spikeforge
Install from PyPI:
$ pip install spikeforge
Then train a network:
from spikeforge import TrainingEngine
engine = TrainingEngine(dataset="mnist", hidden=32, epochs=1, num_steps=5)
for metrics in engine.train():
last = metrics
print(last)
engine = TrainingEngine(dataset="mnist", hidden=32, epochs=1, num_steps=5)
for metrics in engine.train():
last = metrics
print(last)
Open dashboardTrain, inspect, and deploy in the browser
Read the docsGuides, reference, examples, and plans
View project statusSee what is ready and what is in progress
Explore on GitHubSource, issues, releases, and contributions
Leaky integrate-and-fire neuron
β=0.90, threshold=1.00, reset=subtract
β=0.90, threshold=1.00, reset=subtract
I[t]
U[t]
S[t]
Why spiking neural networks
Most AI processes every value at every step, even when nothing meaningful changes. Spiking neural networks work more like a nervous system: they send brief signals when events happen. That can help always-on sensors, robots, cameras, and audio tools react quickly while using less energy.
Spikeforge covers the whole workflow: encoding datasets into spikes, training and inspecting networks, comparing deployment targets, and exporting deployment bundles.
Encoding
Training
Topologies
Datasets
Interpreter spine
Introspection
Deployment
Model hub
Dashboard
Status
| Component | Status | Note |
|---|---|---|
| NIR export & validation | done | |
| Determinism & run tracking | done | |
spikeforge-serve |
done | |
| Pipeline builder | done | |
lava_loihi2 backend |
done | |
| Model hub — architectures | done | |
| Model hub — pretrained weights | in progress | |
| On-chip / local learning rules | in progress | |
speck / xylo / spinnaker2 |
in progress | |
| PyPI release | done |