Layer Design Details: dense_w7() - 2024.2 English - XD100

Vitis Tutorials: AI Engine Development (XD100)

Document ID
XD100
Release Date
2024-12-06
Version
2024.2 English

The figure below summarizes the key aspects of the design of the dense_w7() layer. The Jupyter Notebook used for validation is gen_vectors.ipynb. This tile includes three functions, the flatten_w6() layer, the dense_w7() layer, and the final softmax() computation. The layer uses the mac_elem_16_2() intrinsic to compute outputs two at a time. Each intrinsic consumes 32 input channels. The softmax() activation function is computed using the approach outlined in the Softmax Function Vitis Tutorial

figure19