causalpfn

Run it

$ pip install pfnstudio
$ pfnstudio run causalpfn --input <your file>

# ex: each task on the sample input pfnstudio ships
$ pfnstudio run causalpfn --task estimate --input sample
$ pfnstudio run causalpfn --task rank_causes --input sample

# download it without running · its tasks, inputs, license and digest
$ pfnstudio pull causalpfn
$ pfnstudio show causalpfn

# models loaded now · delete it from this machine
$ pfnstudio ps
$ pfnstudio rm causalpfn

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

Provenance

Readme

A ready-to-use causal-effect model you can point at your own historical data — it estimates how much a treatment, price change, or intervention actually moves the outcome for each unit, without running a live experiment or training a model.

CausalPFN (Balazadeh, Chandak, et al., NeurIPS 2025) is a prior-fitted transformer for amortized causal effect estimation. Given an observational dataset of covariates, treatment, and outcome, it predicts the conditional average treatment effect (CATE) for each unit and the population ATE in a single in-context forward pass. This listing installs the real pretrained weights as an inference-only base model: the checkpoint is pulled from HuggingFace (vdblm/causalpfn) at run time and served through the upstream CATEEstimator. Continued pretraining is not offered — the upstream package ships no training prior.