Slide 30
Slide 30 text
ソフトウェアスタック
mlir
py-videocore
Writing CNN kernels
Doing tests and optimization
qdnn
libnn
Graph Compiler
VC4 NEON
AVX/SSE CPU
Computation Graph Representation
Neural Network API
Pure-GPU
Code Generation No-runtime
Code Generation
runtime library
Model Converters
Designed and Developerd by
Deep leaning frameworks