Sample Stan or PyMC models

These details have not been verified by PyPI

Project links

Source Code

Project description

nutpie: A fast sampler for Bayesian posteriors

Installation

nutpie can be installed using Conda or Mamba from conda-forge with

mamba install -c conda-forge nutpie

Or using pip:

pip install nutpie

To install it from source, install a Rust compiler and maturin and then

maturin develop --release

If you want to use the nightly SIMD implementation for some of the math functions, switch to Rust nightly and then install with the simd_support feature in then nutpie directory:

rustup override set nightly
maturin develop --release --features=simd_support

Usage with PyMC

First, PyMC and Numba need to be installed, for example using

mamba install -c conda-forge pymc numba

We need to create a model:

import pymc as pm
import numpy as np
import nutpie
import pandas as pd
import seaborn as sns

# Load the radon dataset
data = pd.read_csv(pm.get_data("radon.csv"))
data["log_radon"] = data["log_radon"].astype(np.float64)
county_idx, counties = pd.factorize(data.county)
coords = {"county": counties, "obs_id": np.arange(len(county_idx))}

# Create a simple hierarchical model for the radon dataset
with pm.Model(coords=coords, check_bounds=False) as pymc_model:
    intercept = pm.Normal("intercept", sigma=10)

    # County effects
    raw = pm.ZeroSumNormal("county_raw", dims="county")
    sd = pm.HalfNormal("county_sd")
    county_effect = pm.Deterministic("county_effect", raw * sd, dims="county")

    # Global floor effect
    floor_effect = pm.Normal("floor_effect", sigma=2)

    # County:floor interaction
    raw = pm.ZeroSumNormal("county_floor_raw", dims="county")
    sd = pm.HalfNormal("county_floor_sd")
    county_floor_effect = pm.Deterministic(
        "county_floor_effect", raw * sd, dims="county"
    )

    mu = (
        intercept
        + county_effect[county_idx]
        + floor_effect * data.floor.values
        + county_floor_effect[county_idx] * data.floor.values
    )

    sigma = pm.HalfNormal("sigma", sigma=1.5)
    pm.Normal(
        "log_radon", mu=mu, sigma=sigma, observed=data.log_radon.values, dims="obs_id"
    )

We then compile this model and sample form the posterior:

compiled_model = nutpie.compile_pymc_model(pymc_model)
trace_pymc = nutpie.sample(compiled_model)

trace_pymc now contains an ArviZ InferenceData object, including sampling statistics and the posterior of the variables defined above.

We can also control the sampler in a non-blocking way:

# The sampler will now run the the background
sampler = nutpie.sample(compiled_model, blocking=False)

# Pause and resume the sampling
sampler.pause()
sampler.resume()

# Wait for the sampler to finish (up to timeout seconds)
# sampler.wait(timeout=0.1)

# or we can also abort the sampler (and return the incomplete trace)
incomplete_trace = sampler.abort()

# or cancel and discard all progress:
sampler.cancel()

Usage with Stan

In order to sample from Stan model, bridgestan needs to be installed. A pip package is available, but right now this can not be installed using Conda.

pip install bridgestan

When we install nutpie with pip, we can also specify that we want optional dependencies for Stan models using

pip install 'nutpie[stan]'

In addition, a C++ compiler needs to be available. For details see the Stan docs.

We can then compile a Stan model, and sample using nutpie:

import nutpie

code = """
data {
    real mu;
}
parameters {
    real x;
}
model {
    x ~ normal(mu, 1);
}
"""

compiled = nutpie.compile_stan_model(code=code)
# Provide data
compiled = compiled.with_data(mu=3.)
trace = nutpie.sample(compiled)

Advantages

nutpie uses nuts-rs, a library written in Rust, that implements NUTS as in PyMC and Stan, but with a slightly different mass matrix tuning method as those. It often produces a higher effective sample size per gradient evaluation, and tends to converge faster and with fewer gradient evaluation.

Project details

These details have not been verified by PyPI

Project links

Source Code

Release history Release notifications | RSS feed

0.13.2

Jul 26, 2024

This version

0.13.1

Jul 9, 2024

0.13.0

Jul 5, 2024

0.12.0

Jun 29, 2024

0.11.1

Jun 16, 2024

0.11.0

May 29, 2024

0.10.0

Mar 20, 2024

0.9.2

Feb 19, 2024

0.9.1

Sep 14, 2023

0.9.0

Sep 12, 2023

0.8.1

Aug 24, 2023

0.8.0

Aug 18, 2023

0.7.0

Jul 21, 2023

0.6.0

Jul 3, 2023

0.5.0

Jan 3, 2023

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

nutpie-0.13.1.tar.gz (183.9 kB view details)

Uploaded Jul 9, 2024 Source

Built Distributions

nutpie-0.13.1-pp310-pypy310_pp73-manylinux_2_28_x86_64.whl (9.5 MB view details)

Uploaded Jul 9, 2024 PyPy manylinux: glibc 2.28+ x86-64

nutpie-0.13.1-pp310-pypy310_pp73-manylinux_2_28_aarch64.whl (9.7 MB view details)

Uploaded Jul 9, 2024 PyPy manylinux: glibc 2.28+ ARM64

nutpie-0.13.1-cp312-none-win_amd64.whl (1.3 MB view details)

Uploaded Jul 9, 2024 CPython 3.12 Windows x86-64

nutpie-0.13.1-cp312-cp312-manylinux_2_28_x86_64.whl (9.5 MB view details)

Uploaded Jul 9, 2024 CPython 3.12 manylinux: glibc 2.28+ x86-64

nutpie-0.13.1-cp312-cp312-manylinux_2_28_aarch64.whl (9.7 MB view details)

Uploaded Jul 9, 2024 CPython 3.12 manylinux: glibc 2.28+ ARM64

nutpie-0.13.1-cp312-cp312-macosx_11_0_arm64.whl (1.4 MB view details)

Uploaded Jul 9, 2024 CPython 3.12 macOS 11.0+ ARM64

nutpie-0.13.1-cp312-cp312-macosx_10_12_x86_64.whl (1.5 MB view details)

Uploaded Jul 9, 2024 CPython 3.12 macOS 10.12+ x86-64

nutpie-0.13.1-cp311-none-win_amd64.whl (1.3 MB view details)

Uploaded Jul 9, 2024 CPython 3.11 Windows x86-64

nutpie-0.13.1-cp311-cp311-manylinux_2_28_x86_64.whl (9.5 MB view details)

Uploaded Jul 9, 2024 CPython 3.11 manylinux: glibc 2.28+ x86-64

nutpie-0.13.1-cp311-cp311-manylinux_2_28_aarch64.whl (9.7 MB view details)

Uploaded Jul 9, 2024 CPython 3.11 manylinux: glibc 2.28+ ARM64

nutpie-0.13.1-cp311-cp311-macosx_11_0_arm64.whl (1.4 MB view details)

Uploaded Jul 9, 2024 CPython 3.11 macOS 11.0+ ARM64

nutpie-0.13.1-cp311-cp311-macosx_10_12_x86_64.whl (1.5 MB view details)

Uploaded Jul 9, 2024 CPython 3.11 macOS 10.12+ x86-64

nutpie-0.13.1-cp310-none-win_amd64.whl (1.3 MB view details)

Uploaded Jul 9, 2024 CPython 3.10 Windows x86-64

nutpie-0.13.1-cp310-cp310-manylinux_2_28_x86_64.whl (9.5 MB view details)

Uploaded Jul 9, 2024 CPython 3.10 manylinux: glibc 2.28+ x86-64

nutpie-0.13.1-cp310-cp310-manylinux_2_28_aarch64.whl (9.7 MB view details)

Uploaded Jul 9, 2024 CPython 3.10 manylinux: glibc 2.28+ ARM64

nutpie-0.13.1-cp310-cp310-macosx_11_0_arm64.whl (1.4 MB view details)

Uploaded Jul 9, 2024 CPython 3.10 macOS 11.0+ ARM64

nutpie-0.13.1-cp310-cp310-macosx_10_12_x86_64.whl (1.5 MB view details)

Uploaded Jul 9, 2024 CPython 3.10 macOS 10.12+ x86-64

File details

Details for the file nutpie-0.13.1.tar.gz.

File metadata

Download URL: nutpie-0.13.1.tar.gz
Upload date: Jul 9, 2024
Size: 183.9 kB
Tags: Source
Uploaded using Trusted Publishing? No
Uploaded via: maturin/1.7.0

File hashes

Hashes for nutpie-0.13.1.tar.gz
Algorithm	Hash digest
SHA256	`225354701958c0fcf020e87a96335600f157777bbd5c6e64994ff5f12c3dec54`
MD5	`15173fc796ac51ea2bf0241888544b31`
BLAKE2b-256	`89c61a293f75b8171976efbc0b51566deee41118c73c3ae0467b8783fe7cb40c`

Algorithm	Hash digest
SHA256	`a4ed934f76ef789e441b2000446155d24dd9e066908aa3de1dd9821516439ee6`
MD5	`769c146c4ccb10d5c84094798502834a`
BLAKE2b-256	`71b1d4b603201f97b63f00fbf9f45791e17a6f341bb5140080f445a931780447`

Algorithm	Hash digest
SHA256	`6ed2c7dbdbfa7160b61746148ef981073d0631265aec1241ebddee0d03922ea7`
MD5	`0f2baaafa7c311fa0175b7cb144bc501`
BLAKE2b-256	`82e983527511f7730d5141f9302b4cdb527d4a6c1bd436f44365f981a2046072`

Algorithm	Hash digest
SHA256	`1c2d9cf2386029ba71244a56ab802f9da6b9fe85dcb8ff9afd116bbd5cc88032`
MD5	`744ac731ff2ed30a162315f68884b865`
BLAKE2b-256	`ec2339b1a541850b56083756d00e713e906a7a1d8132042672f05c7271e6d534`

Algorithm	Hash digest
SHA256	`a079bf98a9ee678195c02f702895f519a145df807a9ce092ffb0de75938b4f0c`
MD5	`3b3023e390cbbe209347e97af09c47dc`
BLAKE2b-256	`352977e052b53a92545c06e5fb6cbfb6657d19849ad5632222c238048289f825`

Algorithm	Hash digest
SHA256	`98b8868c9f3ef6a91011d81590a73bcae950c06e775cbdad419615450eb6918b`
MD5	`2311baba7bf163a97dc7c546f7ce7fc1`
BLAKE2b-256	`c968b9352cae62f182be30cb766e7b009c95730ad3685e1a53fa286c434e3b62`

Algorithm	Hash digest
SHA256	`4f9cab8165634dfd201e8f1809281bed0e2f95cd99f7b166fc28db5b27c58ef9`
MD5	`5b127904612e80250137a92142f9d917`
BLAKE2b-256	`715c0608b8e36c66ff7b2f16eb15a3c9c7ce1fee601ab14f95252f94594b9a24`

Algorithm	Hash digest
SHA256	`50f7471f431eb934bad34ecdb50d3064a19ade3325190a5b8c05e177f31a7d74`
MD5	`2d3509b19e81d2461344210a84fe74e7`
BLAKE2b-256	`2e24eb9ae157b6e45f1aff912f55e39bc3a22c048f30f69a53fb9a15a1c67dbd`

Algorithm	Hash digest
SHA256	`86722be3d567e89cb2eb7de90ffbb45ce2b0eef93e74f22b1fc49f367cdf8f6d`
MD5	`023bdd2e87dc7f8a6463c496d0085559`
BLAKE2b-256	`4b662426a6e0b1b7224a42c2a678460ca64ae084e332f855d8b444990818301c`

Algorithm	Hash digest
SHA256	`02601da8b7feef640a8ae6f335838689837ad0ac3b25c5b0dad32daa74e112e1`
MD5	`38680ad225f935193c4df9de3f3ba06a`
BLAKE2b-256	`0b3246fb4e4b943c6111997d59f863fb0de9254010b85ac729d348a283820a67`

Algorithm	Hash digest
SHA256	`5dab08b58818e378dc90d6c1a6d470d61fe8e48de3f05b2995342411b277766f`
MD5	`b236b0db4b811928d86f16639f525890`
BLAKE2b-256	`2f573f8b8bff480f963d38e319394476561f9655c4c56ce4996072d4e1ca67de`

Algorithm	Hash digest
SHA256	`6011f21450315f8f98133414360ee6d9ee07df401cf3c8dcfad3bf17bec0df4c`
MD5	`d56507456e7fcb95ef26ccbb845c59b6`
BLAKE2b-256	`f8657b6fa2eb0b073233f6f08f55480766497aa04cf01b7bc3c9f1d213efe83d`

Algorithm	Hash digest
SHA256	`32f12de62b5b5145a5123e4f4c31688a3b889173b3ae958184de0678b19febb5`
MD5	`c7ba09b24674ccd19701a05b70ae87b8`
BLAKE2b-256	`40a154263c596061520357f5db180ad244efd4c71115d1360f559a17c82e2bc5`

Algorithm	Hash digest
SHA256	`ee05ea6a7062bfedfad99d1f472d299a2486ed9a8c8dfa9c9efd8c51a92c1ba2`
MD5	`f78ff3d40c2003f14e73c8ecc88bb9f2`
BLAKE2b-256	`65721a1989ab91116c45b4aacb86fc51b4bb5fcb45b10f072d3607c6406130a3`

Algorithm	Hash digest
SHA256	`141a47590ab210a0a11b6fc20ce73c2cfee6ad8f56ad54e12a50088af75aea71`
MD5	`6f0c1674f33bdd10c93979c44addd444`
BLAKE2b-256	`6c4b0c55c1c34738b2eebf135bbffbe1657bb949f73d7e050e4fe8d9a334879d`

Algorithm	Hash digest
SHA256	`ab9757ebccef606c65a83e335235f7fb331315aac42d5d6b1213cd58d71a917d`
MD5	`b2da8d28998e297bf3d8dbb18fd4882d`
BLAKE2b-256	`550f14ff8122b37a47579f2004da702fa2cd2cc66912d219828b97ac67b330cd`

Algorithm	Hash digest
SHA256	`8b6d3039328005932b1f9e96347d4094cbc26f4678909bdd588df465f2d7c7c0`
MD5	`a61896d7c76e04d4d8a246dc3a1db4ab`
BLAKE2b-256	`90445b43b91615e47e2d1c3ce2ca64db8ed57b7ff879553da3845e809997f678`

Algorithm	Hash digest
SHA256	`81937c68dd0a1e3f1b3103aa475644fdca3daa681f4ce891e5d5becce13b7bd7`
MD5	`c78ca4b883573b3f37f04b50ee2e2c60`
BLAKE2b-256	`ba6ef34f90adfd1f0db393b311d84aac388044ccc970311623cc657bd71ca2ea`

nutpie 0.13.1

Navigation

Verified details

Maintainers

Unverified details

Project links

Meta

Classifiers

Project description

nutpie: A fast sampler for Bayesian posteriors

Installation

Usage with PyMC

Usage with Stan

Advantages

Project details

Verified details

Maintainers

Unverified details

Project links

Meta

Classifiers

Release history Release notifications | RSS feed

Download files

Source Distribution

Built Distributions

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes