SNN Oxford (LAVA-DL)¶

A complete, end-to-end example: a two-layer CubaLIF network trained with LAVA-DL on the Oxford spike-train dataset, exported to NIR, and run through snn-mlir all the way to a compiled binary.

Reference model & training: LAVA-DL SLAYER Oxford tutorial — https://lava-nc.org/lava-lib-dl/slayer/notebooks/oxford/train.html

Network structure¶

Linear(200 → 256) → CubaLIF(256) → Linear(256 → 200) → CubaLIF(200)

Exported as examples/snn_oxford/network.nir. Two Affine/Linear synapse layers feeding two CubaLIF neuron populations, ending on a spike-output layer.

Files in the example¶

File	Role
`network.nir`	The trained network in NIR — the input to `snn-mlir`.
`input.h`	Pre-baked input spike trains (`L0_input[N_STEPS][INPUT_SIZE]`), used by `main.c`.
`target.csv`	Reference output (one row per timestep) for comparing against the compiled run.
`run.py`	Driver: exports `network.mlir` and generates the C runtime into `build/`.
`build/`	Generated artefacts (after running `run.py`): `network.mlir`, `snn_data.h/.c`, `main.c`, `input.h`.

Run it¶

1. Generate MLIR + C runtime files¶

uv run python examples/snn_oxford/run.py              # float32
uv run python examples/snn_oxford/run.py --quantize   # int8 weights, Q12 state
uv run python examples/snn_oxford/run.py --n-steps 50 # fewer timesteps (default 100)

run.py options:

Flag	Default	Effect
`--quantize`	off	int8 weights + Q12 fixed-point neuron state
`--n-steps N`	`100`	Number of simulation timesteps baked into `main.c`

This writes examples/snn_oxford/build/:

network.mlir   ← SNN dialect IR (weights baked in as constant globals)
snn_data.h     ← layer-size constants
main.c         ← memref descriptors + timestep loop + CSV output
input.h        ← copied input data

2. Lower and compile (requires the full toolchain)¶

export MLIR_DIR=/path/to/llvm-project/build/lib/cmake/mlir
bash pipelines/lower_cpu_linux.sh examples/snn_oxford/build/network.mlir
# → examples/snn_oxford/build/network.ll

clang examples/snn_oxford/build/network.ll \
      examples/snn_oxford/build/main.c \
      -o examples/snn_oxford/build/sim

./examples/snn_oxford/build/sim          # prints one CSV row per timestep

3. Compare against the reference¶

The binary prints a CSV row per timestep; compare it against target.csv to confirm the compiled network reproduces the reference output (exactly in float mode; within the expected quantization tolerance in --quantize mode).