Merge remote-tracking branch 'origin/main' into 467-update-docs

Author: sgreenbury

autoemulate/experimental/data/preprocessors.py

@@ -9,7 +9,6 @@
 from torch.distributions import TransformedDistribution
 from torch.optim.lr_scheduler import ExponentialLR, LRScheduler
 
-from autoemulate.experimental.data.preprocessors import Preprocessor
 from autoemulate.experimental.data.utils import ConversionMixin, ValidationMixin
 from autoemulate.experimental.device import TorchDeviceMixin
 from autoemulate.experimental.transforms.standardize import StandardizeTransform
@@ -392,7 +391,7 @@ def predict(self, x: TensorLike, with_grad: bool = False) -> GaussianLike:
         return pred
 
 
-class PyTorchBackend(nn.Module, Emulator, Preprocessor):
+class PyTorchBackend(nn.Module, Emulator):
     """
     `PyTorchBackend` provides a backend for PyTorch models.
 
@@ -410,19 +409,13 @@ class PyTorchBackend(nn.Module, Emulator, Preprocessor):
     epochs: int = 10
     loss_history: ClassVar[list[float]] = []
     verbose: bool = False
-    preprocessor: Preprocessor | None = None
     loss_fn: nn.Module = nn.MSELoss()
     optimizer_cls: type[optim.Optimizer] = optim.Adam
     optimizer: optim.Optimizer
     supports_grad: bool = True
     lr: float = 1e-1
     scheduler_cls: type[LRScheduler] | None = None
 
-    def preprocess(self, x: TensorLike) -> TensorLike:  # noqa: D102
-        if self.preprocessor is None:
-            return x
-        return self.preprocessor.preprocess(x)
-
     def loss_func(self, y_pred, y_true):
         """Loss function to be used for training the model."""
         return nn.MSELoss()(y_pred, y_true)
@@ -455,9 +448,6 @@ def _fit(
             batches = 0
 
             for x_batch, y_batch in dataloader:
-                # Preprocess x_batch
-                x = self.preprocess(x_batch)
-
                 # Forward pass
                 y_pred = self.forward(x_batch)
                 loss = self.loss_func(y_pred, y_batch)
@@ -535,7 +525,6 @@ def _initialize_weights(
     def _predict(self, x: TensorLike, with_grad: bool) -> OutputLike:
         self.eval()
         with torch.set_grad_enabled(with_grad):
-            x = self.preprocess(x)
             return self(x)
 
 

autoemulate/experimental/emulators/base.py

@@ -268,10 +268,7 @@ def _predict(self, x: Tensor, with_grad: bool) -> GaussianLike:
         samples = []
         with torch.set_grad_enabled(with_grad):
             for _ in range(self.n_samples):
-                # apply any preprocessing the model expects
-                x_proc = self.model.preprocess(x)
-
-                out = self.model.forward(x_proc)
+                out = self.model.forward(x)
                 # out: Tensor of shape (batch_size, output_dim)
                 samples.append(out)
 

autoemulate/experimental/emulators/ensemble.py

@@ -89,7 +89,6 @@ def _predict(self, x: TensorLike, with_grad: bool) -> OutputLike:
             msg = "Gradient calculation is not supported."
             raise ValueError(msg)
         self.eval()
-        x = self.preprocess(x)
         return self(x)
 
     @staticmethod

autoemulate/experimental/emulators/radial_basis_functions.py

@@ -1,7 +1,6 @@
 import numpy as np
 import pytest
 import torch
-from autoemulate.experimental.data.preprocessors import Standardizer
 from autoemulate.experimental.data.utils import set_random_seed
 from autoemulate.experimental.device import TorchDeviceMixin
 from autoemulate.experimental.emulators.base import PyTorchBackend
@@ -32,9 +31,6 @@ def __init__(self, x=None, y=None, random_seed=None, device=None, **kwargs):
             self.scheduler_setup(kwargs)
             self.epochs = kwargs.get("epochs", 10)
             self.batch_size = kwargs.get("batch_size", 16)
-            self.preprocessor = Standardizer(
-                torch.Tensor([[-0.5]]), torch.Tensor([[0.5]])
-            )
 
         def forward(self, x):
             return self.linear(x)
@@ -110,24 +106,6 @@ def test_tune_xy(self):
         tuner = Tuner(x_train, y_train, n_iter=10)
         tuner.run(self.DummyModel)
 
-    def test_standardizer(self):
-        x_train = torch.Tensor(
-            [[1.0, 1.0], [2.0, 2.0], [3.0, 3.0], [4.0, 4.0], [5.0, 5.0]]
-        )
-        x_train_preprocessed = self.model.preprocess(x_train)
-        assert isinstance(x_train_preprocessed, torch.Tensor)
-        assert torch.allclose(
-            x_train_preprocessed,
-            torch.Tensor([[3.0], [5.0], [7.0], [9.0], [11.0]]),
-        )
-
-    def test_standardizer_fail(self):
-        x_train = torch.Tensor([0.1, 2.0, 6.0, 0.2])
-        with pytest.raises(
-            ValueError, match="Expected 2D TensorLike, actual shape dim 1"
-        ):
-            self.model.preprocess(x_train)
-
     def test_tune_dataset(self):
         """
         Test that Tuner accepts a single Dataset input.