Merge branch 'master' into master

CHANGELOG.md

@@ -31,7 +31,8 @@
 - Fix bug where `dtype` was not used when creating a `ParameterBeam` from Twiss parameters (see #206) (@jank324)
 - Fix bug after running `Segment.inactive_elements_as_drifts` the drifts could have the wrong `dtype` (see #206) (@jank324)
 - Fix an issue where splitting elements would result in splits with a different `dtype` (see #211) (@jank324)
-- Pin to NumPy 1.X due to compatibility issues on Windows (see #220) (@hespe)
+- Fix issue in Bmad import where collimators had no length by interpreting them as `Drift` + `Aperture` (see #249) (@jank324)
+- Fix NumPy 2 compatibility issues with PyTorch on Windows (see #220, #242) (@hespe)
 
 ### 🐆 Other
 
@@ -45,6 +46,7 @@
 - Fix the quadrupole strength units in the quadrupole docstring (see #202) (@ansantam)
 - Add CI runs for macOS (arm64) and Windows (see #226) (@cr-xu, @jank324, @hespe)
 - Clean up CI pipelines (see #243, #244) (@jank324)
+- Fix logo display in README (see #252) (@jank324)
 
 ## [v0.6.3](https://github.com/desy-ml/cheetah/releases/tag/v0.6.3) (2024-03-28)
 

README.md

@@ -1,5 +1,3 @@
-<img src="images/logo.png" align="right" width="25%"/>
-
 ![format](https://github.com/desy-ml/cheetah/actions/workflows/format.yaml/badge.svg)
 ![pytest](https://github.com/desy-ml/cheetah/actions/workflows/pytest.yaml/badge.svg)
 [![Documentation Status](https://readthedocs.org/projects/cheetah-accelerator/badge/?version=latest)](https://cheetah-accelerator.readthedocs.io/en/latest/?badge=latest)
@@ -10,7 +8,13 @@
 
 # Cheetah
 
-Cheetah is a particle tracking accelerator we built specifically to speed up the training of reinforcement learning models.
+<img src="images/logo.png" align="right" width="25%"/>
+
+Cheetah is a high-speed differentiable beam dynamics code specifically design to support machine learning applications for particle accelerators.
+
+Its speed helps generate data fast, for example for extremely data-hungry reinforcement learning algorithms, while its differentiability can be used for a variety of applications, including accelerator tuning, system identification and physics-informed prior means for Bayesian optimisation. Its native integration with machine learning toolchains around PyTorch also makes Cheetah an ideal candidate for coupling of physics-based and neural network beam dynamics models that remain fast and differentiable.
+
+To learn more about what Cheetah can do, we recommend reading our [PRAB paper](https://doi.org/10.1103/PhysRevAccelBeams.27.054601). To learn how to use Cheetah, we refer to the example notebooks in the [Cheetah documentation](https://cheetah-accelerator.readthedocs.io/).
 
 ## Installation
 

cheetah/converters/bmad.py

@@ -216,26 +216,47 @@ def convert_element(
                 ["element_type", "l", "alias", "type", "x_limit", "y_limit"],
                 bmad_parsed,
             )
-            return cheetah.Aperture(
-                x_max=torch.tensor(bmad_parsed.get("x_limit", np.inf)),
-                y_max=torch.tensor(bmad_parsed.get("y_limit", np.inf)),
-                shape="rectangular",
+            return cheetah.Segment(
+                elements=[
+                    cheetah.Drift(
+                        length=torch.tensor(bmad_parsed.get("l", 0.0)),
+                        name=name + "_drift",
+                        device=device,
+                        dtype=dtype,
+                    ),
+                    cheetah.Aperture(
+                        x_max=torch.tensor(bmad_parsed.get("x_limit", np.inf)),
+                        y_max=torch.tensor(bmad_parsed.get("y_limit", np.inf)),
+                        shape="rectangular",
+                        name=name + "_aperture",
+                        device=device,
+                        dtype=dtype,
+                    ),
+                ],
                 name=name,
-                device=device,
-                dtype=dtype,
             )
         elif bmad_parsed["element_type"] == "ecollimator":
             validate_understood_properties(
                 ["element_type", "l", "alias", "type", "x_limit", "y_limit"],
                 bmad_parsed,
             )
-            return cheetah.Aperture(
-                x_max=torch.tensor(bmad_parsed.get("x_limit", np.inf)),
-                y_max=torch.tensor(bmad_parsed.get("y_limit", np.inf)),
-                shape="elliptical",
-                name=name,
-                device=device,
-                dtype=dtype,
+            return cheetah.Segment(
+                elements=[
+                    cheetah.Drift(
+                        length=torch.tensor(bmad_parsed.get("l", 0.0)),
+                        name=name + "_drift",
+                        device=device,
+                        dtype=dtype,
+                    ),
+                    cheetah.Aperture(
+                        x_max=torch.tensor(bmad_parsed.get("x_limit", np.inf)),
+                        y_max=torch.tensor(bmad_parsed.get("y_limit", np.inf)),
+                        shape="elliptical",
+                        name=name + "_aperture",
+                        device=device,
+                        dtype=dtype,
+                    ),
+                ],
             )
         elif bmad_parsed["element_type"] == "wiggler":
             validate_understood_properties(

setup.py

@@ -19,5 +19,5 @@
     long_description_content_type="text/markdown",
     packages=[package for package in find_packages() if package.startswith("cheetah")],
     python_requires=">=3.9",
-    install_requires=["matplotlib", "numpy<2.0", "scipy", "torch"],
+    install_requires=["matplotlib", "numpy", "scipy", "torch"],
 )

tests/test_bmad_conversion.py

@@ -13,18 +13,19 @@ def test_bmad_tutorial():
 
     correct = cheetah.Segment(
         [
-            cheetah.Drift(length=torch.tensor(0.5), name="d"),
+            cheetah.Drift(length=torch.tensor([0.5]), name="d"),
             cheetah.Dipole(
-                length=torch.tensor(0.5), e1=torch.tensor(0.1), name="b"
+                length=torch.tensor([0.5]), e1=torch.tensor([0.1]), name="b"
             ),  # TODO: What are g and dg?
             cheetah.Quadrupole(
-                length=torch.tensor(0.6), k1=torch.tensor(0.23), name="q"
+                length=torch.tensor([0.6]), k1=torch.tensor([0.23]), name="q"
             ),
         ],
         name="bmad_tutorial",
     )
 
     assert converted.name == correct.name
+    assert converted.length == correct.length
     assert [element.name for element in converted.elements] == [
         element.name for element in correct.elements
     ]