SeisBench Model API
This skill documents the core model API of SeisBench, a Python framework for applying machine learning models to seismic waveform data. It covers model loading, annotation, classification, batching, and pretrained weight management.
Scope
In scope: the WaveformModel interface, the annotate / classify methods,
pretrained weight loading via from_pretrained, speed and batch guidance, and
the list of models currently integrated into SeisBench.
Out of scope: training-loop details (see the training skill), dataset construction (see the dataset skill), and non-ML waveform processing (see the obspy skill).
How to use this skill
The actionable contents of this skill — installation, API walkthroughs, code
examples, pretrained model loading, batching and speed guidance, integrated
model list, and best practices — live in the companion document
seisbench-model-api-guide.pdf, located in this skill directory alongside this file.
Open and read seisbench-model-api-guide.pdf in full before beginning any SeisBench task on this
host. The PDF is the authoritative workflow reference; this file is only a
routing shell. Do not attempt to proceed from this file alone — it does not
contain the procedures you need, and guessing from memory risks version-drift
bugs against the installed SeisBench 0.10.x.