Merge branch 'main' into main

Author: esantorella

botorch/acquisition/input_constructors.py

@@ -843,6 +843,7 @@ def construct_inputs_qNEHVI(
         "cache_pending": kwargs.get("cache_pending", True),
         "max_iep": kwargs.get("max_iep", 0),
         "incremental_nehvi": kwargs.get("incremental_nehvi", True),
+        "cache_root": kwargs.get("cache_root", True),
     }
 
 

botorch/acquisition/multi_objective/objective.py

@@ -82,7 +82,6 @@ def __init__(
         r"""Initialize Objective.
 
         Args:
-            weights: `m'`-dim tensor of outcome weights.
             outcomes: A list of the `m'` indices that the weights should be
                 applied to.
             num_outcomes: The total number of outcomes `m`

botorch/exceptions/warnings.py

@@ -49,3 +49,9 @@ class BotorchTensorDimensionWarning(BotorchWarning):
     r"""Warning raised when a tensor possibly violates a botorch convention."""
 
     pass
+
+
+class UserInputWarning(BotorchWarning):
+    r"""Warning raised when a potential issue is detected with user provided inputs."""
+
+    pass

botorch/models/transforms/input.py

@@ -22,6 +22,7 @@
 
 import torch
 from botorch.exceptions.errors import BotorchTensorDimensionError
+from botorch.exceptions.warnings import UserInputWarning
 from botorch.models.transforms.utils import subset_transform
 from botorch.models.utils import fantasize
 from botorch.utils.rounding import approximate_round, OneHotArgmaxSTE, RoundSTE
@@ -505,6 +506,7 @@ def __init__(
         transform_on_fantasize: bool = True,
         reverse: bool = False,
         min_range: float = 1e-8,
+        learn_bounds: Optional[bool] = None,
     ) -> None:
         r"""Normalize the inputs to the unit cube.
 
@@ -527,7 +529,13 @@ def __init__(
                 the inputs.
             min_range: Amount of noise to add to the range to ensure no division by
                 zero errors.
+            learn_bounds: Whether to learn the bounds in train mode. Defaults
+                to False if bounds are provided, otherwise defaults to True.
         """
+        if learn_bounds is not None:
+            self.learn_coefficients = learn_bounds
+        else:
+            self.learn_coefficients = bounds is None
         transform_dimension = d if indices is None else len(indices)
         if bounds is not None:
             if indices is not None and bounds.size(-1) == d:
@@ -544,7 +552,12 @@ def __init__(
         else:
             coefficient = torch.ones(*batch_shape, 1, transform_dimension)
             offset = torch.zeros(*batch_shape, 1, transform_dimension)
-            self.learn_coefficients = True
+            if self.learn_coefficients is False:
+                warn(
+                    "learn_bounds is False and no bounds were provided. The bounds "
+                    "will not be updated and the transform will be a no-op.",
+                    UserInputWarning,
+                )
         super().__init__(
             d=d,
             coefficient=coefficient,
@@ -586,6 +599,21 @@ def _update_coefficients(self, X) -> None:
         self._coefficient = torch.amax(X, dim=reduce_dims).unsqueeze(-2) - self.offset
         self._coefficient.clamp_(min=self.min_range)
 
+    def get_init_args(self) -> Dict[str, Any]:
+        r"""Get the arguments necessary to construct an exact copy of the transform."""
+        return {
+            "d": self._d,
+            "indices": getattr(self, "indices", None),
+            "bounds": self.bounds,
+            "batch_shape": self.batch_shape,
+            "transform_on_train": self.transform_on_train,
+            "transform_on_eval": self.transform_on_eval,
+            "transform_on_fantasize": self.transform_on_fantasize,
+            "reverse": self.reverse,
+            "min_range": self.min_range,
+            "learn_bounds": self.learn_bounds,
+        }
+
 
 class InputStandardize(AffineInputTransform):
     r"""Standardize inputs (zero mean, unit variance).
@@ -796,6 +824,18 @@ def equals(self, other: InputTransform) -> bool:
             and self.tau == other.tau
         )
 
+    def get_init_args(self) -> Dict[str, Any]:
+        r"""Get the arguments necessary to construct an exact copy of the transform."""
+        return {
+            "integer_indices": self.integer_indices,
+            "categorical_features": self.categorical_features,
+            "transform_on_train": self.transform_on_train,
+            "transform_on_eval": self.transform_on_eval,
+            "transform_on_fantasize": self.transform_on_fantasize,
+            "approximate": self.approximate,
+            "tau": self.tau,
+        }
+
 
 class Log10(ReversibleInputTransform, Module):
     r"""A base-10 log transformation."""

scripts/run_tutorials.py

@@ -10,16 +10,13 @@
 import datetime
 import os
 import subprocess
-import tempfile
 import time
 from pathlib import Path
 from subprocess import CalledProcessError
 from typing import Any, Dict, Optional, Tuple
 
-import nbformat
 import pandas as pd
 from memory_profiler import memory_usage
-from nbconvert import PythonExporter
 
 
 IGNORE_ALWAYS = {  # ignored in smoke tests and full runs
@@ -30,7 +27,6 @@
 RUN_IF_SMOKE_TEST_IGNORE_IF_STANDARD = {  # only used in smoke tests
     "thompson_sampling.ipynb",  # very slow without KeOps + GPU
     "composite_mtbo.ipynb",  # TODO: very slow, figure out if we can make it faster
-    "Multi_objective_multi_fidelity_BO.ipynb",  # TODO: very slow, speed up
     # Causing the tutorials to crash when run without smoke test. Likely OOM.
     # Fix planned.
     "constraint_active_search.ipynb",
@@ -65,33 +61,18 @@ def get_output_file_path(smoke_test: bool) -> str:
     return fname
 
 
-def parse_ipynb(file: Path) -> str:
-    with open(file, "r") as nb_file:
-        nb_str = nb_file.read()
-    nb = nbformat.reads(nb_str, nbformat.NO_CONVERT)
-    exporter = PythonExporter()
-    script, _ = exporter.from_notebook_node(nb)
-    return script
-
-
 def run_script(
-    script: str, timeout_minutes: int, env: Optional[Dict[str, str]] = None
+    tutorial: Path, timeout_minutes: int, env: Optional[Dict[str, str]] = None
 ) -> None:
-    # need to keep the file around & close it so subprocess does not run into I/O issues
-    with tempfile.NamedTemporaryFile(delete=False) as tf:
-        tf_name = tf.name
-        with open(tf_name, "w") as tmp_script:
-            tmp_script.write(script)
     if env is not None:
         env = {**os.environ, **env}
     run_out = subprocess.run(
-        ["ipython", tf_name],
+        ["papermill", tutorial, "|"],
         capture_output=True,
         text=True,
         env=env,
         timeout=timeout_minutes * 60,
     )
-    os.remove(tf_name)
     return run_out
 
 
@@ -103,13 +84,12 @@ def run_tutorial(
     them as a string, and returns runtime and memory information as a dict.
     """
     timeout_minutes = 5 if smoke_test else 30
-    script = parse_ipynb(tutorial)
     tic = time.monotonic()
     print(f"Running tutorial {tutorial.name}.")
     env = {"SMOKE_TEST": "True"} if smoke_test else None
     try:
         mem_usage, run_out = memory_usage(
-            (run_script, (script, timeout_minutes), {"env": env}),
+            (run_script, (tutorial, timeout_minutes), {"env": env}),
             retval=True,
             include_children=True,
         )

setup.py

@@ -23,6 +23,7 @@
     "kaleido",
     "matplotlib",
     "memory_profiler",
+    "papermill",
     "pykeops",
     "torchvision",
 ]

test/acquisition/test_input_constructors.py

@@ -715,6 +715,7 @@ def test_construct_inputs_qNEHVI(self):
         self.assertTrue(kwargs["cache_pending"])
         self.assertEqual(kwargs["max_iep"], 0)
         self.assertTrue(kwargs["incremental_nehvi"])
+        self.assertTrue(kwargs["cache_root"])
 
         # Test check for block designs
         mock_model = mock.Mock()
@@ -748,6 +749,7 @@ def test_construct_inputs_qNEHVI(self):
             cache_pending=False,
             max_iep=1,
             incremental_nehvi=False,
+            cache_root=False,
         )
         ref_point_expected = objective(objective_thresholds)
         self.assertTrue(torch.equal(kwargs["ref_point"], ref_point_expected))
@@ -768,6 +770,7 @@ def test_construct_inputs_qNEHVI(self):
         self.assertFalse(kwargs["cache_pending"])
         self.assertEqual(kwargs["max_iep"], 1)
         self.assertFalse(kwargs["incremental_nehvi"])
+        self.assertFalse(kwargs["cache_root"])
 
         # Test with risk measures.
         with self.assertRaisesRegex(UnsupportedError, "feasibility-weighted"):

test/exceptions/test_warnings.py

@@ -15,6 +15,7 @@
     InputDataWarning,
     OptimizationWarning,
     SamplingWarning,
+    UserInputWarning,
 )
 from botorch.utils.testing import BotorchTestCase
 
@@ -28,6 +29,7 @@ def test_botorch_warnings_hierarchy(self):
         self.assertIsInstance(OptimizationWarning(), BotorchWarning)
         self.assertIsInstance(SamplingWarning(), BotorchWarning)
         self.assertIsInstance(BotorchTensorDimensionWarning(), BotorchWarning)
+        self.assertIsInstance(UserInputWarning(), BotorchWarning)
 
     def test_botorch_warnings(self):
         for WarningClass in (
@@ -38,6 +40,7 @@ def test_botorch_warnings(self):
             InputDataWarning,
             OptimizationWarning,
             SamplingWarning,
+            UserInputWarning,
         ):
             with warnings.catch_warnings(record=True) as ws, settings.debug(True):
                 warnings.warn("message", WarningClass)

test/models/test_deterministic.py

@@ -96,7 +96,7 @@ def test_AffineDeterministicModel(self):
             X = torch.rand(*shape)
             p = model.posterior(X)
             mean_exp = model.b + (X.unsqueeze(-1) * a).sum(dim=-2)
-            self.assertTrue(torch.equal(p.mean, mean_exp))
+            self.assertAllClose(p.mean, mean_exp)
         # # test two-dim output
         a = torch.rand(3, 2)
         model = AffineDeterministicModel(a)
@@ -105,7 +105,7 @@ def test_AffineDeterministicModel(self):
             X = torch.rand(*shape)
             p = model.posterior(X)
             mean_exp = model.b + (X.unsqueeze(-1) * a).sum(dim=-2)
-            self.assertTrue(torch.equal(p.mean, mean_exp))
+            self.assertAllClose(p.mean, mean_exp)
         # test subset output
         X = torch.rand(4, 3)
         subset_model = model.subset_output([0])

test/models/transforms/test_input.py

@@ -11,6 +11,7 @@
 import torch
 from botorch import settings
 from botorch.exceptions.errors import BotorchTensorDimensionError
+from botorch.exceptions.warnings import UserInputWarning
 from botorch.models.transforms.input import (
     AffineInputTransform,
     AppendFeatures,
@@ -155,16 +156,29 @@ def test_normalize(self):
             self.assertEqual(nlz._d, 2)
             self.assertEqual(nlz.mins.shape, torch.Size([3, 1, 2]))
             self.assertEqual(nlz.ranges.shape, torch.Size([3, 1, 2]))
+            self.assertTrue(nlz.equals(Normalize(**nlz.get_init_args())))
 
-            # basic init, fixed bounds
+            # learn_bounds=False with no bounds.
+            with self.assertWarnsRegex(UserInputWarning, "learn_bounds"):
+                Normalize(d=2, learn_bounds=False)
+
+            # learn_bounds=True with bounds provided.
             bounds = torch.zeros(2, 2, device=self.device, dtype=dtype)
+            nlz = Normalize(d=2, bounds=bounds, learn_bounds=True)
+            self.assertTrue(nlz.learn_bounds)
+            self.assertTrue(torch.equal(nlz.mins, bounds[..., 0:1, :]))
+            self.assertTrue(
+                torch.equal(nlz.ranges, bounds[..., 1:2, :] - bounds[..., 0:1, :])
+            )
+
+            # basic init, fixed bounds
             nlz = Normalize(d=2, bounds=bounds)
             self.assertFalse(nlz.learn_bounds)
             self.assertTrue(nlz.training)
             self.assertEqual(nlz._d, 2)
             self.assertTrue(torch.equal(nlz.mins, bounds[..., 0:1, :]))
             self.assertTrue(
-                torch.equal(nlz.mins, bounds[..., 1:2, :] - bounds[..., 0:1, :])
+                torch.equal(nlz.ranges, bounds[..., 1:2, :] - bounds[..., 0:1, :])
             )
             # with grad
             bounds.requires_grad = True
@@ -180,6 +194,7 @@ def test_normalize(self):
             nlz.eval()
             self.assertIsNone(nlz.coefficient.grad_fn)
             self.assertIsNone(nlz.offset.grad_fn)
+            self.assertTrue(nlz.equals(Normalize(**nlz.get_init_args())))
 
             # basic init, provided indices
             with self.assertRaises(ValueError):
@@ -204,6 +219,7 @@ def test_normalize(self):
                     == torch.tensor([0], dtype=torch.long, device=self.device)
                 ).all()
             )
+            self.assertTrue(nlz.equals(Normalize(**nlz.get_init_args())))
 
             # test .to
             other_dtype = torch.float if dtype == torch.double else torch.double
@@ -594,13 +610,15 @@ def test_round_transform(self):
             self.assertTrue(round_tf.training)
             self.assertFalse(round_tf.approximate)
             self.assertEqual(round_tf.tau, 1e-3)
+            self.assertTrue(round_tf.equals(Round(**round_tf.get_init_args())))
 
             # With tensor indices.
             round_tf = Round(
                 integer_indices=torch.tensor(int_idcs, dtype=dtype, device=self.device),
                 categorical_features=categorical_feats,
             )
             self.assertEqual(round_tf.integer_indices.tolist(), int_idcs)
+            self.assertTrue(round_tf.equals(Round(**round_tf.get_init_args())))
 
             # basic usage
             for batch_shape, approx, categorical_features in itertools.product(