Record. Track. Visualize. Deploy.
DevTools for
neuromorphic
hardware.
Nuro is the first experiment tracking platform for neuromorphic research. Capture spike trains from any hardware - Loihi, SpiNNaker, DYNAP-SE2, or GPU simulations. Track experiments with metadata and metrics. Visualize with one line. No experiment tool existed for this field. Now one does.
Early Access — Cloud & Hardware Backends
At a Glance
Every neuromorphic lab has ad-hoc scripts for data analysis. No structured experiment tracking. No cross-hardware recording format. Nuro fixes this.
0
Experiment tools exist
For neuromorphic hardware
6
Hardware adapters
AER, GPU, Loihi, Samna, CSV, HDF5
227
Tests passing
Full stack coverage
1
pip install
To start tracking experiments
How It Works
Record from any hardware. Track experiments with metadata. Visualize results. The compiler handles training and deployment.
Record
Capture spike trains, membrane voltages, weights, and arbitrary signals from any hardware. Hardware-agnostic Recording class works with GPU simulations, Loihi monitors, Samna event streams, or offline CSV/HDF5 data. Numpy-first - no PyTorch required.
Track
Wrap recordings in Experiments with metadata, hardware configs, parameters, and metrics. Save to disk as JSON + HDF5. Compare runs across different chips, neuron models, and hyperparameters. Reproducible by default.
Visualize
Spike rasters, voltage traces, firing rate histograms, population activity, weight matrices, and multi-experiment comparisons. One-line dashboard for any recording. All plots composable with matplotlib.
Train & Deploy
Define spiking networks with a clean Python API. Train on GPU with surrogate gradients. Deploy to Intel Loihi 2, SpiNNaker 2, or BrainChip Akida - zero code changes. The compiler is still here, now with DevTools built in.
Quick Start
Record from any hardware. Track experiments. Visualize with one line. The compiler is still there when you need it.
import numpy as np
import nuro
from nuro.adapters.samna import SamnaAdapter
# Record from hardware — or import from files
# Option A: Import offline event data (no SDK needed)
events = [(0.001, 3), (0.002, 7), (0.005, 3), (0.010, 12)]
rec = SamnaAdapter.from_events(events, num_neurons=128, dt=1e-3)
# Option B: Import from CSV, HDF5, or numpy
rec = nuro.adapters.from_csv("chip_recording.csv", dt=1e-3)
rec = nuro.adapters.from_numpy(spike_array, probe_name="spikes")
# Track experiments with metadata
exp = nuro.experiment("dynap_se2_baseline", project="analog_vision")
exp.set_hardware("dynap-se2", board_id="board_03")
exp.set_params(tau=20e-3, threshold=-50.0, num_neurons=128)
exp.add_recording("main", rec)
exp.log_metrics({"mean_rate": 42.5, "accuracy": 0.89})
exp.save("./experiments") # JSON metadata + HDF5 recordings
# Visualize — one line
nuro.plot.experiment_dashboard(rec, save_path="dashboard.png")
nuro.plot.spike_raster(rec.get("spikes"), dt=rec.dt)
nuro.plot.compare_recordings({"chip_A": rec_a, "chip_B": rec_b})
# The compiler is still here for GPU training + hardware deployment
model = nuro.compile(graph, target="gpu", requires_grad=True)
model = nuro.compile(graph, target="loihi") # auto-quantizesInstall
pip install nuro[devtools]
numpy + h5py + matplotlib
License
Apache 2.0
free to use & modify
No PyTorch needed
numpy-first
analog researchers welcome
Why DevTools
27+ SNN frameworks exist for simulation and training. Zero tools exist for experiment tracking on neuromorphic hardware. We built what researchers actually asked for.
No Experiment Tracking Exists
Every neuromorphic lab has ad-hoc Python scripts for data analysis. No structured way to capture, store, compare, or reproduce experiments across hardware platforms. Nuro fills this gap.
Analog + Digital
Works with Intel Loihi, SpiNNaker, SynSense DYNAP-SE2, BrainScaleS-2, BrainChip Akida, and any chip that produces spike events. Numpy-first means no PyTorch dependency for analysis.
HDF5 + JSON Persistence
Recordings serialize to compressed HDF5 (matches NIR, language-agnostic). Experiment metadata in JSON. Load in Python, MATLAB, Julia, or any HDF5-compatible tool.
The Compiler is Still Here
Train on GPU with surrogate gradients. Deploy to Loihi, SpiNNaker, or Akida with zero code changes. DevTools wraps the compiler - now you can track what happens on the hardware too.
Built For
Analog Neuromorphic Labs
You run experiments on DYNAP-SE2, BrainScaleS-2, or custom analog chips. You need data analysis, logging, and plotting - not another simulator. Nuro captures your event streams and gives you structured experiment tracking.
Digital Neuromorphic Teams
You have Loihi or SpiNNaker access. Nuro compiles your SNN to hardware AND tracks what happens on the chip. Train on GPU, deploy to silicon, compare recordings across platforms.
SNN Researchers
You want reproducible experiments. Log parameters, metrics, and recordings in one place. Compare runs. Share HDF5 files with collaborators. Stop losing results in ad-hoc scripts.
The Vision
DevTools
The first experiment tracking platform for neuromorphic hardware. Record, track, visualize, compare. Works with any chip that produces spike events.
Recording — hardware-agnostic data capture, numpy-first
Experiment — metadata, params, metrics, HDF5 persistence
Plot — rasters, traces, dashboards, multi-experiment compare
Adapters — GPU, Loihi, Samna, CSV, HDF5, NIR
Compiler
The universal abstraction layer between your SNN and any hardware. Train on GPU with surrogate gradients, deploy to neuromorphic silicon with zero code changes.
GPU — PyTorch + surrogate gradients + BPTT
Loihi 2 — Lava backend with auto-quantization
SpiNNaker 2 — weight transfer + delay mapping
Akida — BrainChip edge deployment
Vantar Cloud
Push experiments to the cloud. Compare runs across labs. Share recordings with collaborators. Remote access to neuromorphic chips. Coming v0.9.
Experiment storage — push and pull experiments
Cross-lab sharing — share recordings via link
Remote hardware — run on Loihi/SpiNNaker without owning the chip
Compare — side-by-side experiment analysis
Roadmap
Three layers. SDK to cloud to silicon. Each one is useful standalone — together they are the full stack for event-based neuromorphic AI.
Nuro SDK + Compiler
Open-source Python SDK. Train spiking neural networks on GPU with surrogate gradients. Deploy to Intel Loihi 2, SpiNNaker 2, or BrainChip Akida. NIR interop, ANN-to-SNN conversion, auto-quantization, synaptic delays, and neuromorphic datasets built in.
Neuromorphic DevTools
The first experiment tracking platform for neuromorphic hardware. Hardware-agnostic Recording, Experiment tracking with metadata and metrics, visualization (raster, traces, dashboard), and adapters for GPU, Loihi, Samna, CSV/HDF5. Built from researcher feedback.
Vantar Cloud
Push experiments to the cloud. Compare runs across hardware platforms. Share recordings with collaborators. Remote access to neuromorphic chips you don't own yet. The experiment tracking layer goes online.
Vantar Dev Kit
A complete edge AI module for event-based vision. Hybrid event camera paired with a neuromorphic processor. Nuro pre-installed with DevTools. One device from sensor to inference - at under 1mW.
The missing toolchain for neuromorphic research.
Open source. Apache 2.0. Record, track, visualize, deploy. pip install nuro[devtools]