CNN.c

A scale of https://github.com/euske/nn1.

How to compile and train a CNN on RISCV (Chipyard edition)

• Requirements:
  1. spike
  2. RISCV toolchain
  3. GNU Make
  4. CMake

To run, simply call make mnist_sim. Make sure to have sourced ee290 environment variables to avoid any dependency issues, but the code should theoretically work on any environment containing the requirements I've listed above.

Organization:

Adding a new neural network

"networks" folder contains the list of different neural networks you can train, which is then exported as a static library.

Adding a new neural network application

Place any new applications within the main folder.