Module diskchef.chemistry.scikit
Class to read the scikit-learn models
Expand source code
"""Class to read the scikit-learn models"""
from typing import Union
from dataclasses import dataclass
import pickle
import os
import copy
import logging
import numpy as np
from astropy import units as u
from astropy import constants as c
import sklearn
from diskchef.engine.other import PathLike
from diskchef.chemistry.base import ChemistryBase
from diskchef.maps.radiation_fields import DRAINE_UV_FIELD, WEINGARTNER_DRAINE_ISRF, ANDES2_G0
from diskchef.physics.williams_best import WilliamsBest2014
from diskchef.engine.exceptions import CHEFNotImplementedError, CHEFRuntimeError
@dataclass
class SciKitChemistry(ChemistryBase):
"""Predicts chemistry based on `sklearn` trained models
Usage:
>>> physics = WilliamsBest2014(radial_bins=3, vertical_bins=3)
>>> chem = SciKitChemistry(physics)
>>> chem._model # doctest: +NORMALIZE_WHITESPACE
TransformedTargetRegressor(regressor=Pipeline(steps=[('quantiletransformer',
QuantileTransformer(n_quantiles=100)),
('kneighborsregressor',
KNeighborsRegressor(n_neighbors=62))]),
transformer=QuantileTransformer(n_quantiles=100,
output_distribution='normal'))
>>> chem._model.X
('log(numberdens)', 'log(temperature)')
>>> chem._model.y
('CO', 'CN', 'HCN', 'HNC', 'HCO+', 'H2CO', 'N2H+', 'CS', 'C2H', 'e-')
>>> chem.run_chemistry()
>>> chem.table # doctest: +NORMALIZE_WHITESPACE
<CTable length=9>
Radius Height Height to radius Gas density Dust density Gas temperature Dust temperature n(H+2H2) log(numberdens) log(temperature) CO CN HCN HNC HCO+ H2CO N2H+ CS C2H e-
AU AU g / cm3 g / cm3 K K 1 / cm3
float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64
------------ ------------ ---------------- ------------ ------------ --------------- ---------------- ------------ --------------- ---------------- ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------
1.000000e-01 0.000000e+00 0.000000e+00 1.153290e-15 1.153290e-17 7.096268e+02 7.096268e+02 5.095483e+08 8.474334e+00 2.851030e+00 1.664660e-10 2.075371e-13 7.722633e-14 4.447148e-14 5.587394e-12 5.853986e-16 3.863739e-17 3.872823e-16 3.844251e-13 7.300494e-05
1.000000e-01 3.500000e-02 3.500000e-01 5.024587e-34 5.024587e-36 3.548134e+03 3.548134e+03 2.219970e-10 -9.886505e+00 3.550000e+00 7.899202e-20 6.030650e-22 1.383617e-33 3.440258e-34 1.810430e-25 8.080346e-35 3.133545e-28 8.086328e-35 3.225259e-34 6.885677e-03
1.000000e-01 7.000000e-02 7.000000e-01 5.921268e-72 5.921268e-74 3.548134e+03 3.548134e+03 2.616143e-48 -4.781519e+01 3.550000e+00 7.899202e-20 6.030650e-22 1.383617e-33 3.440258e-34 1.810430e-25 8.080346e-35 3.133545e-28 8.086328e-35 3.225259e-34 6.885677e-03
7.071068e+00 0.000000e+00 0.000000e+00 2.285168e-13 2.285168e-15 6.820453e+01 6.820453e+01 1.009636e+11 1.077131e+01 1.833813e+00 6.431126e-05 2.138011e-18 3.018989e-14 2.846458e-14 4.012340e-12 4.361136e-15 3.347467e-15 9.598290e-12 3.676756e-18 3.737908e-11
7.071068e+00 2.474874e+00 3.500000e-01 2.716386e-17 2.716386e-19 3.410227e+02 3.410227e+02 1.200157e+07 6.846386e+00 2.532783e+00 3.196870e-15 9.588096e-18 7.029195e-25 5.756870e-26 2.279574e-19 9.920758e-30 2.743343e-21 2.316110e-28 1.619243e-28 2.640124e-04
7.071068e+00 4.949747e+00 7.000000e-01 7.189026e-23 7.189026e-25 3.410227e+02 3.410227e+02 3.176265e+01 1.269065e+00 2.532783e+00 7.899202e-20 6.030650e-22 1.383617e-33 3.440258e-34 1.810430e-25 8.080346e-35 3.133545e-28 8.086328e-35 3.225259e-34 6.885677e-03
5.000000e+02 0.000000e+00 0.000000e+00 2.871710e-20 2.871710e-22 6.555359e+00 6.555359e+00 1.268783e+04 3.870536e+00 8.165965e-01 7.246437e-09 1.639502e-09 9.789595e-12 6.854840e-12 1.985437e-12 3.611351e-12 1.609762e-16 2.563294e-12 2.442295e-10 4.478516e-05
5.000000e+02 1.750000e+02 3.500000e-01 6.929036e-22 6.929036e-24 2.625083e+01 2.625083e+01 3.061396e+02 2.253068e+00 1.419143e+00 9.592822e-12 3.264297e-14 2.652106e-14 1.248681e-14 8.773964e-14 4.373150e-18 6.765207e-18 1.123771e-21 1.185412e-18 7.771940e-05
5.000000e+02 3.500000e+02 7.000000e-01 8.170464e-23 8.170464e-25 3.277680e+01 3.277680e+01 3.609885e+01 1.324642e+00 1.515566e+00 3.185463e-12 1.816816e-16 2.245415e-18 1.057603e-18 5.366938e-14 2.360676e-22 1.899195e-20 9.770146e-23 1.644348e-23 8.516644e-05
"""
model: Union[PathLike, sklearn.base.TransformerMixin] = os.path.join(
os.path.dirname(os.path.abspath(__file__)),
"scikit_estimators", "andes2_atomic_knearest_temp_dens.pkl"
)
def __post_init__(self):
super().__post_init__()
if hasattr(self.model, "predict"):
self._model = self.model
self.logger.info("SciKit Transformer object is recognized")
else:
self._model = self.load_scikit_model(self.model)
def run_chemistry(self):
all_arguments = copy.copy(self._model.X)
existing_arguments_with_log = {f"log({colname})" for colname in self.table.colnames}
for argument in all_arguments:
if argument in self.table.colnames:
pass
elif argument in existing_arguments_with_log:
self.table[f"log({argument})"] = np.log10(
(self.table[argument]).cgs.value
)
self.logger.info("log(%s) was taken in CGS system", argument)
elif argument == "log(numberdens)":
self.table["log(numberdens)"] = np.log10(
(self.table["Gas density"] / self.molar_mass * c.N_A).to(u.cm ** -3).value
)
elif argument == "log(temperature)":
self.table["log(temperature)"] = np.log10(self.table["Gas temperature"].to(u.K).value)
elif argument == "log(uv)":
self.table["log(uv)"] = np.log10(
(self.table["UV radiation strength"]).to_value(ANDES2_G0))
elif argument == "log(ionization)":
self.table["log(ionization)"] = np.log10(self.table["Ionization rate"].to(1 / u.s).value)
else:
raise CHEFRuntimeError(f"{argument} or its precursor is not found in the original data")
X = np.nan_to_num(np.array([self.table[col] for col in self._model.X]).T)
y = self._model.predict(X)
for species, abunds in zip(self._model.y, y.T):
self.table[species] = 10 ** abunds
@staticmethod
def load_scikit_model(filename: PathLike) -> sklearn.base.RegressorMixin:
"""
Read pickled scikit regressor. Can be used as model argument to SciKitChemistry to avoid multiple loads
Args:
filename: path to file to be unpickled
Returns:
Regressor-like object with .predict(X), X, and y attributes
Raises:
FileNotFoundError if file is not found
CHEFNotImplementedError / NotImplementedError if unpickled object does not have the needed attributes
"""
with open(filename, 'rb') as fff:
model = pickle.load(fff)
if hasattr(model, "predict") and hasattr(model, "X") and hasattr(model, 'y'):
logging.info("SciKit Transformer object is read and recognized")
else:
raise CHEFNotImplementedError
return modelClasses
class SciKitChemistry (physics: PhysicsBase = None, molar_mass: Unit("g / mol") = <Quantity 2.33 g / mol>, hydrogen_mass_fraction: float = 0.739, model: Union[str, os.PathLike, sklearn.base.TransformerMixin] = '/builds/SmirnGreg/diskchef/diskchef/chemistry/scikit_estimators/andes2_atomic_knearest_temp_dens.pkl')-
Predicts chemistry based on
sklearntrained modelsUsage:
>>> physics = WilliamsBest2014(radial_bins=3, vertical_bins=3) >>> chem = SciKitChemistry(physics) >>> chem._model # doctest: +NORMALIZE_WHITESPACE TransformedTargetRegressor(regressor=Pipeline(steps=[('quantiletransformer', QuantileTransformer(n_quantiles=100)), ('kneighborsregressor', KNeighborsRegressor(n_neighbors=62))]), transformer=QuantileTransformer(n_quantiles=100, output_distribution='normal')) >>> chem._model.X ('log(numberdens)', 'log(temperature)') >>> chem._model.y ('CO', 'CN', 'HCN', 'HNC', 'HCO+', 'H2CO', 'N2H+', 'CS', 'C2H', 'e-') >>> chem.run_chemistry() >>> chem.table # doctest: +NORMALIZE_WHITESPACE <CTable length=9> Radius Height Height to radius Gas density Dust density Gas temperature Dust temperature n(H+2H2) log(numberdens) log(temperature) CO CN HCN HNC HCO+ H2CO N2H+ CS C2H e- AU AU g / cm3 g / cm3 K K 1 / cm3 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 ------------ ------------ ---------------- ------------ ------------ --------------- ---------------- ------------ --------------- ---------------- ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ 1.000000e-01 0.000000e+00 0.000000e+00 1.153290e-15 1.153290e-17 7.096268e+02 7.096268e+02 5.095483e+08 8.474334e+00 2.851030e+00 1.664660e-10 2.075371e-13 7.722633e-14 4.447148e-14 5.587394e-12 5.853986e-16 3.863739e-17 3.872823e-16 3.844251e-13 7.300494e-05 1.000000e-01 3.500000e-02 3.500000e-01 5.024587e-34 5.024587e-36 3.548134e+03 3.548134e+03 2.219970e-10 -9.886505e+00 3.550000e+00 7.899202e-20 6.030650e-22 1.383617e-33 3.440258e-34 1.810430e-25 8.080346e-35 3.133545e-28 8.086328e-35 3.225259e-34 6.885677e-03 1.000000e-01 7.000000e-02 7.000000e-01 5.921268e-72 5.921268e-74 3.548134e+03 3.548134e+03 2.616143e-48 -4.781519e+01 3.550000e+00 7.899202e-20 6.030650e-22 1.383617e-33 3.440258e-34 1.810430e-25 8.080346e-35 3.133545e-28 8.086328e-35 3.225259e-34 6.885677e-03 7.071068e+00 0.000000e+00 0.000000e+00 2.285168e-13 2.285168e-15 6.820453e+01 6.820453e+01 1.009636e+11 1.077131e+01 1.833813e+00 6.431126e-05 2.138011e-18 3.018989e-14 2.846458e-14 4.012340e-12 4.361136e-15 3.347467e-15 9.598290e-12 3.676756e-18 3.737908e-11 7.071068e+00 2.474874e+00 3.500000e-01 2.716386e-17 2.716386e-19 3.410227e+02 3.410227e+02 1.200157e+07 6.846386e+00 2.532783e+00 3.196870e-15 9.588096e-18 7.029195e-25 5.756870e-26 2.279574e-19 9.920758e-30 2.743343e-21 2.316110e-28 1.619243e-28 2.640124e-04 7.071068e+00 4.949747e+00 7.000000e-01 7.189026e-23 7.189026e-25 3.410227e+02 3.410227e+02 3.176265e+01 1.269065e+00 2.532783e+00 7.899202e-20 6.030650e-22 1.383617e-33 3.440258e-34 1.810430e-25 8.080346e-35 3.133545e-28 8.086328e-35 3.225259e-34 6.885677e-03 5.000000e+02 0.000000e+00 0.000000e+00 2.871710e-20 2.871710e-22 6.555359e+00 6.555359e+00 1.268783e+04 3.870536e+00 8.165965e-01 7.246437e-09 1.639502e-09 9.789595e-12 6.854840e-12 1.985437e-12 3.611351e-12 1.609762e-16 2.563294e-12 2.442295e-10 4.478516e-05 5.000000e+02 1.750000e+02 3.500000e-01 6.929036e-22 6.929036e-24 2.625083e+01 2.625083e+01 3.061396e+02 2.253068e+00 1.419143e+00 9.592822e-12 3.264297e-14 2.652106e-14 1.248681e-14 8.773964e-14 4.373150e-18 6.765207e-18 1.123771e-21 1.185412e-18 7.771940e-05 5.000000e+02 3.500000e+02 7.000000e-01 8.170464e-23 8.170464e-25 3.277680e+01 3.277680e+01 3.609885e+01 1.324642e+00 1.515566e+00 3.185463e-12 1.816816e-16 2.245415e-18 1.057603e-18 5.366938e-14 2.360676e-22 1.899195e-20 9.770146e-23 1.644348e-23 8.516644e-05Expand source code
@dataclass class SciKitChemistry(ChemistryBase): """Predicts chemistry based on `sklearn` trained models Usage: >>> physics = WilliamsBest2014(radial_bins=3, vertical_bins=3) >>> chem = SciKitChemistry(physics) >>> chem._model # doctest: +NORMALIZE_WHITESPACE TransformedTargetRegressor(regressor=Pipeline(steps=[('quantiletransformer', QuantileTransformer(n_quantiles=100)), ('kneighborsregressor', KNeighborsRegressor(n_neighbors=62))]), transformer=QuantileTransformer(n_quantiles=100, output_distribution='normal')) >>> chem._model.X ('log(numberdens)', 'log(temperature)') >>> chem._model.y ('CO', 'CN', 'HCN', 'HNC', 'HCO+', 'H2CO', 'N2H+', 'CS', 'C2H', 'e-') >>> chem.run_chemistry() >>> chem.table # doctest: +NORMALIZE_WHITESPACE <CTable length=9> Radius Height Height to radius Gas density Dust density Gas temperature Dust temperature n(H+2H2) log(numberdens) log(temperature) CO CN HCN HNC HCO+ H2CO N2H+ CS C2H e- AU AU g / cm3 g / cm3 K K 1 / cm3 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 float64 ------------ ------------ ---------------- ------------ ------------ --------------- ---------------- ------------ --------------- ---------------- ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ 1.000000e-01 0.000000e+00 0.000000e+00 1.153290e-15 1.153290e-17 7.096268e+02 7.096268e+02 5.095483e+08 8.474334e+00 2.851030e+00 1.664660e-10 2.075371e-13 7.722633e-14 4.447148e-14 5.587394e-12 5.853986e-16 3.863739e-17 3.872823e-16 3.844251e-13 7.300494e-05 1.000000e-01 3.500000e-02 3.500000e-01 5.024587e-34 5.024587e-36 3.548134e+03 3.548134e+03 2.219970e-10 -9.886505e+00 3.550000e+00 7.899202e-20 6.030650e-22 1.383617e-33 3.440258e-34 1.810430e-25 8.080346e-35 3.133545e-28 8.086328e-35 3.225259e-34 6.885677e-03 1.000000e-01 7.000000e-02 7.000000e-01 5.921268e-72 5.921268e-74 3.548134e+03 3.548134e+03 2.616143e-48 -4.781519e+01 3.550000e+00 7.899202e-20 6.030650e-22 1.383617e-33 3.440258e-34 1.810430e-25 8.080346e-35 3.133545e-28 8.086328e-35 3.225259e-34 6.885677e-03 7.071068e+00 0.000000e+00 0.000000e+00 2.285168e-13 2.285168e-15 6.820453e+01 6.820453e+01 1.009636e+11 1.077131e+01 1.833813e+00 6.431126e-05 2.138011e-18 3.018989e-14 2.846458e-14 4.012340e-12 4.361136e-15 3.347467e-15 9.598290e-12 3.676756e-18 3.737908e-11 7.071068e+00 2.474874e+00 3.500000e-01 2.716386e-17 2.716386e-19 3.410227e+02 3.410227e+02 1.200157e+07 6.846386e+00 2.532783e+00 3.196870e-15 9.588096e-18 7.029195e-25 5.756870e-26 2.279574e-19 9.920758e-30 2.743343e-21 2.316110e-28 1.619243e-28 2.640124e-04 7.071068e+00 4.949747e+00 7.000000e-01 7.189026e-23 7.189026e-25 3.410227e+02 3.410227e+02 3.176265e+01 1.269065e+00 2.532783e+00 7.899202e-20 6.030650e-22 1.383617e-33 3.440258e-34 1.810430e-25 8.080346e-35 3.133545e-28 8.086328e-35 3.225259e-34 6.885677e-03 5.000000e+02 0.000000e+00 0.000000e+00 2.871710e-20 2.871710e-22 6.555359e+00 6.555359e+00 1.268783e+04 3.870536e+00 8.165965e-01 7.246437e-09 1.639502e-09 9.789595e-12 6.854840e-12 1.985437e-12 3.611351e-12 1.609762e-16 2.563294e-12 2.442295e-10 4.478516e-05 5.000000e+02 1.750000e+02 3.500000e-01 6.929036e-22 6.929036e-24 2.625083e+01 2.625083e+01 3.061396e+02 2.253068e+00 1.419143e+00 9.592822e-12 3.264297e-14 2.652106e-14 1.248681e-14 8.773964e-14 4.373150e-18 6.765207e-18 1.123771e-21 1.185412e-18 7.771940e-05 5.000000e+02 3.500000e+02 7.000000e-01 8.170464e-23 8.170464e-25 3.277680e+01 3.277680e+01 3.609885e+01 1.324642e+00 1.515566e+00 3.185463e-12 1.816816e-16 2.245415e-18 1.057603e-18 5.366938e-14 2.360676e-22 1.899195e-20 9.770146e-23 1.644348e-23 8.516644e-05 """ model: Union[PathLike, sklearn.base.TransformerMixin] = os.path.join( os.path.dirname(os.path.abspath(__file__)), "scikit_estimators", "andes2_atomic_knearest_temp_dens.pkl" ) def __post_init__(self): super().__post_init__() if hasattr(self.model, "predict"): self._model = self.model self.logger.info("SciKit Transformer object is recognized") else: self._model = self.load_scikit_model(self.model) def run_chemistry(self): all_arguments = copy.copy(self._model.X) existing_arguments_with_log = {f"log({colname})" for colname in self.table.colnames} for argument in all_arguments: if argument in self.table.colnames: pass elif argument in existing_arguments_with_log: self.table[f"log({argument})"] = np.log10( (self.table[argument]).cgs.value ) self.logger.info("log(%s) was taken in CGS system", argument) elif argument == "log(numberdens)": self.table["log(numberdens)"] = np.log10( (self.table["Gas density"] / self.molar_mass * c.N_A).to(u.cm ** -3).value ) elif argument == "log(temperature)": self.table["log(temperature)"] = np.log10(self.table["Gas temperature"].to(u.K).value) elif argument == "log(uv)": self.table["log(uv)"] = np.log10( (self.table["UV radiation strength"]).to_value(ANDES2_G0)) elif argument == "log(ionization)": self.table["log(ionization)"] = np.log10(self.table["Ionization rate"].to(1 / u.s).value) else: raise CHEFRuntimeError(f"{argument} or its precursor is not found in the original data") X = np.nan_to_num(np.array([self.table[col] for col in self._model.X]).T) y = self._model.predict(X) for species, abunds in zip(self._model.y, y.T): self.table[species] = 10 ** abunds @staticmethod def load_scikit_model(filename: PathLike) -> sklearn.base.RegressorMixin: """ Read pickled scikit regressor. Can be used as model argument to SciKitChemistry to avoid multiple loads Args: filename: path to file to be unpickled Returns: Regressor-like object with .predict(X), X, and y attributes Raises: FileNotFoundError if file is not found CHEFNotImplementedError / NotImplementedError if unpickled object does not have the needed attributes """ with open(filename, 'rb') as fff: model = pickle.load(fff) if hasattr(model, "predict") and hasattr(model, "X") and hasattr(model, 'y'): logging.info("SciKit Transformer object is read and recognized") else: raise CHEFNotImplementedError return modelAncestors
Class variables
var model : Union[str, os.PathLike, sklearn.base.TransformerMixin]
Static methods
def load_scikit_model(filename: Union[str, os.PathLike]) ‑> sklearn.base.RegressorMixin-
Read pickled scikit regressor. Can be used as model argument to SciKitChemistry to avoid multiple loads
Args
filename- path to file to be unpickled
Returns
Regressor-like object with .predict(X), X, and y attributes
Raises
FileNotFoundError if file is not found CHEFNotImplementedError / NotImplementedError if unpickled object does not have the needed attributes
Expand source code
@staticmethod def load_scikit_model(filename: PathLike) -> sklearn.base.RegressorMixin: """ Read pickled scikit regressor. Can be used as model argument to SciKitChemistry to avoid multiple loads Args: filename: path to file to be unpickled Returns: Regressor-like object with .predict(X), X, and y attributes Raises: FileNotFoundError if file is not found CHEFNotImplementedError / NotImplementedError if unpickled object does not have the needed attributes """ with open(filename, 'rb') as fff: model = pickle.load(fff) if hasattr(model, "predict") and hasattr(model, "X") and hasattr(model, 'y'): logging.info("SciKit Transformer object is read and recognized") else: raise CHEFNotImplementedError return model
Inherited members