Matériel d’IAIn preparation
RTL Design: CNN Accelerators
Map convolution and data reuse to hardware and implement the operations of a small CNN inference workload.
ConvolutionQuantizationInference accelerators
Ce que vous apprendrez
- Map convolution inputs, weights and outputs
- Build quantized, fixed-point reference results
- Supply data through buffers and an arithmetic array
- Connect activation and output processing and compare results
Cours préalable recommandé
RTL Design: Systolic Arrays ↗Programme
- CNN operations and convolution
- Quantization and a reference model
- Data reuse and buffering
- Connecting and checking inference operations