cmdstanpy 1.2.0__tar.gz → 1.2.1__tar.gz

{cmdstanpy-1.2.0/cmdstanpy.egg-info → cmdstanpy-1.2.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: cmdstanpy
-Version: 1.2.0
+Version: 1.2.1
 Summary: Python interface to CmdStan
 Home-page: https://github.com/stan-dev/cmdstanpy
 Author: Stan Dev Team

cmdstanpy-1.2.1/cmdstanpy/_version.py

@@ -0,0 +1,3 @@
+"""PyPi Version"""
+
+__version__ = '1.2.1'

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/cmdstanpy/cmdstan_args.py

@@ -7,7 +7,7 @@ from time import time
 from typing import Any, Dict, List, Mapping, Optional, Union
 
 import numpy as np
-from numpy.random import RandomState
+from numpy.random import default_rng
 
 from cmdstanpy import _TMPDIR
 from cmdstanpy.utils import (
@@ -139,10 +139,9 @@ class SamplerArgs:
                     'Value for iter_warmup must be a non-negative integer,'
                     ' found {}.'.format(self.iter_warmup)
                 )
-            if self.iter_warmup > 0 and not self.adapt_engaged:
+            if self.iter_warmup == 0 and self.adapt_engaged:
                 raise ValueError(
-                    'Argument "adapt_engaged" is False, '
-                    'cannot specify warmup iterations.'
+                    'Must specify iter_warmup > 0 when adapt_engaged=True.'
                 )
         if self.iter_sampling is not None:
             if self.iter_sampling < 0 or not isinstance(
@@ -541,6 +540,8 @@ class PathfinderArgs:
         num_draws: Optional[int] = None,
         num_elbo_draws: Optional[int] = None,
         save_single_paths: bool = False,
+        psis_resample: bool = True,
+        calculate_lp: bool = True,
     ) -> None:
         self.init_alpha = init_alpha
         self.tol_obj = tol_obj
@@ -557,6 +558,8 @@ class PathfinderArgs:
         self.num_elbo_draws = num_elbo_draws
 
         self.save_single_paths = save_single_paths
+        self.psis_resample = psis_resample
+        self.calculate_lp = calculate_lp
 
     def validate(self, _chains: Optional[int] = None) -> None:
         """
@@ -609,6 +612,12 @@ class PathfinderArgs:
         if self.save_single_paths:
             cmd.append('save_single_paths=1')
 
+        if not self.psis_resample:
+            cmd.append('psis_resample=0')
+
+        if not self.calculate_lp:
+            cmd.append('calculate_lp=0')
+
         return cmd
 
 
@@ -868,8 +877,8 @@ class CmdStanArgs:
                 )
 
         if self.seed is None:
-            rng = RandomState()
-            self.seed = rng.randint(1, 99999 + 1)
+            rng = default_rng()
+            self.seed = rng.integers(low=1, high=99999, size=1).item()
         else:
             if not isinstance(self.seed, (int, list, np.integer)):
                 raise ValueError(

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/cmdstanpy/model.py

@@ -80,12 +80,12 @@ class CmdStanModel:
     """
     The constructor method allows model instantiation given either the
     Stan program source file or the compiled executable, or both.
-    By default, the constructor will compile the Stan program on instantiation
-    unless the argument ``compile=False`` is specified.
-    The set of constructor arguments are:
+    This will compile the model if provided a Stan file and no executable,
 
     :param model_name: Model name, used for output file names.
         Optional, default is the base filename of the Stan program file.
+        Deprecated: In version 2.0.0, model name cannot be
+        specified and will always be taken from executable.
 
     :param stan_file: Path to Stan program file.
 
@@ -94,9 +94,8 @@ class CmdStanModel:
         the compiled executable file are specified, the base filenames
         must match, (but different directory locations are allowed).
 
-    :param compile: Whether or not to compile the model.  Default is ``True``.
-        If set to the string ``"force"``, it will always compile even if
-        an existing executable is found.
+    :param force_compile: If ``True``, always compile, even if there
+        is an existing executable file for this model.
 
     :param stanc_options: Options for stanc compiler, specified as a Python
         dictionary containing Stanc3 compiler option name, value pairs.
@@ -109,6 +108,13 @@ class CmdStanModel:
     :param user_header: A path to a header file to include during C++
         compilation.
         Optional.
+
+    :param compile: Whether or not to compile the model.  Default is ``True``.
+        If set to the string ``"force"``, it will always compile even if
+        an existing executable is found.
+        Deprecated: Use ``force_compile`` instead. The ability to instantiate
+        a CmdStanModel without an executable will be removed in version 2.0.0.
+
     """
 
     def __init__(
@@ -126,14 +132,16 @@ class CmdStanModel:
         """
         Initialize object given constructor args.
 
-        :param model_name: Model name, used for output file names.
+        :param model_name: Deprecated. Model name, used for output file names.
         :param stan_file: Path to Stan program file.
         :param exe_file: Path to compiled executable file.
-        :param compile: Whether or not to compile the model.
+        :param force_compile: Whether or not to force recompilation if
+            executable file already exists.
         :param stanc_options: Options for stanc compiler.
         :param cpp_options: Options for C++ compiler.
         :param user_header: A path to a header file to include during C++
             compilation.
+        :param compile: Deprecated. Whether or not to compile the model.
         """
         self._name = ''
         self._stan_file = None
@@ -446,6 +454,9 @@ class CmdStanModel:
         _internal: bool = False,
     ) -> None:
         """
+        Deprecated: To compile a model, use the :class:`~cmdstanpy.CmdStanModel`
+        constructor or :func:`cmdstanpy.compile_stan_file()`.
+
         Compile the given Stan program file.  Translates the Stan code to
         C++, then calls the C++ compiler.
 
@@ -562,7 +573,7 @@ class CmdStanModel:
 
         :param seed: The seed for random number generator. Must be an integer
             between 0 and 2^32 - 1. If unspecified,
-            :class:`numpy.random.RandomState` is used to generate a seed.
+            :func:`numpy.random.default_rng` is used to generate a seed.
 
         :param inits:  Specifies how the sampler initializes parameter values.
             Initialization is either uniform random on a range centered on 0,
@@ -789,7 +800,7 @@ class CmdStanModel:
 
         :param seed: The seed for random number generator. Must be an integer
             between 0 and 2^32 - 1. If unspecified,
-            :class:`numpy.random.RandomState`
+            :func:`numpy.random.default_rng`
             is used to generate a seed which will be used for all chains.
             When the same seed is used across all chains,
             the chain-id is used to advance the RNG to avoid dependent samples.
@@ -1231,7 +1242,7 @@ class CmdStanModel:
 
         :param seed: The seed for random number generator. Must be an integer
             between 0 and 2^32 - 1. If unspecified,
-            :class:`numpy.random.RandomState`
+            :func:`numpy.random.default_rng`
             is used to generate a seed which will be used for all chains.
             *NOTE: Specifying the seed will guarantee the same result for
             multiple invocations of this method with the same inputs.  However
@@ -1432,7 +1443,7 @@ class CmdStanModel:
 
         :param seed: The seed for random number generator. Must be an integer
             between 0 and 2^32 - 1. If unspecified,
-            :class:`numpy.random.RandomState`
+            :func:`numpy.random.default_rng`
             is used to generate a seed which will be used for all chains.
 
         :param inits:  Specifies how the sampler initializes parameter values.
@@ -1612,6 +1623,8 @@ class CmdStanModel:
         draws: Optional[int] = None,
         num_single_draws: Optional[int] = None,
         num_elbo_draws: Optional[int] = None,
+        psis_resample: bool = True,
+        calculate_lp: bool = True,
         # arguments standard to all methods
         seed: Optional[int] = None,
         inits: Union[Dict[str, float], float, str, os.PathLike, None] = None,
@@ -1645,9 +1658,17 @@ class CmdStanModel:
 
         :param num_elbo_draws: Number of Monte Carlo draws to evaluate ELBO.
 
+        :param psis_resample: Whether or not to use Pareto Smoothed Importance
+            Sampling on the result of the individual Pathfinders. If False, the
+            result contains the draws from each path.
+
+        :param calculate_lp: Whether or not to calculate the log probability
+            for approximate draws. If False, this also implies that
+            ``psis_resample`` will be set to False.
+
         :param seed: The seed for random number generator. Must be an integer
             between 0 and 2^32 - 1. If unspecified,
-            :class:`numpy.random.RandomState` is used to generate a seed.
+            :func:`numpy.random.default_rng` is used to generate a seed.
 
         :param inits: Specifies how the algorithm initializes parameter values.
             Initialization is either uniform random on a range centered on 0,
@@ -1726,12 +1747,22 @@ class CmdStanModel:
         Research, 23(306), 1–49. Retrieved from
         http://jmlr.org/papers/v23/21-0889.html
         """
-        if cmdstan_version_before(2, 33, self.exe_info()):
+
+        exe_info = self.exe_info()
+        if cmdstan_version_before(2, 33, exe_info):
             raise ValueError(
                 "Method 'pathfinder' not available for CmdStan versions "
                 "before 2.33"
             )
 
+        if (not psis_resample or not calculate_lp) and cmdstan_version_before(
+            2, 34, exe_info
+        ):
+            raise ValueError(
+                "Arguments 'psis_resample' and 'calculate_lp' are only "
+                "available for CmdStan versions 2.34 and later"
+            )
+
         if num_paths == 1:
             if num_single_draws is None:
                 num_single_draws = draws
@@ -1754,6 +1785,8 @@ class CmdStanModel:
             max_lbfgs_iters=max_lbfgs_iters,
             num_draws=num_single_draws,
             num_elbo_draws=num_elbo_draws,
+            psis_resample=psis_resample,
+            calculate_lp=calculate_lp,
         )
 
         with temp_single_json(data) as _data, temp_inits(inits) as _inits:

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/cmdstanpy/stanfit/gq.py

@@ -323,6 +323,8 @@ class CmdStanGQ(Generic[Fit]):
 
         self._assemble_generated_quantities()
 
+        all_columns = ['chain__', 'iter__', 'draw__'] + list(self.column_names)
+
         gq_cols: List[str] = []
         mcmc_vars: List[str] = []
         if vars is not None:
@@ -341,10 +343,12 @@ class CmdStanGQ(Generic[Fit]):
                             info.start_idx : info.end_idx
                         ]
                     )
+                elif var in ['chain__', 'iter__', 'draw__']:
+                    gq_cols.append(var)
                 else:
                     raise ValueError('Unknown variable: {}'.format(var))
         else:
-            gq_cols = list(self.column_names)
+            gq_cols = all_columns
             vars_list = gq_cols
 
         previous_draws_pd = self._previous_draws_pd(mcmc_vars, inc_warmup)
@@ -369,13 +373,9 @@ class CmdStanGQ(Generic[Fit]):
         )
         draws = np.concatenate([chains_col, iter_col, draw_col, draws], axis=2)
 
-        vars_list = ['chain__', 'iter__', 'draw__'] + vars_list
-        if gq_cols:
-            gq_cols = ['chain__', 'iter__', 'draw__'] + gq_cols
-
         draws_pd = pd.DataFrame(
             data=flatten_chains(draws),
-            columns=['chain__', 'iter__', 'draw__'] + list(self.column_names),
+            columns=all_columns,
         )
 
         if inc_sample and mcmc_vars:

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/cmdstanpy/stanfit/mcmc.py

@@ -441,6 +441,12 @@ class CmdStanMCMC:
                                 self._metric[chain, i, :] = [
                                     float(x) for x in xs
                                 ]
+                    else:  # unit_e changed in 2.34 to have an extra line
+                        pos = fd.tell()
+                        line = fd.readline().strip()
+                        if not line.startswith('#'):
+                            fd.seek(pos)
+
                 # process draws
                 for i in range(sampling_iter_start, num_draws):
                     line = fd.readline().strip()
@@ -609,10 +615,12 @@ class CmdStanMCMC:
                     cols.extend(
                         self.column_names[info.start_idx : info.end_idx]
                     )
+                elif var in ['chain__', 'iter__', 'draw__']:
+                    cols.append(var)
                 else:
                     raise ValueError(f'Unknown variable: {var}')
         else:
-            cols = list(self.column_names)
+            cols = ['chain__', 'iter__', 'draw__'] + list(self.column_names)
 
         draws = self.draws(inc_warmup=inc_warmup)
         # add long-form columns for chain, iteration, draw
@@ -634,8 +642,6 @@ class CmdStanMCMC:
         )
         draws = np.concatenate([chains_col, iter_col, draw_col, draws], axis=2)
 
-        cols = ['chain__', 'iter__', 'draw__'] + cols
-
         return pd.DataFrame(
             data=flatten_chains(draws),
             columns=['chain__', 'iter__', 'draw__'] + list(self.column_names),

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/cmdstanpy/stanfit/pathfinder.py

@@ -206,6 +206,18 @@ class CmdStanPathfinder:
         """
         return self._metadata.cmdstan_config['column_names']  # type: ignore
 
+    @property
+    def is_resampled(self) -> bool:
+        """
+        Returns True if the draws were resampled from several Pathfinder
+        approximations, False otherwise.
+        """
+        return (  # type: ignore
+            self._metadata.cmdstan_config.get("num_paths", 4) > 1
+            and self._metadata.cmdstan_config.get('psis_resample', 1) == 1
+            and self._metadata.cmdstan_config.get('calculate_lp', 1) == 1
+        )
+
     def save_csvfiles(self, dir: Optional[str] = None) -> None:
         """
         Move output CSV files to specified directory.  If files were

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/cmdstanpy/utils/stancsv.py

@@ -289,10 +289,16 @@ def scan_hmc_params(
         raise ValueError(
             'line {}: invalid step size: {}'.format(lineno, step_size)
         ) from e
-    if metric == 'unit_e':
-        return lineno
+    before_metric = fd.tell()
     line = fd.readline().strip()
     lineno += 1
+    if metric == 'unit_e':
+        if line.startswith("# No free parameters"):
+            return lineno
+        else:
+            fd.seek(before_metric)
+            return lineno - 1
+
     if not (
         (
             metric == 'diag_e'

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1/cmdstanpy.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: cmdstanpy
-Version: 1.2.0
+Version: 1.2.1
 Summary: Python interface to CmdStan
 Home-page: https://github.com/stan-dev/cmdstanpy
 Author: Stan Dev Team

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/test/test_cmdstan_args.py

@@ -111,7 +111,7 @@ def test_bad() -> None:
     with pytest.raises(ValueError):
         args.validate(chains=2)
 
-    args = SamplerArgs(iter_warmup=10, adapt_engaged=False)
+    args = SamplerArgs(iter_warmup=0, adapt_engaged=True)
     with pytest.raises(ValueError):
         args.validate(chains=2)
 

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/test/test_generate_quantities.py

@@ -85,9 +85,11 @@ def test_from_csv_files(caplog: pytest.LogCaptureFixture) -> None:
         - 3  # chain, iter, draw duplicates
     )
 
-    assert list(bern_gqs.draws_pd(vars=['y_rep']).columns) == (
-        ["chain__", "iter__", "draw__"] + column_names
-    )
+    assert list(bern_gqs.draws_pd(vars=['y_rep']).columns) == (column_names)
+
+    assert list(
+        bern_gqs.draws_pd(vars=["chain__", "iter__", "draw__", 'y_rep']).columns
+    ) == (["chain__", "iter__", "draw__"] + column_names)
 
 
 def test_pd_xr_agreement():
@@ -315,9 +317,9 @@ def test_save_warmup(caplog: pytest.LogCaptureFixture) -> None:
     assert bern_gqs.draws_pd(inc_warmup=True).shape == (800, 13)
     assert bern_gqs.draws_pd(vars=['y_rep'], inc_warmup=False).shape == (
         400,
-        13,
+        10,
     )
-    assert bern_gqs.draws_pd(vars='y_rep', inc_warmup=False).shape == (400, 13)
+    assert bern_gqs.draws_pd(vars='y_rep', inc_warmup=False).shape == (400, 10)
 
     theta = bern_gqs.stan_variable(var='theta')
     assert theta.shape == (400,)

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/test/test_log_prob.py

@@ -21,13 +21,21 @@ BERN_EXE = os.path.join(DATAFILES_PATH, 'bernoulli' + EXTENSION)
 BERN_BASENAME = 'bernoulli'
 
 
-@pytest.mark.parametrize("sig_figs, expected, expected_unadjusted", [
-    (11, ["-7.0214667713","-1.188472607"], ["-5.5395901199", "-1.4903938392"]),
-    (3, ["-7.02", "-1.19"], ["-5.54", "-1.49"]),
-    (None, ["-7.02147", "-1.18847"], ["-5.53959", "-1.49039"])
-])
-def test_lp_good(sig_figs: Optional[int], expected: List[str],
-                 expected_unadjusted: List[str]) -> None:
+@pytest.mark.parametrize(
+    "sig_figs, expected, expected_unadjusted",
+    [
+        (
+            11,
+            ["-7.0214667713", "-1.188472607"],
+            ["-5.5395901199", "-1.4903938392"],
+        ),
+        (3, ["-7.02", "-1.19"], ["-5.54", "-1.49"]),
+        (None, ["-7.02147", "-1.18847"], ["-5.53959", "-1.49039"]),
+    ],
+)
+def test_lp_good(
+    sig_figs: Optional[int], expected: List[str], expected_unadjusted: List[str]
+) -> None:
     model = CmdStanModel(stan_file=BERN_STAN)
     params = {"theta": 0.34903938392023830482}
     out = model.log_prob(params, data=BERN_DATA, sig_figs=sig_figs)

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/test/test_pathfinder.py

@@ -4,6 +4,7 @@
 
 from pathlib import Path
 
+import numpy as np
 import pytest
 
 import cmdstanpy
@@ -31,6 +32,8 @@ def test_pathfinder_outputs():
     assert theta.shape == (draws,)
     assert 0.23 < theta.mean() < 0.27
 
+    assert pathfinder.is_resampled
+
     assert pathfinder.draws().shape == (draws, 3)
 
 
@@ -58,6 +61,8 @@ def test_single_pathfinder():
         draws=draws,
     )
 
+    assert not pathfinder.is_resampled
+
     theta = pathfinder.theta
     assert theta.shape == (draws,)
 
@@ -122,3 +127,28 @@ def test_pathfinder_init_sampling():
 
     assert fit.chains == 4
     assert fit.draws().shape == (1000, 4, 9)
+
+
+def test_pathfinder_no_psis():
+    stan = DATAFILES_PATH / 'bernoulli.stan'
+    bern_model = cmdstanpy.CmdStanModel(stan_file=stan)
+    jdata = str(DATAFILES_PATH / 'bernoulli.data.json')
+
+    pathfinder = bern_model.pathfinder(data=jdata, psis_resample=False)
+
+    assert not pathfinder.is_resampled
+    assert pathfinder.draws().shape == (4000, 3)
+
+
+def test_pathfinder_no_lp_calc():
+    stan = DATAFILES_PATH / 'bernoulli.stan'
+    bern_model = cmdstanpy.CmdStanModel(stan_file=stan)
+    jdata = str(DATAFILES_PATH / 'bernoulli.data.json')
+
+    pathfinder = bern_model.pathfinder(data=jdata, calculate_lp=False)
+
+    assert not pathfinder.is_resampled
+    assert pathfinder.draws().shape == (4000, 3)
+    n_lp_nan = np.sum(np.isnan(pathfinder.method_variables()['lp__']))
+    assert n_lp_nan < 4000  # some lp still calculated during pathfinder
+    assert n_lp_nan > 3000  # but most are not

{cmdstanpy-1.2.0 → cmdstanpy-1.2.1}/test/test_sample.py

@@ -778,24 +778,19 @@ def test_validate_good_run() -> None:
         fit.runset.chains * fit.num_draws_sampling,
         len(fit.column_names) + 3,
     )
-    assert fit.draws_pd(vars=['theta']).shape == (400, 4)
-    assert fit.draws_pd(vars=['lp__', 'theta']).shape == (400, 5)
-    assert fit.draws_pd(vars=['theta', 'lp__']).shape == (400, 5)
-    assert fit.draws_pd(vars='theta').shape == (400, 4)
+    assert fit.draws_pd(vars=['theta']).shape == (400, 1)
+    assert fit.draws_pd(vars=['lp__', 'theta']).shape == (400, 2)
+    assert fit.draws_pd(vars=['theta', 'lp__']).shape == (400, 2)
+    assert fit.draws_pd(vars='theta').shape == (400, 1)
 
     assert list(fit.draws_pd(vars=['theta', 'lp__']).columns) == [
-        'chain__',
-        'iter__',
-        'draw__',
         'theta',
         'lp__',
     ]
-    assert list(fit.draws_pd(vars=['lp__', 'theta']).columns) == [
-        'chain__',
-        'iter__',
-        'draw__',
+    assert list(fit.draws_pd(vars=['lp__', 'theta', 'iter__']).columns) == [
         'lp__',
         'theta',
+        'iter__',
     ]
 
     summary = fit.summary()
@@ -854,7 +849,7 @@ def test_validate_big_run() -> None:
     assert fit.step_size.shape == (2,)
     assert fit.metric.shape == (2, 2095)
     assert fit.draws().shape == (1000, 2, 2102)
-    assert fit.draws_pd(vars=['phi']).shape == (2000, 2098)
+    assert fit.draws_pd(vars=['phi']).shape == (2000, 2095)
     with raises_nested(ValueError, r'Unknown variable: gamma'):
         fit.draws_pd(vars=['gamma'])
 
@@ -1440,6 +1435,26 @@ def test_dont_save_warmup(caplog: pytest.LogCaptureFixture) -> None:
     )
 
 
+def test_warmup_no_adapt() -> None:
+    # we may want to have a "burn-in" period, even without adaptation
+    stan = os.path.join(DATAFILES_PATH, 'bernoulli.stan')
+    jdata = os.path.join(DATAFILES_PATH, 'bernoulli.data.json')
+
+    bern_model = CmdStanModel(stan_file=stan)
+    bern_fit = bern_model.sample(
+        data=jdata,
+        chains=2,
+        seed=12345,
+        iter_warmup=200,
+        iter_sampling=100,
+        adapt_engaged=False,
+    )
+
+    assert bern_fit.column_names == tuple(BERNOULLI_COLS)
+    assert bern_fit.num_draws_sampling == 100
+    assert bern_fit.draws().shape == (100, 2, len(BERNOULLI_COLS))
+
+
 def test_sampler_diags() -> None:
     stan = os.path.join(DATAFILES_PATH, 'bernoulli.stan')
     jdata = os.path.join(DATAFILES_PATH, 'bernoulli.data.json')

cmdstanpy-1.2.0/cmdstanpy/_version.py

@@ -1,3 +0,0 @@
-"""PyPi Version"""
-
-__version__ = '1.2.0'