ceptron-xs-discovery
Ceptron XS-discovery — Temporal Causal Discovery
which variable drives which, at which lag · one forward pass
Choose this for a first look at which variables of a time series drive which, with lags. Inference-only. Read the probabilities, not only the list: on very small graphs it tends to mark one link too many.
Run it
$ pip install pfnstudio
$ pfnstudio run ceptron-xs-discovery --input <your file>
# ex:
$ pfnstudio run ceptron-xs-discovery --input sample
# download it without running · its tasks, inputs, license and digest
$ pfnstudio pull ceptron-xs-discovery
$ pfnstudio show ceptron-xs-discovery
# models loaded now · delete it from this machine
$ pfnstudio ps
$ pfnstudio rm ceptron-xs-discovery
Python 3.10+. The first run downloads the model's adapter, asks before installing PyTorch and the model's own packages (-y installs without asking) and pulls the pretrained weights from Hugging Face. --input sample runs the sample input pfnstudio ships and prints its path: copy it, put your data in, and pass it as <your file>. Later runs reuse the cache.
Serves
| Task | Answers |
|---|---|
| discover | what drives what? |
Headline numbers
Provenance
| Publisher | ProfitOps ↗ |
| Weights | pretrained Ceptron XS-discovery · model.safetensors from HuggingFace |
| Input | (T, V) time steps × variables |
| License | CC BY-NC 4.0 (weights; commercial use needs a ProfitOps license) · ceptron code Apache-2.0 |
| Adapter | basemodels/ceptron/adapters/ · ceptron |
| Source | huggingface.co/ProfitOpsAI/ceptron ↗ |
Readme
A fast first causal map of your operations or sales data: which numbers drive which others, and how many days later — before you invest in a deeper analysis.
Ceptron XS-discovery is a prior-fitted network with 65.5M parameters. From a table of observations over time (one column per variable, one row per step) it answers which variables cause which and after how many steps, with a graph-level head deciding whether there is any edge. Strongest on process/operations and retail/demand series (AUROC 94 % and 88 % on its development set). The weights are pulled from HuggingFace (ProfitOpsAI/ceptron) at run time and served through the upstream ceptron package. Weights are CC BY-NC 4.0: non-commercial use; commercial use needs a license from ProfitOps.