pymc-forecast 0.0.1__tar.gz

pymc_forecast-0.0.1/.gitignore

@@ -0,0 +1,27 @@
+# Claude Code session data
+.claude/
+
+# Python
+__pycache__/
+*.py[cod]
+*.egg-info/
+dist/
+build/
+.venv/
+uv.lock
+
+# Tooling caches
+.pytest_cache/
+.ruff_cache/
+.coverage
+htmlcov/
+
+# Notebooks / editors
+.ipynb_checkpoints/
+.DS_Store
+.idea/
+.vscode/
+
+# Docs
+docs/_build/
+docs/jupyter_execute/

pymc_forecast-0.0.1/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

pymc_forecast-0.0.1/PKG-INFO

@@ -0,0 +1,138 @@
+Metadata-Version: 2.4
+Name: pymc-forecast
+Version: 0.0.1
+Summary: Bayesian time-series forecasting with PyMC: train/forecast plumbing, backtesting, and metrics.
+Project-URL: Repository, https://github.com/pymc-labs/pymc_forecast
+Project-URL: Issues, https://github.com/pymc-labs/pymc_forecast/issues
+Author: PyMC Labs
+License-Expression: Apache-2.0
+License-File: LICENSE
+Keywords: bayesian,forecasting,pymc,time-series
+Classifier: Development Status :: 2 - Pre-Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Mathematics
+Requires-Python: >=3.11
+Requires-Dist: arviz>=0.20
+Requires-Dist: numpy
+Requires-Dist: pooch
+Requires-Dist: pymc>=5.20
+Requires-Dist: xarray
+Provides-Extra: dataframes
+Requires-Dist: pandas; extra == 'dataframes'
+Provides-Extra: extras
+Requires-Dist: pymc-extras>=0.10; extra == 'extras'
+Description-Content-Type: text/markdown
+
+# pymc_forecast
+
+[![CI](https://github.com/pymc-labs/pymc_forecast/actions/workflows/ci.yml/badge.svg)](https://github.com/pymc-labs/pymc_forecast/actions/workflows/ci.yml)
+[![Docs](https://github.com/pymc-labs/pymc_forecast/actions/workflows/docs.yml/badge.svg)](https://pymc-labs.github.io/pymc_forecast/)
+
+Bayesian time-series forecasting with [PyMC](https://www.pymc.io): you write the
+generative model; the package handles the train/forecast plumbing, inference,
+backtesting, and evaluation.
+
+A PyMC port of [numpyro_forecast](https://github.com/juanitorduz/numpyro_forecast)
+(itself a port of Pyro's `pyro.contrib.forecast`) — redesigned around PyMC idioms
+rather than a 1:1 translation.
+
+> **Status: early development.** The design and roadmap live in
+> [PLAN.md](PLAN.md) and the
+> [issue tracker](https://github.com/pymc-labs/pymc_forecast/issues).
+
+**Documentation:** <https://pymc-labs.github.io/pymc_forecast/> — API reference and
+executed example notebooks (univariate, hierarchical, covariates, state-space).
+
+## Quickstart
+
+Write a model with one `predict()` call, fit it, and forecast:
+
+```python
+import numpy as np, pandas as pd, pymc as pm, pytensor.tensor as pt
+from pymc_forecast import Forecaster, backtest, evaluate_forecast, predict, time_series
+
+# a trending weekly series; hold out the last 8 weeks
+dates = pd.date_range("2024-01-07", periods=60, freq="W")
+y = pd.Series(np.cumsum(np.random.default_rng(0).normal(0.2, 1.0, 60)) + 10, index=dates)
+train, test = y.iloc[:52], y.iloc[52:]
+
+def model(h, covariates):
+    # a per-step drift latent; time_series adds the matching `_future` latent
+    drift = time_series(h, "drift", lambda name, dims: pm.Normal(name, 0.0, 0.5, dims=dims))
+    sigma = pm.HalfNormal("sigma", 1.0)
+    predict(
+        h,
+        lambda name, mu, dims, obs: pm.Normal(name, mu, sigma, dims=dims, observed=obs),
+        pt.cumsum(drift),                       # local-linear trend
+    )
+
+fc = Forecaster(model, train, num_steps=5_000, random_seed=0)   # ADVI
+idata = fc.forecast(horizon=8, num_samples=500, random_seed=0)
+forecast = idata["predictions"]["forecast"]     # dims: (chain, draw, time_future)
+
+# score against the held-out weeks (aligned by dim name, not axis position)
+truth = test.to_xarray().rename({"index": "time_future"})
+print(evaluate_forecast(forecast, truth))       # {'mae': ..., 'rmse': ..., 'crps': ..., 'coverage': ...}
+
+# rolling-origin backtest over the whole series
+results = backtest(y, None, model, min_train_window=48, test_window=4, stride=4,
+                   num_samples=200, forecaster_options={"num_steps": 3_000}, random_seed=0)
+```
+
+Swap `Forecaster` for `HMCForecaster` (NUTS, with `nuts_sampler="nutpie"/"numpyro"/...`)
+or `PathfinderForecaster` (pymc-extras) — the fit/forecast interface is identical.
+For models with real covariates, pass full-horizon `covariates` to `.forecast()`
+instead of `horizon=`. See `markov_time_series` for state-space latents and
+`predict_mvn` for observation noise correlated across time.
+
+[pymc-extras statespace](https://github.com/pymc-devs/pymc-extras) structural models
+(level/trend, seasonality, SARIMAX, ...) are first-class citizens too: define one as a
+`StatespaceModel` and fit it with `StatespaceForecaster` — the same `forecast`
+(including exogenous-regression covariates), `predict_in_sample`, `backtest`, and
+metrics calls apply, with the Kalman filter marginalizing the latent states instead
+of sampling them (see `docs/examples/scan_vs_statespace_local_level.ipynb`).
+The pymc-extras integrations (`PathfinderForecaster`, `StatespaceForecaster`) are an
+optional extra: install with `pip install 'pymc-forecast[extras]'`.
+
+## Design principles
+
+- **One model trains and forecasts.** In-sample time latents are fitted; the
+  forecast horizon uses separate `{name}_future` variables that
+  `pm.sample_posterior_predictive` draws from the prior while replaying the
+  posterior for everything else.
+- **Dims and coords everywhere.** No positional axis conventions: variables carry
+  named dims (`"time"`, `"time_future"`, `"obs"`, batch dims), results are
+  `arviz.InferenceData` / `xarray` objects with real coordinates, and metrics are
+  dim-aware.
+- **Not AutoML.** No model zoo, no automatic feature pipelines — a clean path from
+  a hand-written PyMC model to probabilistic forecasts and scores.
+- **Leverage the ecosystem.** ADVI/NUTS from PyMC core, Pathfinder and state-space
+  models from [pymc-extras](https://github.com/pymc-devs/pymc-extras), diagnostics
+  and storage from [ArviZ](https://python.arviz.org).
+
+## Development
+
+Requires Python >= 3.11 and [uv](https://docs.astral.sh/uv/):
+
+```bash
+uv sync --all-extras
+uv run pytest
+uv run ruff check .
+```
+
+To build the documentation (the example notebooks are committed fully executed;
+CI re-executes them with reduced sampling settings):
+
+```bash
+uv sync --all-extras --group docs
+uv run sphinx-build -b html docs docs/_build/html
+```
+
+## License
+
+Apache-2.0. Portions derived from
+[numpyro_forecast](https://github.com/juanitorduz/numpyro_forecast) (Apache-2.0).

pymc_forecast-0.0.1/PLAN.md

@@ -0,0 +1,129 @@
+# Porting `numpyro_forecast` to PyMC
+
+Design and porting plan for building `pymc_forecast` from
+[juanitorduz/numpyro_forecast](https://github.com/juanitorduz/numpyro_forecast)
+(itself a JAX/NumPyro port of Pyro's `pyro.contrib.forecast`).
+
+**This is a redesign, not a translation**: we adopt PyMC idioms throughout —
+named dims/coords instead of positional axis conventions, `InferenceData`/xarray
+results instead of raw arrays, and pymc-extras for Pathfinder and state-space
+models. No API or layout compatibility with the NumPyro version is kept.
+
+The actionable roadmap lives in the
+[issue tracker](https://github.com/pymc-labs/pymc_forecast/issues); this document
+records the source-package analysis and the concept mapping behind those issues.
+
+## What the source package is
+
+A small, focused toolkit for Bayesian time-series forecasting. The user writes the
+generative model; the package handles train/forecast plumbing, inference, and
+evaluation. ~3.3k lines in the main package + ~2k in `functional/` and `contrib/`,
+plus ~40 test modules.
+
+Core design invariants of the source:
+
+1. **One model both trains and forecasts.** In-sample time latents use a fixed site
+   name (e.g. `drift`); the forecast horizon uses a separate `{name}_future` site.
+   The guide/posterior never sees the future site, so `Predictive` replays the
+   posterior for in-sample latents and draws the forecast suffix from the prior.
+2. **Horizon inferred from shapes**: `future = covariates.shape[-2] - data.shape[-2]`.
+3. **Array layout**: time at axis `-2`, observation dim at `-1`, batch dims left
+   (dropped in this port in favor of dims/coords).
+
+Module inventory of the source:
+
+| Module | Lines | Role |
+|---|---|---|
+| `functional/models.py` | 417 | `Horizon`, `time_series`, `markov_time_series` (scan), `predict`, `predict_glm`, `forecasting_model` |
+| `surgery.py` | 368 | Distribution surgery: `shift_loc`, `slice_time`, `prefix_condition` (singledispatch; elementwise families + MVN Gaussian conditional) |
+| `functional/svi.py` | 278 | Guide/optimizer resolution, `fit_svi` (AutoNormal default) |
+| `functional/mcmc.py` | 249 | Kernel resolution, `fit_mcmc` (NUTS default) |
+| `functional/prediction.py` | 279 | `forecast`, `predict_in_sample` via `Predictive`, chunked/vmapped sampling |
+| `functional/posterior.py` | 135 | `draw_posterior` dispatch over fit types |
+| `forecaster.py` | 556 | OOP shims: `ForecastingModel`, `Forecaster` (SVI), `HMCForecaster`, `PathfinderForecaster` |
+| `contrib/blackjax.py` | 581 | Pathfinder VI via BlackJAX |
+| `evaluate.py` | 1090 | `backtest` (expanding/rolling windows), `backtest_vectorized` (vmapped SVI), `evaluate_forecast`, result dataclasses |
+| `metrics.py` | 220 | CRPS, pinball, interval score, MASE factory |
+| `convert.py` | 380 | Fits → ArviZ-schema `xarray.DataTree`, forecast groups |
+| `features.py` | 80 | Fourier design matrices, periodic tiling |
+| `datasets.py` | 137 | BART ridership, Victoria electricity |
+| `arrays.py`, `typing.py`, `exceptions.py`, `optional.py` | ~390 | Helpers, jaxtyping/beartype runtime typing, error taxonomy |
+
+## Concept mapping: NumPyro → PyMC
+
+| numpyro_forecast concept | pymc_forecast equivalent |
+|---|---|
+| Model = pure callable `(covariates, data=None)` | Model-builder callable returning a `pm.Model` (built fresh per fit/forecast call with different horizons) |
+| `Horizon` from array shapes, plates `time` / `time_future` | `Horizon` from coords: dims `"time"` / `"time_future"` with real coordinate values (dates, periods) |
+| `time_series`: site + `{name}_future` site, concat on axis -2 | Same trick: an RV with `dims="time"` plus a fresh `{name}_future` RV with `dims="time_future"`, `pt.concatenate` — `sample_posterior_predictive` replays trace vars and **samples vars missing from the trace from the prior**, which is exactly the `_future`-site mechanism |
+| `markov_time_series` (numpyro `scan`) | `pytensor.scan` inside the model (with `collect_default_updates`) or `CustomDist` with a scan-based dist; forecast scan seeded from the final in-sample state. For linear-Gaussian cases: `pymc_extras.statespace` |
+| `predict(noise_dist, prediction)` + `shift_loc` | Unnecessary as surgery — PyMC likelihoods take `mu` directly. `predict()` becomes API sugar registering `obs` (observed, dims `("time", ...)`) + `obs_future` (unobserved, dims `("time_future", ...)`) + a `forecast` deterministic |
+| `slice_time` / `prefix_condition` (elementwise) | Free: index the latent/params on the future coords when building the future likelihood; conditional = marginal for elementwise noise |
+| `prefix_condition` (MultivariateNormal over time) | Must be ported: explicit Gaussian conditional (Cholesky/solve) in PyTensor, mirroring `_mvn_prefix_condition` |
+| `fit_svi` + `AutoNormal` guide | `pm.fit(method="advi")` (mean-field ≙ AutoNormal); `fullrank_advi` ≙ AutoMultivariateNormal; optimizer via `obj_optimizer=pm.adam(...)` |
+| `fit_mcmc` + NUTS, kernel resolution | `pm.sample()`; backend choice via `nuts_sampler=` (`"pymc"`, `"nutpie"`, `"numpyro"`, `"blackjax"`) replaces kernel resolution |
+| Pathfinder via BlackJAX (581 lines) | Collapses to a thin wrapper over `pymc_extras.fit_pathfinder` |
+| `draw_posterior` (fit-type dispatch) | `approx.sample(n)` for VI; thin/subsample `idata.posterior` for MCMC — normalize both to `InferenceData` |
+| `Predictive` → `forecast` site | `pm.sample_posterior_predictive(idata, var_names=[...], predictions=True)` on the extended-horizon model — the predictions group comes labeled with forecast-time coords |
+| `backtest` (per-window refit) | Same structure, framework-agnostic loop; PRNG keys → integer `random_seed` streams |
+| `backtest_vectorized` (vmapped SVI over windows) | **No vmap equivalent.** Mitigation: rolling windows have fixed shapes → build the window model once with `pm.Data`, iterate via `pm.set_data` to avoid recompiles; nutpie compiled-model reuse for MCMC. Statistical parity, not wall-clock parity |
+| `convert.py` → ArviZ DataTree | Mostly free: PyMC returns `InferenceData` natively and `predictions=True` creates the predictions groups. Keep a small helper for forecast-time coord labeling |
+| `metrics.py` / `evaluate_forecast` (jitted JAX, axis conventions) | Dim-aware xarray/NumPy implementations: metrics take forecast draws and truth as labeled arrays, reduce over named dims |
+| `features.py`, `datasets.py` | Near-verbatim port to NumPy; feature builders return labeled outputs (coords for the feature dim) |
+| jaxtyping/beartype runtime typing | Drop; plain annotations + dim/coord validation at API boundaries |
+| Exceptions taxonomy | Port with renames (guide/kernel resolution errors → VI-method/sampler resolution errors) |
+
+## Where pymc-extras fits
+
+- **`pymc_extras.statespace`**: structural time series (level/trend, seasonality,
+  cycles, AR, regression components), SARIMAX, VARMAX — with Kalman filtering and
+  built-in `.forecast()`. This covers the linear-Gaussian slice of what
+  `markov_time_series` is used for, with exact marginalization instead of sampling
+  per-step latents (usually better posteriors and faster). The port should
+  interoperate: statespace models as first-class citizens in `backtest`/metrics.
+- **`fit_pathfinder`**: replaces the entire BlackJAX contrib module.
+- Scan-based `markov_time_series` remains valuable for arbitrary nonlinear /
+  non-Gaussian transitions that statespace can't express.
+
+## Key design decisions
+
+1. **Dims/coords everywhere (decided).** No positional axis conventions anywhere in
+   the API: model variables carry named dims, forecast outputs are
+   `InferenceData`/xarray with real time coordinates, metrics reduce over named
+   dims. Coords (e.g. `pandas.DatetimeIndex`) flow from input data through to
+   forecast outputs.
+2. **Model API shape.** Keep the two-level design — a functional core where the
+   user writes a model body against a `Horizon` (primitives: `time_series`,
+   `markov_time_series`, `predict`, `predict_glm`) executed inside a managed
+   `pm.Model` context, plus an OOP `ForecastingModel` facade. The builder
+   constructs a fresh model per call (train: no future coords; forecast: extended
+   coords), which is idiomatic PyMC and keeps the "one model definition" invariant.
+3. **Randomness.** `rng_key` threading → `random_seed` integers / numpy Generators;
+   derive per-window seeds deterministically in `backtest`.
+4. **Performance posture.** Accept that per-window refits recompile unless shapes
+   are fixed; document `nuts_sampler="nutpie"` / `"numpyro"` for speed; use
+   `pm.Data` + `set_data` wherever shapes allow.
+
+## What gets dropped or shrinks
+
+- `surgery.py` singledispatch machinery — dissolves into direct likelihood
+  construction (only the MVN conditional survives).
+- `contrib/blackjax.py` (581 lines) → thin wrapper around pymc-extras.
+- `convert.py` (380 lines) → small coords/labeling helper on native `InferenceData`.
+- `arrays.py` axis-layout helpers → replaced by coord handling.
+- jaxtyping/beartype import hook, JAX-specific vmap invariant tests.
+
+## Risks / open questions
+
+- **Posterior replay semantics**: upstream relies on "guide never sees `_future`
+  sites". PyMC's `sample_posterior_predictive` prior-samples missing vars, but
+  deterministic replay vs. resampling of *dependent* vars needs careful
+  `var_names` handling — cover with the upstream consistency tests (SVI vs. MCMC
+  forecasts agree on conjugate examples). Prototype first.
+- **Scan + posterior replay**: replaying scan-internal latents from a trace while
+  continuing the state forward is the trickiest mechanism; prototype early.
+- **Wall-clock**: no vmapped multi-window SVI; decide how much of
+  `backtest_vectorized`'s speed promise to keep.
+- **Multivariate/hierarchical batch semantics**: upstream leans on NumPyro plates +
+  broadcasting; dims/coords cover it, but the shape-heavy tests should be ported
+  wholesale (translated to dims).

pymc_forecast-0.0.1/README.md

@@ -0,0 +1,109 @@
+# pymc_forecast
+
+[![CI](https://github.com/pymc-labs/pymc_forecast/actions/workflows/ci.yml/badge.svg)](https://github.com/pymc-labs/pymc_forecast/actions/workflows/ci.yml)
+[![Docs](https://github.com/pymc-labs/pymc_forecast/actions/workflows/docs.yml/badge.svg)](https://pymc-labs.github.io/pymc_forecast/)
+
+Bayesian time-series forecasting with [PyMC](https://www.pymc.io): you write the
+generative model; the package handles the train/forecast plumbing, inference,
+backtesting, and evaluation.
+
+A PyMC port of [numpyro_forecast](https://github.com/juanitorduz/numpyro_forecast)
+(itself a port of Pyro's `pyro.contrib.forecast`) — redesigned around PyMC idioms
+rather than a 1:1 translation.
+
+> **Status: early development.** The design and roadmap live in
+> [PLAN.md](PLAN.md) and the
+> [issue tracker](https://github.com/pymc-labs/pymc_forecast/issues).
+
+**Documentation:** <https://pymc-labs.github.io/pymc_forecast/> — API reference and
+executed example notebooks (univariate, hierarchical, covariates, state-space).
+
+## Quickstart
+
+Write a model with one `predict()` call, fit it, and forecast:
+
+```python
+import numpy as np, pandas as pd, pymc as pm, pytensor.tensor as pt
+from pymc_forecast import Forecaster, backtest, evaluate_forecast, predict, time_series
+
+# a trending weekly series; hold out the last 8 weeks
+dates = pd.date_range("2024-01-07", periods=60, freq="W")
+y = pd.Series(np.cumsum(np.random.default_rng(0).normal(0.2, 1.0, 60)) + 10, index=dates)
+train, test = y.iloc[:52], y.iloc[52:]
+
+def model(h, covariates):
+    # a per-step drift latent; time_series adds the matching `_future` latent
+    drift = time_series(h, "drift", lambda name, dims: pm.Normal(name, 0.0, 0.5, dims=dims))
+    sigma = pm.HalfNormal("sigma", 1.0)
+    predict(
+        h,
+        lambda name, mu, dims, obs: pm.Normal(name, mu, sigma, dims=dims, observed=obs),
+        pt.cumsum(drift),                       # local-linear trend
+    )
+
+fc = Forecaster(model, train, num_steps=5_000, random_seed=0)   # ADVI
+idata = fc.forecast(horizon=8, num_samples=500, random_seed=0)
+forecast = idata["predictions"]["forecast"]     # dims: (chain, draw, time_future)
+
+# score against the held-out weeks (aligned by dim name, not axis position)
+truth = test.to_xarray().rename({"index": "time_future"})
+print(evaluate_forecast(forecast, truth))       # {'mae': ..., 'rmse': ..., 'crps': ..., 'coverage': ...}
+
+# rolling-origin backtest over the whole series
+results = backtest(y, None, model, min_train_window=48, test_window=4, stride=4,
+                   num_samples=200, forecaster_options={"num_steps": 3_000}, random_seed=0)
+```
+
+Swap `Forecaster` for `HMCForecaster` (NUTS, with `nuts_sampler="nutpie"/"numpyro"/...`)
+or `PathfinderForecaster` (pymc-extras) — the fit/forecast interface is identical.
+For models with real covariates, pass full-horizon `covariates` to `.forecast()`
+instead of `horizon=`. See `markov_time_series` for state-space latents and
+`predict_mvn` for observation noise correlated across time.
+
+[pymc-extras statespace](https://github.com/pymc-devs/pymc-extras) structural models
+(level/trend, seasonality, SARIMAX, ...) are first-class citizens too: define one as a
+`StatespaceModel` and fit it with `StatespaceForecaster` — the same `forecast`
+(including exogenous-regression covariates), `predict_in_sample`, `backtest`, and
+metrics calls apply, with the Kalman filter marginalizing the latent states instead
+of sampling them (see `docs/examples/scan_vs_statespace_local_level.ipynb`).
+The pymc-extras integrations (`PathfinderForecaster`, `StatespaceForecaster`) are an
+optional extra: install with `pip install 'pymc-forecast[extras]'`.
+
+## Design principles
+
+- **One model trains and forecasts.** In-sample time latents are fitted; the
+  forecast horizon uses separate `{name}_future` variables that
+  `pm.sample_posterior_predictive` draws from the prior while replaying the
+  posterior for everything else.
+- **Dims and coords everywhere.** No positional axis conventions: variables carry
+  named dims (`"time"`, `"time_future"`, `"obs"`, batch dims), results are
+  `arviz.InferenceData` / `xarray` objects with real coordinates, and metrics are
+  dim-aware.
+- **Not AutoML.** No model zoo, no automatic feature pipelines — a clean path from
+  a hand-written PyMC model to probabilistic forecasts and scores.
+- **Leverage the ecosystem.** ADVI/NUTS from PyMC core, Pathfinder and state-space
+  models from [pymc-extras](https://github.com/pymc-devs/pymc-extras), diagnostics
+  and storage from [ArviZ](https://python.arviz.org).
+
+## Development
+
+Requires Python >= 3.11 and [uv](https://docs.astral.sh/uv/):
+
+```bash
+uv sync --all-extras
+uv run pytest
+uv run ruff check .
+```
+
+To build the documentation (the example notebooks are committed fully executed;
+CI re-executes them with reduced sampling settings):
+
+```bash
+uv sync --all-extras --group docs
+uv run sphinx-build -b html docs docs/_build/html
+```
+
+## License
+
+Apache-2.0. Portions derived from
+[numpyro_forecast](https://github.com/juanitorduz/numpyro_forecast) (Apache-2.0).