SambaFlow learning map

Welcome! This doc page is a learning map for users new to SambaNova. It helps you see the big picture and find the information you need quickly. Here’s an overview:

Tutorials

Many of us learn best by doing. A set of tutorials includes sample code, instructions for running the example, and code discussion in this doc set. The code discussion in this doc set has a special focus on how code for running on RDU is different from code in other environments.

The learning map above points to some additional materials — for example, even if you’re trying out the simplest model, you most likely want to go to the API Reference .

The tutorials in this doc set use different code than tutorials included in /opt/sambaflow/apps. Tutorial examples have been updated and streamlined.

Hello SambaFlow!

The Hello SambaFlow tutorial uses logreg and the MNIST dataset for a simple model. By default, the tutorial code downloads the dataset.

Hello SambaFlow! Compile and run a model includes the code and instructions for how to perform compilation and training.
Examine logreg model code explains the main components of the sample code.

Intermediate tutorial

The intermediate tutorial uses LeNet and the Fashion MNIST dataset. In addition to instructions for compiling and running the tutorial, this tutorial includes dataset download, restarting training from a checkpoint, and running inference.

Compilation, training, and inference steps through a complete machine learning workflow. It includes running inference and visually inspecting labels assigned to data by the inference run.
Code elements of the LeNet tutorial discusses the model code.

Conversion 101

The Conversion 101 tutorial looks at a simple CNN model. Includes original model code and two conversions to RDU: One uses an integrated loss function, another uses an external loss function.

Convert a simple model to SambaFlow includes introduction, planning questions, and how to run the converted model.
Model functions and changes discusses the code that uses an integrated loss function.
Model with an external loss function discusses the code that uses an external loss function.

Concepts

Many of us learn best by understanding the big picture first — having a look at a map before exploring unknown territory. The doc set includes several pages that help you get oriented (or dig deep after initial exploration with the code).

Architecture and workflows. Explains how the SambaFlow components fits into the SambaNova hardware and software stack.
SambaNova compilation overview. Discusses the compiler stack and explains how model compilation works.
White paper. SambaNova Accelerated Computing with a Reconfigurable Dataflow Architecture . Discusses the architecture in some detail. Not required reading, but might be of interest.

Reference

All developers have to rely on reference documentation to get their job done. For SambaFlow, we include the following:

Data preparation and other doc

The following resources in this doc set or elsewhere might help you learn more:

Data preparation scripts. We have a public GitHub repository with two scripts for pretraining data creation, pipeline.py and data_prep.py.
SambaNova Runtime documentation. Information on logs, fault management, and other lower-level procedures.
SambaTune documentation. SambaNova tool for performance optimization (advanced).