This document is a glossary of technical terms and acronyms that are specific to SambaNova.
In contrast to some other AI systems, SambaFlow includes a compilation step. You submit your model’s python code to the compiler and the compiler traces the model and generates a PEF file. You can then run the model by passing the PEF file in to the run command. Running a model can mean training the model or performing inference.
SambaNova DataScale® is a fully integrated hardware-software system, powered by a dataflow architecture, that enables organizations to train and deploy models.
- DataScale hardware
SambaNova DataScale is a complete, rack-level, data-center-ready accelerated computing system. Each DataScale system configuration consists of one or more DataScale nodes, integrated networking, and management infrastructure in a standards-compliant data center rack. SambaNova ships several versions of the hardware. For example, the DataScale SN30 is self-contained in a standard 42 rack unit (RU) datacenter rack.
- PEF file
Processor Executable Format file. When you compile a model, the compiler traces the model graph to map it onto an RDU. The PEF file optimally lays out your model on the RDU.
Power Distribution Unit. SambaNova distinguishes two different versions of DataScale SN10 hardware by specifying the number of PDUs. The SN10 2PDU model is the more recent version of DataScale SN10.
SambaNova Runtime allows system administrators perform low-level management tasks. SambaNova Runtime talks directly to the hardware and includes the SambaNova daemon and some configuration tools. See Get started with SambaNova Runtime
SambaNova Reconfigurable Dataflow Unit is a processor that provide native dataflow processing and programmable acceleration. It has a tiled architecture that consists of a network of reconfigurable functional units. The architecture enables a broad set of highly parallelizable patterns contained within dataflow graphs. See the white paper Accelerated computing with a Reconfigurable Dataflow Architecture.
SambaFlow™ is a software stack that is running on DataScale hardware. It consists of several components and includes hardware, software, and models.
- SambaFlow Python API
A software library for compiling and running PyTorch models on the RDU. It contains APIs for model definition, graph tracing, interacting with SambaNova Runtime, and performance benchmarking, as well as the model analyzer and compiler.
The SambaNova Fault Management (SNFM) framework supports reporting, diagnosing, and analyzing the system error and fault events associated with a DataScale system. See SambaNova Fault Management.
- SNML APIs
SambaNova Management Layer APIs. Support many information retrieval tasks and some administration tasks. See SambaNova Management Layer.
- SND (SambaNova Daemon)
The SambaNova Daemon (SND) starts automatically when the host boots. When SND starts, it initializes the hardware resources and loads the driver. When SND is running, you can run applications on the hardware. See Get started with SambaNova Runtime
SambaStudio is a rich, GUI-based platform that provides the functionality to train, deploy, and manage models. See SambaStudio documentation
- SambaNova API (snapi)
Command-line interface (CLI) that is supported in SambaStudio. See SambaStudio CLI setup guide