Symbol API
E814034
Symbol API is MXNet’s symbolic computation interface for defining, composing, and optimizing deep learning models as static computation graphs.
Statements (49)
| Predicate | Object |
|---|---|
| instanceOf |
software library component
ⓘ
symbolic computation interface ⓘ |
| allows |
declarative model specification
ⓘ
deployment without source code ⓘ serialization of computation graphs ⓘ |
| contrastedWith | MXNet Gluon API NERFINISHED ⓘ |
| developedBy | Apache MXNet community NERFINISHED ⓘ |
| enables |
automatic differentiation
ⓘ
graph-level optimization ⓘ memory optimization ⓘ operator fusion ⓘ |
| graphType | static graph ⓘ |
| hasAdvantage |
performance optimizations at graph level
ⓘ
portable model representation ⓘ |
| hasConcept |
Symbol
ⓘ
computation graph ⓘ dataflow graph ⓘ symbolic operator ⓘ symbolic variable ⓘ |
| hasDisadvantage |
harder to debug than imperative execution
ⓘ
less flexible than imperative APIs ⓘ |
| hasPurpose |
compose deep learning models
ⓘ
define deep learning models ⓘ optimize deep learning models ⓘ |
| introducedIn | early versions of MXNet prior to Gluon ⓘ |
| license | Apache License 2.0 ⓘ |
| optimizedBy | MXNet backend engine NERFINISHED ⓘ |
| partOf |
Apache MXNet
NERFINISHED
ⓘ
MXNet deep learning framework NERFINISHED ⓘ |
| relatedTo |
TensorFlow static graph API
NERFINISHED
ⓘ
Theano symbolic graph API ⓘ |
| status | considered lower-level than Gluon in MXNet ⓘ |
| supports | static computation graphs ⓘ |
| supportsFeature |
custom operators
ⓘ
distributed training ⓘ mixed precision computation ⓘ multi-GPU training ⓘ |
| supportsLanguage |
C++
ⓘ
Julia NERFINISHED ⓘ Python ⓘ R NERFINISHED ⓘ Scala NERFINISHED ⓘ |
| usedFor |
model inference
ⓘ
model training ⓘ neural network definition ⓘ symbolic computation ⓘ |
| usedIn |
computer vision models
ⓘ
natural language processing models ⓘ recommendation systems ⓘ |
Referenced by (1)
Full triples — surface form annotated when it differs from this entity's canonical label.