How-to Guides¶

Practical guides for setting up your environment, managing data, and customizing training and inference.

Setup & Environment¶

Advanced Linux & NVIDIA Setup: Step-by-step instructions for configuring Linux and CUDA.
Working with the CLI: How to use Noether’s command-line interface tools.
Experiment Tracking: Integrating experiment tracking tools like Weights & Biases or trackio.

How to Use Private Data Source: Configuring private data storage like AWS S3.
How to Implement a Custom Dataset: A guide to implementing your own dataset classes.
How to Implement a Custom Multistage Pipeline: Designing complex, multi-step data pipelines.
How to Implement a Custom Sample Processor: Customizing how individual data samples are processed.

How to Implement a Custom Model: Defining new model architectures and layers.
How to Implement a Custom Trainer: Tailoring the training loop for specific research or production needs.
How to Use and Build Callbacks: Extending the framework with custom training callbacks.
How to run distributed training jobs: Running data-parallel training on multiple GPUs and nodes, managed (SLURM) or unmanaged.
Noether Development CLI: How to develop individual components using Noether’s CLI tools.
Configuring AB-UPT: Composing physics blocks and per-domain decoders for AB-UPT.

How to Run Inference and Evaluation on Trained Models: Running pre-trained models for evaluation, prediction generation, or visualization.