Metadata-Version: 2.1
Name: lbw-guard
Version: 1.1.3
Summary: Learn-By-Wire Guard optimizer runtime for PyTorch
License: Copyright (c) Qluon Inc. All rights reserved. Licensed under the applicable Qluon license terms.
Requires-Python: >=3.10
Description-Content-Type: text/markdown
Requires-Dist: torch

# LBW_Guard

Learn-By-Wire Guard is a PyTorch optimizer runtime for controlled training workflows.

It provides the public `Guard` API through the `lbw` package and supports freemium and licensed editions for evaluation, standard training, and Guard Pro behavior.

## Install

```bash
pip install LBW_Guard
```

Requirements:

- Python 3.10+
- PyTorch compatible with your environment

## Quick Start

```python
import torch
from lbw import Guard

model = torch.nn.Linear(8, 4)

optimizer = Guard(
    model.parameters(),
    lr=1e-3,
    mode="eval",
)
```

## Editions

- `LBW_Guard_Eval`: freemium evaluation edition with one visible GPU up to 24 GiB RAM, 5,000 optimizer steps, and a 60-day build-based trial window.
- `LBW_Guard_Std`: licensed standard edition.
- `LBW_Guard_Pro`: licensed Guard Pro edition.

Guard Std and Guard Pro require a signed license token.

```python
from lbw import Guard

optimizer = Guard(
    model.parameters(),
    mode="lbw_guard_std",
    license_key=license_token,
)
```

## Public API

Use:

```python
from lbw import Guard
```

`Guard` is the supported public interface for `LBW_Guard_Eval`, `LBW_Guard_Std`, and `LBW_Guard_Pro`.

## Product Information

- Product: `LBW_Guard` / Learn-By-Wire Guard
- Company: Qluon Inc.
- Website: https://www.learnbywire.ai
- Company domain: https://qluon.ai
- Support: support@qluon.ai

Copyright (c) Qluon Inc. All rights reserved.

Licensed under the applicable Qluon license terms.
