SynapseAI
E660953
SynapseAI is Intel’s software stack designed to optimize and manage AI workloads on Gaudi accelerators for high-performance deep learning.
Statements (48)
| Predicate | Object |
|---|---|
| instanceOf |
AI software platform
ⓘ
software stack ⓘ |
| associatedWith | Intel Gaudi software ecosystem NERFINISHED ⓘ |
| category |
AI accelerator software stack
ⓘ
deep learning framework backend ⓘ |
| compatibleWith |
PCIe Gaudi cards
NERFINISHED
ⓘ
x86 host servers ⓘ |
| designedFor |
Gaudi accelerators
NERFINISHED
ⓘ
Intel Gaudi NERFINISHED ⓘ |
| developer |
Intel
NERFINISHED
ⓘ
Intel Habana Labs NERFINISHED ⓘ |
| includes |
communication libraries
ⓘ
compilers ⓘ drivers ⓘ firmware components ⓘ graph optimizations ⓘ kernel libraries ⓘ monitoring tools ⓘ profiling tools ⓘ runtime libraries ⓘ |
| integratesWith |
ONNX Runtime
NERFINISHED
ⓘ
PyTorch NERFINISHED ⓘ TensorFlow NERFINISHED ⓘ |
| optimizationGoal |
cost efficiency
ⓘ
latency optimization ⓘ power efficiency ⓘ throughput optimization ⓘ |
| primaryUse |
high-performance deep learning
ⓘ
manage AI workloads ⓘ optimize AI workloads ⓘ |
| provides |
automatic kernel selection
ⓘ
cluster management integration ⓘ collective communication primitives ⓘ graph compilation ⓘ memory optimization ⓘ |
| supports |
deep learning models
ⓘ
distributed training ⓘ inference workloads ⓘ multi-device training ⓘ multi-node training ⓘ training workloads ⓘ |
| supportsModelType |
computer vision models
ⓘ
large language models ⓘ recommendation models ⓘ |
| targetDomain |
cloud AI infrastructure
ⓘ
data center AI ⓘ |
| targetHardware |
Intel Gaudi2
NERFINISHED
ⓘ
Intel Gaudi3 NERFINISHED ⓘ |
Referenced by (1)
Full triples — surface form annotated when it differs from this entity's canonical label.