do-pfn
Do-PFN — Interventional
pretrained interventional PFN · warm-start from here
Choose this to continue pretraining or fine-tune an interventional / treatment-effect model on your own data. The target model must be architecture-compatible with Do-PFN (or supply a keymap); the trainer verifies the weights load before continuing.
Run it
$ pip install pfnstudio
$ pfnstudio run do-pfn --input <your file>
# ex:
$ pfnstudio run do-pfn --input sample
# download it without running · its tasks, inputs, license and digest
$ pfnstudio pull do-pfn
$ pfnstudio show do-pfn
# models loaded now · delete it from this machine
$ pfnstudio ps
$ pfnstudio rm do-pfn
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 |
|---|---|
| estimate | what does this lever do? |
Headline numbers
Provenance
| Paper | Do-PFN: In-Context Learning for Causal Effect Estimation ↗ |
| Venue | NeurIPS, 2025 |
| Authors | Jake Robertson, Arik Reuter, Siyuan Guo, Noah Hollmann, Frank Hutter, Bernhard Schölkopf |
| Publisher | Prior Labs ↗ |
| Weights | 7.3M params · upstream .cpkt from GitHub |
| Input | (B, N, d) |
| License | See the upstream Do-PFN repository for license terms |
| Adapter | basemodels/dopfn/adapters/ · do-pfn |
| Source | github.com/jr2021/Do-PFN ↗ |
Readme
A ready-trained interventional model you can build on — start from its learned weights and keep training on your own experiment/observational data instead of starting from zero.
Do-PFN answers "what if we changed this?" — estimating interventional / treatment-effect (CATE) outcomes in-context. This listing publishes its pretrained weights as a base model so a researcher can warm-start a new run from them. The checkpoint is pulled from the upstream Do-PFN GitHub repository at run time and staged into the trainer's init_checkpoint.