting better execution
efficiency.


* * *

### **TFLite Profiler**

The **{doc}`TF Lite Profiler <TFLiteProfiler>`** analyzes runtime characteristics such as operator execution cost,
memory usage, and bottlenecks. Its output helps guide optimization and quantization decisions based on real performance
data rather than assumptions.


* * *

### **TFLite Quantizer**

The **{doc}`TF Lite Quantizer <TFLiteQuantizer>`** reduces model size and improves execution efficiency by converting
floating‑point models to lower‑precision representations. It supports multiple quantization strategies and integrates
seamlessly with profiling and optimization steps.


* * *

### **Neutron Runner**

The **{doc}`Neutron Runner <NeutronRunner>`** is a validation and execution tool for TensorFlow Lite models that
integrates Neutron's golden kernels into the TFLite runtime. It enables bit-exact validation of models compiled with the
Neutron converter by executing them using reference implementations of Neutron operators.