eiq time series 生成模型使用PowerQuad DSP or Neutron N1-16 NPU加速

[xiaobai_226]

问题：eiq time series 生成模型怎么使用PowerQuad DSP or Neutron N1-16 NPU加速

如图：是否需要单独加库文件或者是生成的模型静态库已经包含了PowerQuad DSP or Neutron N1-16 NPU加速？
期望回复：提供使用PowerQuad DSP or Neutron N1-16 NPU加速的文档或者是工程，在tss里面是否有相应的工程可以更新？

[Harry_Zhang]

Hi @xiaobai_226 

Accelerators support. PowerQuad: The NXP hardware IP is useful for speeding up FFT calculation. It is shown and enabled by default when the selected target contains this hardware module.

Neutron NPU: The NXP hardware IP is useful for speeding up deep learning model predictions, which is supported when the selected device contains this hardware module.

BR

Harry

[xiaobai_226 · Accepted solution]

hi Harry
Are you indicating that the library files generated by TSS already incorporate the acceleration processing unit? Since the .lib files produced by TSS do not explicitly indicate whether the processing units are included, and given that in both the SDK examples and TensorFlow for NPU, the NPU is provided in the form of an .a file, it remains unclear whether the TSS-generated files contain the acceleration units. The evaluation board used for testing is the MXCN947EVK.