Metadata-Version: 2.5
Name: pfnstudio
Version: 0.8.24
Summary: CLI for the PFN Studio framework — scaffold, validate, lint, and run prior-fitted foundation model projects.
Project-URL: Homepage, https://pfnstudio.com
Project-URL: Repository, https://github.com/profitopsai/pfnstudio
Project-URL: Issues, https://github.com/profitopsai/pfnstudio/issues
Project-URL: Documentation, https://github.com/profitopsai/pfnstudio/tree/main/docs
Author: PFN Studio contributors
License: Apache-2.0
Keywords: causal-inference,foundation-models,ml,pfn,prior-fitted-networks
Classifier: Development Status :: 3 - Alpha
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Programming Language :: Python :: 3
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.10
Requires-Dist: jsonschema>=4.21
Requires-Dist: packaging>=23
Requires-Dist: pfnstudio-core[serving]>=0.9.10
Requires-Dist: pyyaml>=6.0
Requires-Dist: requests>=2.31
Requires-Dist: rich>=13.7
Requires-Dist: typer>=0.12
Provides-Extra: all
Requires-Dist: huggingface-hub>=0.20; extra == 'all'
Requires-Dist: mlflow>=2.10; extra == 'all'
Requires-Dist: modal>=0.62; extra == 'all'
Requires-Dist: pfnstudio-core[torch]; extra == 'all'
Requires-Dist: pfnstudio-studio>=0.4.0; extra == 'all'
Requires-Dist: runpod>=1.6; extra == 'all'
Requires-Dist: wandb>=0.16; extra == 'all'
Provides-Extra: dev
Requires-Dist: pytest-cov>=5.0; extra == 'dev'
Requires-Dist: pytest>=8.0; extra == 'dev'
Requires-Dist: ruff>=0.4; extra == 'dev'
Provides-Extra: hf
Requires-Dist: huggingface-hub>=0.20; extra == 'hf'
Provides-Extra: mlflow
Requires-Dist: mlflow>=2.10; extra == 'mlflow'
Provides-Extra: modal
Requires-Dist: modal>=0.62; extra == 'modal'
Provides-Extra: runner
Requires-Dist: pfnstudio-core[torch]; extra == 'runner'
Provides-Extra: runpod
Requires-Dist: runpod>=1.6; extra == 'runpod'
Provides-Extra: studio
Requires-Dist: pfnstudio-studio>=0.4.0; extra == 'studio'
Provides-Extra: torch
Requires-Dist: pfnstudio-core[torch]; extra == 'torch'
Provides-Extra: wandb
Requires-Dist: wandb>=0.16; extra == 'wandb'
Description-Content-Type: text/markdown

# pfnstudio — CLI

The command-line interface for [PFN Studio](https://github.com/profitopsai/pfnstudio),
the toolkit for training [prior-fitted foundation models](https://arxiv.org/abs/2112.10510).

## Install

```bash
pip install pfnstudio
```

Run a pretrained base model. The first run lists what the model needs and
this Python lacks (PyTorch, the model's own packages) and asks before
installing it; `-y` installs without asking. Use a virtual environment of its
own so those packages don't change another project's.

```bash
pip install pfnstudio
pfnstudio run ceptron-xs-discovery --input sample
```

`--input sample` runs the sample input pfnstudio ships for that model and
prints the file's path: copy it, put your data in, and pass
`--input your.json`.

For training (requires PyTorch):

```bash
pip install "pfnstudio[torch]"
```

## Commands

```text
pfnstudio init <dir>            # scaffold a new FM project
pfnstudio validate <path>       # check artifacts against JSON Schema
pfnstudio lint <project>        # cross-reference + style checks
pfnstudio sample <prior.yaml>   # draw N tasks from a prior
pfnstudio run <run.yaml>        # execute a training run end-to-end
pfnstudio predict <run-dir>     # inference against a trained checkpoint
pfnstudio export <project>     # tar-gzipped project archive
```

Base models from the marketplace, Ollama-shaped — the CLI is a client of a
local daemon that hosts the models:

```text
pfnstudio run <slug>[:tag]      # pull, pin (lock.json + digest), load into the daemon, run its sample
                                #   asks before installing PyTorch + the model's packages (-y: don't ask)
                                #   --input sample: the shipped sample input · --input f.json: your data
pfnstudio pull <slug>[:tag]     # fetch the adapter bundle + checkpoint into the local store, no run
pfnstudio ls                    # what is in the local store (slug, tag, digest, sizes, loaded?)
pfnstudio ps                    # what the daemon has loaded right now
pfnstudio show <slug>           # listing, lock, tasks (capabilities) of one model
pfnstudio rm <slug>[:tag]       # delete it from the store (asks first; shared weights are kept)
pfnstudio serve                 # run the daemon in the foreground (`run` starts it on demand)
pfnstudio create <name> -f Pfnfile   # compose a model in the studio FROM a base model; `run <name>` uses it
```

`pfnstudio serve <run.yaml> --checkpoint <dir>` is the single-model form the
model container and the studio's Deployments use.

Run `pfnstudio --help` for the full list and `<cmd> --help` for each
subcommand's flags.

## What this CLI is for

PFN Studio organises every PFN project around five first-class
artifacts: **priors** (synthetic data generators), **models** (block
compositions), **evals** (benchmarks + metrics), **runs** (training
manifests), and **initiatives** (research workstreams). This CLI
operates on the file layout those artifacts produce — scaffolding new
projects, validating them, running training, and exporting them for
sharing.

The full story (concepts, architecture, examples, marketplace catalog)
lives at the main repo:
**[github.com/profitopsai/pfnstudio](https://github.com/profitopsai/pfnstudio)**

## License

Apache-2.0.
