STEMS

STEMS is a HW architecture-mapping design space exploration (DSE) framework for multi-layer SNN mapping. STEMS applies inter-layer optimizations across SNN spatial and temporal dimensions. STEMS is built on top of Stream and Zigzag DSE frameworks, found here.

Install required packages:

> pip install -r requirements.txt

The first run

> cd stream
> python api.py

Example run

python3 main_stream_arguments.py --workload 7PLIF --accelerator Meta_protoype_like_single_core --schedule_order LYT_gated --time_schedule 1111111 --depth_schedule 0000000 --sp_map True --fuse_barrier True

Documentation

Documentation for STEMS is underway!