Add protocol for estimator builder

sbi/inference/trainers/base.py

@@ -7,7 +7,16 @@
 from dataclasses import asdict
 from datetime import datetime
 from pathlib import Path
-from typing import Any, Callable, Dict, List, Literal, Optional, Tuple, Union
+from typing import (
+    Any,
+    Callable,
+    Dict,
+    List,
+    Literal,
+    Optional,
+    Tuple,
+    Union,
+)
 from warnings import warn
 
 import torch

sbi/inference/trainers/nle/mnle.py

@@ -1,7 +1,7 @@
 # This file is part of sbi, a toolkit for simulation-based inference. sbi is licensed
 # under the Apache License Version 2.0, see <https://www.apache.org/licenses/>
 
-from typing import Any, Callable, Dict, Literal, Optional, Union
+from typing import Any, Dict, Literal, Optional, Union
 
 from torch.distributions import Distribution
 
@@ -14,6 +14,7 @@
 )
 from sbi.inference.trainers.nle.nle_base import LikelihoodEstimatorTrainer
 from sbi.neural_nets.estimators import MixedDensityEstimator
+from sbi.neural_nets.estimators.base import DensityEstimatorBuilder
 from sbi.sbi_types import TensorBoardSummaryWriter
 from sbi.utils.sbiutils import del_entries
 
@@ -33,7 +34,7 @@ class MNLE(LikelihoodEstimatorTrainer):
     def __init__(
         self,
         prior: Optional[Distribution] = None,
-        density_estimator: Union[str, Callable] = "mnle",
+        density_estimator: Union[Literal["mnle"], DensityEstimatorBuilder] = "mnle",
         device: str = "cpu",
         logging_level: Union[int, str] = "WARNING",
         summary_writer: Optional[TensorBoardSummaryWriter] = None,
@@ -47,12 +48,12 @@ def __init__(
                 prior must be passed to `.build_posterior()`.
             density_estimator: If it is a string, it must be "mnle" to use the
                 preconfiugred neural nets for MNLE. Alternatively, a function
-                that builds a custom neural network can be provided. The function will
+                that builds a custom neural network, which adheres to
+                `DensityEstimatorBuilder` protocol can be provided. The function will
                 be called with the first batch of simulations (theta, x), which can
-                thus be used for shape inference and potentially for z-scoring. It
-                needs to return a PyTorch `nn.Module` implementing the density
-                estimator. The density estimator needs to provide the methods
-                `.log_prob`, `.log_prob_iid()` and `.sample()`.
+                thus be used for shape inference and potentially for z-scoring. The
+                density estimator needs to provide the methods `.log_prob` and
+                `.sample()`.
             device: Training device, e.g., "cpu", "cuda" or "cuda:{0, 1, ...}".
             logging_level: Minimum severity of messages to log. One of the strings
                 INFO, WARNING, DEBUG, ERROR and CRITICAL.

sbi/inference/trainers/nle/nle_a.py

@@ -1,11 +1,12 @@
 # This file is part of sbi, a toolkit for simulation-based inference. sbi is licensed
 # under the Apache License Version 2.0, see <https://www.apache.org/licenses/>
 
-from typing import Callable, Optional, Union
+from typing import Literal, Optional, Union
 
 from torch.distributions import Distribution
 
 from sbi.inference.trainers.nle.nle_base import LikelihoodEstimatorTrainer
+from sbi.neural_nets.estimators.base import DensityEstimatorBuilder
 from sbi.sbi_types import TensorBoardSummaryWriter
 from sbi.utils.sbiutils import del_entries
 
@@ -21,7 +22,9 @@ class NLE_A(LikelihoodEstimatorTrainer):
     def __init__(
         self,
         prior: Optional[Distribution] = None,
-        density_estimator: Union[str, Callable] = "maf",
+        density_estimator: Union[
+            Literal["nsf", "maf", "mdn", "made"], DensityEstimatorBuilder
+        ] = "maf",
         device: str = "cpu",
         logging_level: Union[int, str] = "WARNING",
         summary_writer: Optional[TensorBoardSummaryWriter] = None,
@@ -35,12 +38,12 @@ def __init__(
                 prior must be passed to `.build_posterior()`.
             density_estimator: If it is a string, use a pre-configured network of the
                 provided type (one of nsf, maf, mdn, made). Alternatively, a function
-                that builds a custom neural network can be provided. The function will
+                that builds a custom neural network, which adheres to
+                `DensityEstimatorBuilder` protocol can be provided. The function will
                 be called with the first batch of simulations (theta, x), which can
-                thus be used for shape inference and potentially for z-scoring. It
-                needs to return a PyTorch `nn.Module` implementing the density
-                estimator. The density estimator needs to provide the methods
-                `.log_prob` and `.sample()`.
+                thus be used for shape inference and potentially for z-scoring. The
+                density estimator needs to provide the methods `.log_prob` and
+                `.sample()`.
             device: Training device, e.g., "cpu", "cuda" or "cuda:{0, 1, ...}".
             logging_level: Minimum severity of messages to log. One of the strings
                 INFO, WARNING, DEBUG, ERROR and CRITICAL.

sbi/inference/trainers/nle/nle_base.py

@@ -4,7 +4,7 @@
 import warnings
 from abc import ABC
 from copy import deepcopy
-from typing import Any, Callable, Dict, Literal, Optional, Tuple, Union
+from typing import Any, Dict, Literal, Optional, Tuple, Union
 
 import torch
 from torch import Tensor
@@ -26,6 +26,7 @@
 from sbi.inference.trainers.base import NeuralInference
 from sbi.neural_nets import likelihood_nn
 from sbi.neural_nets.estimators import ConditionalDensityEstimator
+from sbi.neural_nets.estimators.base import DensityEstimatorBuilder
 from sbi.neural_nets.estimators.shape_handling import (
     reshape_to_batch_event,
 )
@@ -38,7 +39,9 @@ class LikelihoodEstimatorTrainer(NeuralInference, ABC):
     def __init__(
         self,
         prior: Optional[Distribution] = None,
-        density_estimator: Union[str, Callable] = "maf",
+        density_estimator: Union[
+            Literal["nsf", "maf", "mdn", "made"], DensityEstimatorBuilder
+        ] = "maf",
         device: str = "cpu",
         logging_level: Union[int, str] = "WARNING",
         summary_writer: Optional[SummaryWriter] = None,
@@ -53,12 +56,12 @@ def __init__(
                 distribution) can be used.
             density_estimator: If it is a string, use a pre-configured network of the
                 provided type (one of nsf, maf, mdn, made). Alternatively, a function
-                that builds a custom neural network can be provided. The function will
+                that builds a custom neural network, which adheres to
+                `DensityEstimatorBuilder` protocol can be provided. The function will
                 be called with the first batch of simulations (theta, x), which can
-                thus be used for shape inference and potentially for z-scoring. It
-                needs to return a PyTorch `nn.Module` implementing the density
-                estimator. The density estimator needs to provide the methods
-                `.log_prob` and `.sample()`.
+                thus be used for shape inference and potentially for z-scoring. The
+                density estimator needs to provide the methods `.log_prob` and
+                `.sample()`.
 
         See docstring of `NeuralInference` class for all other arguments.
         """

sbi/inference/trainers/npe/mnpe.py

@@ -1,7 +1,7 @@
 # This file is part of sbi, a toolkit for simulation-based inference. sbi is licensed
 # under the Apache License Version 2.0, see <https://www.apache.org/licenses/>
 
-from typing import Any, Callable, Dict, Literal, Optional, Union
+from typing import Any, Dict, Literal, Optional, Union
 
 from torch.distributions import Distribution
 
@@ -15,6 +15,7 @@
 )
 from sbi.inference.trainers.npe.npe_c import NPE_C
 from sbi.neural_nets.estimators import MixedDensityEstimator
+from sbi.neural_nets.estimators.base import DensityEstimatorBuilder
 from sbi.sbi_types import TensorBoardSummaryWriter
 from sbi.utils.sbiutils import del_entries
 
@@ -31,7 +32,7 @@ class MNPE(NPE_C):
     def __init__(
         self,
         prior: Optional[Distribution] = None,
-        density_estimator: Union[str, Callable] = "mnpe",
+        density_estimator: Union[Literal["mnpe"], DensityEstimatorBuilder] = "mnpe",
         device: str = "cpu",
         logging_level: Union[int, str] = "WARNING",
         summary_writer: Optional[TensorBoardSummaryWriter] = None,
@@ -45,12 +46,12 @@ def __init__(
                 prior must be passed to `.build_posterior()`.
             density_estimator: If it is a string, it must be "mnpe" to use the
                 preconfigured neural nets for MNPE. Alternatively, a function
-                that builds a custom neural network can be provided. The function will
+                that builds a custom neural network, which adheres to
+                `DensityEstimatorBuilder` protocol can be provided. The function will
                 be called with the first batch of simulations (theta, x), which can
-                thus be used for shape inference and potentially for z-scoring. It
-                needs to return a PyTorch `nn.Module` implementing the density
-                estimator. The density estimator needs to provide the methods
-                `.log_prob`, `.log_prob_iid()` and `.sample()`.
+                thus be used for shape inference and potentially for z-scoring. The
+                density estimator needs to provide the methods `.log_prob` and
+                `.sample()`.
             device: Training device, e.g., "cpu", "cuda" or "cuda:{0, 1, ...}".
             logging_level: Minimum severity of messages to log. One of the strings
                 INFO, WARNING, DEBUG, ERROR and CRITICAL.

sbi/inference/trainers/npe/npe_a.py

@@ -4,7 +4,7 @@
 import warnings
 from copy import deepcopy
 from functools import partial
-from typing import Any, Callable, Dict, Optional, Union
+from typing import Any, Callable, Dict, Literal, Optional, Union
 
 import torch
 from pyknos.mdn.mdn import MultivariateGaussianMDN
@@ -16,7 +16,10 @@
 from sbi.inference.trainers.npe.npe_base import (
     PosteriorEstimatorTrainer,
 )
-from sbi.neural_nets.estimators.base import ConditionalDensityEstimator
+from sbi.neural_nets.estimators.base import (
+    ConditionalDensityEstimator,
+    DensityEstimatorBuilder,
+)
 from sbi.sbi_types import TensorBoardSummaryWriter
 from sbi.utils import torchutils
 from sbi.utils.sbiutils import (
@@ -49,7 +52,9 @@ class NPE_A(PosteriorEstimatorTrainer):
     def __init__(
         self,
         prior: Optional[Distribution] = None,
-        density_estimator: Union[str, Callable] = "mdn_snpe_a",
+        density_estimator: Union[
+            Literal["mdn_snpe_a"], DensityEstimatorBuilder
+        ] = "mdn_snpe_a",
         num_components: int = 10,
         device: str = "cpu",
         logging_level: Union[int, str] = "WARNING",
@@ -65,17 +70,17 @@ def __init__(
                 distribution) can be used.
             density_estimator: If it is a string (only "mdn_snpe_a" is valid), use a
                 pre-configured mixture of densities network. Alternatively, a function
-                that builds a custom neural network can be provided. The function will
+                that builds a custom neural network, which adheres to
+                `DensityEstimatorBuilder` protocol can be provided. The function will
                 be called with the first batch of simulations (theta, x), which can
-                thus be used for shape inference and potentially for z-scoring. It
-                needs to return a PyTorch `nn.Module` implementing the density
-                estimator. The density estimator needs to provide the methods
-                `.log_prob` and `.sample()`. Note that until the last round only a
-                single (multivariate) Gaussian component is used for training (see
-                Algorithm 1 in [1]). In the last round, this component is replicated
-                `num_components` times, its parameters are perturbed with a very small
-                noise, and then the last training round is done with the expanded
-                Gaussian mixture as estimator for the proposal posterior.
+                thus be used for shape inference and potentially for z-scoring. The
+                density estimator needs to provide the methods `.log_prob` and
+                `.sample()`. Note that until the last round only a single (multivariate)
+                Gaussian component is used for training (seeAlgorithm 1 in [1]). In the
+                last round, this component is replicated `num_components` times, its
+                parameters are perturbed with a very small noise, and then the last
+                training round is done with the expanded Gaussian mixture as estimator
+                for the proposal posterior.
             num_components: Number of components of the mixture of Gaussians in the
                 last round. This overrides the `num_components` value passed to
                 `posterior_nn()`.

sbi/inference/trainers/npe/npe_b.py

@@ -1,14 +1,17 @@
 # This file is part of sbi, a toolkit for simulation-based inference. sbi is licensed
 # under the Apache License Version 2.0, see <https://www.apache.org/licenses/>
 
-from typing import Any, Callable, Optional, Union
+from typing import Any, Literal, Optional, Union
 
 import torch
 from torch import Tensor
 from torch.distributions import Distribution
 
 import sbi.utils as utils
-from sbi.inference.trainers.npe.npe_base import PosteriorEstimatorTrainer
+from sbi.inference.trainers.npe.npe_base import (
+    PosteriorEstimatorTrainer,
+)
+from sbi.neural_nets.estimators.base import DensityEstimatorBuilder
 from sbi.neural_nets.estimators.shape_handling import reshape_to_sample_batch_event
 from sbi.sbi_types import TensorBoardSummaryWriter
 from sbi.utils.sbiutils import del_entries
@@ -33,7 +36,9 @@ class NPE_B(PosteriorEstimatorTrainer):
     def __init__(
         self,
         prior: Optional[Distribution] = None,
-        density_estimator: Union[str, Callable] = "maf",
+        density_estimator: Union[
+            Literal["nsf", "maf", "mdn", "made"], DensityEstimatorBuilder
+        ] = "maf",
         device: str = "cpu",
         logging_level: Union[int, str] = "WARNING",
         summary_writer: Optional[TensorBoardSummaryWriter] = None,
@@ -46,12 +51,12 @@ def __init__(
                 parameters, e.g. which ranges are meaningful for them.
             density_estimator: If it is a string, use a pre-configured network of the
                 provided type (one of nsf, maf, mdn, made). Alternatively, a function
-                that builds a custom neural network can be provided. The function will
+                that builds a custom neural network, which adheres to
+                `DensityEstimatorBuilder` protocol can be provided. The function will
                 be called with the first batch of simulations (theta, x), which can
-                thus be used for shape inference and potentially for z-scoring. It
-                needs to return a PyTorch `nn.Module` implementing the density
-                estimator. The density estimator needs to provide the methods
-                `.log_prob` and `.sample()`.
+                thus be used for shape inference and potentially for z-scoring. The
+                density estimator needs to provide the methods `.log_prob` and
+                `.sample()`.
             device: Training device, e.g., "cpu", "cuda" or "cuda:{0, 1, ...}".
             logging_level: Minimum severity of messages to log. One of the strings
                 INFO, WARNING, DEBUG, ERROR and CRITICAL.

sbi/inference/trainers/npe/npe_base.py

@@ -28,9 +28,13 @@
 )
 from sbi.inference.potentials import posterior_estimator_based_potential
 from sbi.inference.potentials.posterior_based_potential import PosteriorBasedPotential
-from sbi.inference.trainers.base import NeuralInference, check_if_proposal_has_default_x
+from sbi.inference.trainers.base import (
+    NeuralInference,
+    check_if_proposal_has_default_x,
+)
 from sbi.neural_nets import posterior_nn
 from sbi.neural_nets.estimators import ConditionalDensityEstimator
+from sbi.neural_nets.estimators.base import DensityEstimatorBuilder
 from sbi.neural_nets.estimators.shape_handling import (
     reshape_to_batch_event,
     reshape_to_sample_batch_event,
@@ -54,7 +58,9 @@ class PosteriorEstimatorTrainer(NeuralInference, ABC):
     def __init__(
         self,
         prior: Optional[Distribution] = None,
-        density_estimator: Union[str, Callable] = "maf",
+        density_estimator: Union[
+            Literal["nsf", "maf", "mdn", "made"], DensityEstimatorBuilder
+        ] = "maf",
         device: str = "cpu",
         logging_level: Union[int, str] = "WARNING",
         summary_writer: Optional[SummaryWriter] = None,
@@ -69,12 +75,12 @@ def __init__(
         Args:
             density_estimator: If it is a string, use a pre-configured network of the
                 provided type (one of nsf, maf, mdn, made). Alternatively, a function
-                that builds a custom neural network can be provided. The function will
+                that builds a custom neural network, which adheres to
+                `DensityEstimatorBuilder` protocol can be provided. The function will
                 be called with the first batch of simulations (theta, x), which can
-                thus be used for shape inference and potentially for z-scoring. It
-                needs to return a PyTorch `nn.Module` implementing the density
-                estimator. The density estimator needs to provide the methods
-                `.log_prob` and `.sample()`.
+                thus be used for shape inference and potentially for z-scoring. The
+                density estimator needs to provide the methods `.log_prob` and
+                `.sample()`.
 
         See docstring of `NeuralInference` class for all other arguments.
         """