tutorials to showcase the usage of deepaudiox

[ChrisNick92]

Proposed tutorial structure:

Notebook 1 — Quick Start

1. Install, load a public dataset (GTZAN is ideal: small, well-known, music genre), train a minimal classifier with default settings in ~10 lines
2. Goal: show the "happy path" — someone can get results without reading anything else

Notebook 2 — Dataset Ingestion

1. Show both loading strategies (directory-based vs. JSON/dict), audio segmentation options, how to inspect the dataset
2. Use SpeechCommands (already in the paper, widely available via torchaudio.datasets)
3. Goal: cover the datasets package thoroughly

Notebook 3 — Backbone & Pooling Exploration

1. Swap backbones and pooling strategies programmatically, compare results on the same dataset
2. Use ESC-50 (also in the paper) — small enough to run quickly
3. Goal: demonstrate the modularity and benchmarking value of the library

Notebook 4 — Advanced Configuration

1. Custom MLPHead, callbacks (EarlyStopper, Checkpointer), learning rate scheduling, loading checkpoints for inference
2. Use CREMA-D or any dataset from previous notebooks
3. Goal: cover production-oriented use cases and extensibility
4. Implement and append a custom backbone following BaseBackbone logic