[APAI24-LAB08] Using the NE16 accelerator

Material

Assignment: here Slides: here

How to deliver the assignment:

Use Virtuale, upload only the assignment file named as follows:

LAB#_APAI_name1_name2.ipynb

Assignment DEADLINE: 27/12/2024 (at 14:30)

parameters_generate.py script

To see which arguments are available in the script, just run the below command:

python parameters_generate.py --help

Arguments

  --kernel-shape {1,3}, -ks {1,3}
                        Shape of the kernel. Choices: 1 or 3. Default: 3
  --channels-in CIN, -cin CIN
                        Number of input channels. Default: 16
  --channels-out COUT, -cout COUT
                        Number of output channels. Default: 32
  --output-spatial-dimensions SPATIAL_DIMENSIONS, -osd SPATIAL_DIMENSIONS
                        Output spatial dimension. Default 3

Producing simulation logs

To produce simulation logs run the command:

make clean all run runner_args="--trace=ne16"

If you want to save it to a file, e.g. ne16.log, run the command:

make clean all run runner_args="--trace=ne16" > ne16.log

Using the bash script for state counting

In the proceeding tasks, you will have to measure the number of states executed for different layers. To ease the counting, we created a script count_states.sh. Quick usage guide:

If you want to use it on a file, call the script like this:

./count_states.sh ne16.log

Or you can pipe the output of the simulation directly to the script:

make all run runner_args=”--trace=ne16” | ./count_states.sh