I like to learn SpikeForge by giving it a small job and keeping the moving pieces visible. This walkthrough uses the public toolkit, a short experiment and the event path that gained a real train/test split in the latest release cycle.
Start with the SpikeForge repository and install the package in a fresh environment. Pick one of the example datasets already supported by the project. Keep the first run short. A small experiment that you can rerun is more useful than a large run that leaves you guessing about which setting changed the result.
The shape of the experiment is straightforward:
- Load the dataset and turn the samples into the event representation.
- Split the samples before training, so the test set stays out of the update loop.
- Choose a compact network and a short number of epochs.
- Save the configuration beside the result.
- Compare the test output with the training output rather than reporting one number alone.
The train/test split is the important detail. It keeps an event-driven experiment honest when the model is changing quickly. The same configuration can then move through a target adapter or into the SpikeForge model hub with a name, version and small amount of provenance.
SpikeForge is built around Python, PyTorch and snnTorch, but the useful interface is the experiment boundary. You can change the model without changing the dataset loader. You can add a target without rewriting the training loop. That is why the project was split into packages rather than kept as a single directory with a longer README.
What to record
Record the dataset, event conversion, seed, model name, number of epochs and the exact package versions. Put those beside the output rather than in a notebook cell that will disappear. If you publish the result, include the short configuration and a link to the source. Small habits like this make the next experiment faster because you can see what actually changed.
The SpikeForge project page has the current package and dashboard links. I also wrote a four-week Capsize overview with the release context and the other projects that moved at the same time.