diffct 2.0.1__tar.gz → 2.0.2__tar.gz

{diffct-2.0.1 → diffct-2.0.2}/CHANGELOG.md

@@ -5,6 +5,25 @@ All notable changes to this project are documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [2.0.2] - 2026-10-09
+
+### Fixed
+
+- The exit-time cleanup from 2.0.1 set numba-cuda's pending-deallocation
+  ratio to infinity. numba-cuda converts ``capacity * ratio`` to ``int``, so
+  each deallocation queued during shutdown raised ``ValueError`` or
+  ``OverflowError``. The exit code stayed 0, but stderr showed the error. The
+  exit hook now disables numba-cuda's pending-deallocation queues instead of
+  changing the limits, so no queue is flushed at exit and no error is raised.
+  The hook no longer imports ``numba.cuda.core.config``, which older
+  numba-cuda releases do not have.
+
+### Documentation
+
+- The install instructions state that ``pip install diffct`` without a CUDA
+  extra needs a system CUDA Toolkit. Use ``diffct[cu12]`` or ``diffct[cu13]``.
+- README images and links use absolute URLs, so they also work on PyPI.
+
 ## [2.0.1] - 2026-10-09
 
 ### Fixed
@@ -13,6 +32,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   weighted backprojections run in the CUDA context of their input. Before,
   inputs on another GPU than the current device could use the wrong context
   and stream.
+- A process that ran diffct kernels on several GPUs no longer crashes at
+  interpreter exit: numba-cuda's flush of pending CUDA module unloads during
+  shutdown is skipped, and process exit releases the modules.
 - ``custom_trajectory_3d`` and ``custom_trajectory_2d_fan`` raise
   ``ValueError`` for a source at the isocenter or a non-finite source,
   instead of returning NaN geometry. The figure-8 example in the

{diffct-2.0.1 → diffct-2.0.2}/CITATION.cff

@@ -13,7 +13,7 @@ authors:
     given-names: Yipeng
     email: yipeng.sun@fau.de
     alias: sypsyp97
-version: 2.0.1
+version: 2.0.2
 date-released: "2026-10-09"
 license: Apache-2.0
 repository-code: "https://github.com/sypsyp97/diffct"

{diffct-2.0.1 → diffct-2.0.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: diffct
-Version: 2.0.1
+Version: 2.0.2
 Summary: Differentiable CUDA projectors for CT: arbitrary trajectories, multi-GPU and multi-node execution, geometry gradients
 Project-URL: Homepage, https://github.com/sypsyp97/diffct
 Project-URL: Documentation, https://sypsyp97.github.io/diffct/
@@ -31,7 +31,7 @@ Description-Content-Type: text/markdown
 
 <h1 align="center">diffct</h1>
 
-<p align="center">English · <a href="README.zh.md">简体中文</a></p>
+<p align="center">English · <a href="https://github.com/sypsyp97/diffct/blob/main/README.zh.md">简体中文</a></p>
 
 <p align="center">
   Differentiable CUDA projectors for CT: arbitrary trajectories, multi-GPU, multi-node, geometry gradients.
@@ -43,15 +43,15 @@ Description-Content-Type: text/markdown
   <a href="https://pypi.org/project/diffct/"><img src="https://img.shields.io/pypi/v/diffct.svg?style=flat-square&logo=pypi&logoColor=white" alt="PyPI version"></a>
   <a href="https://sypsyp97.github.io/diffct/"><img src="https://img.shields.io/github/actions/workflow/status/sypsyp97/diffct/docs.yml?branch=main&label=docs&style=flat-square" alt="Documentation"></a>
   <a href="https://github.com/sypsyp97/diffct/actions"><img src="https://img.shields.io/github/actions/workflow/status/sypsyp97/diffct/ci.yml?branch=main&label=CI&style=flat-square" alt="CI/CD"></a>
-  <a href="https://deepwiki.com/sypsyp97/diffct"><img src="docs/assets/deepwiki-badge.svg" alt="Ask DeepWiki"></a>
+  <a href="https://deepwiki.com/sypsyp97/diffct"><img src="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/deepwiki-badge.svg" alt="Ask DeepWiki"></a>
 </p>
 
 <p align="center">
-  <img src="docs/assets/diffct_intro.gif" width="100%" alt="diffct intro: projection, sinogram, trajectories, multi-GPU split, reconstruction and geometry calibration">
+  <img src="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/diffct_intro.gif" width="100%" alt="diffct intro: projection, sinogram, trajectories, multi-GPU split, reconstruction and geometry calibration">
 </p>
 
 <p align="center">
-  <a href="docs/assets/diffct_intro.mp4">Intro video (MP4)</a> ·
+  <a href="https://github.com/sypsyp97/diffct/blob/main/docs/assets/diffct_intro.mp4">Intro video (MP4)</a> ·
   <a href="https://www.preprints.org/manuscript/202605.1446/v1">Technical report</a> ·
   <a href="https://doi.org/10.20944/preprints202605.1446.v1">DOI</a>
 </p>
@@ -65,7 +65,9 @@ Description-Content-Type: text/markdown
   <a href="#citation">Citation</a>
 </p>
 
-> **Version 2.0** replaces the circular-orbit API of 1.x with per-view trajectories. Code written for 1.x needs changes; see the [migration notes](docs/MIGRATION.md).
+> **Install** with the CUDA extra, `pip install "diffct[cu12]"` or `pip install "diffct[cu13]"`, after PyTorch. A plain `pip install diffct` needs a system CUDA Toolkit.
+>
+> **Version 2.0** replaces the circular-orbit API of 1.x with per-view trajectories. Code written for 1.x needs changes; see the [migration notes](https://github.com/sypsyp97/diffct/blob/main/docs/MIGRATION.md).
 > For Apple Silicon, use the [MLX port by Linda-Sophie Schneider](https://github.com/Linda-SophieSchneider/DiffCT-MLX).
 
 ## Why diffct
@@ -77,16 +79,19 @@ To our knowledge, diffct is the only open-source GPU CT library that combines ar
 - **Multi-GPU and multi-node.** Use `devices=[0, 1, 2, 3]` in one process, or one process per GPU with torchrun and NCCL. Views are partitioned and the volume is replicated. Speedup depends on workload and communication costs.
 - **Analytical helpers.** `diffct.analytical` provides ramp filters (ram-lak, shepp-logan, cosine, hamming, hann), fan, cone and Parker weights, and FBP and FDK backprojection.
 
-Capabilities, limits and isocenter rules: [docs/REFERENCE.md](docs/REFERENCE.md#capabilities-and-limits).
+Capabilities, limits and isocenter rules: [docs/REFERENCE.md](https://github.com/sypsyp97/diffct/blob/main/docs/REFERENCE.md#capabilities-and-limits).
 
 ## Install
 
-You need a CUDA GPU and PyTorch. Install PyTorch for your CUDA version first.
+You need an NVIDIA GPU. Install PyTorch for your CUDA version first. Then install diffct with the extra for the same CUDA major version:
 
 ```bash
-pip install "diffct[cu12]"    # [cu13] for CUDA 13; install PyTorch for your CUDA first
+pip install "diffct[cu12]"    # PyTorch built for CUDA 12
+pip install "diffct[cu13]"    # PyTorch built for CUDA 13
 ```
 
+The extra installs the CUDA compiler libraries that Numba CUDA needs to compile the kernels. A plain `pip install diffct` does not install them. Use it only if a CUDA Toolkit is already installed on the system.
+
 From source, with the examples:
 
 ```bash
@@ -129,11 +134,11 @@ loss = 0.5 * (A.project(x) - sinogram).square().sum()
 loss.backward()                     # x.grad = A^T (A x - y)
 ```
 
-**Custom or calibrated trajectories.** Pass calibrated tensors directly instead of a generator; no circular fit is needed. See [REFERENCE](docs/REFERENCE.md#custom-or-calibrated-trajectories).
+**Custom or calibrated trajectories.** Pass calibrated tensors directly instead of a generator; no circular fit is needed. See [REFERENCE](https://github.com/sypsyp97/diffct/blob/main/docs/REFERENCE.md#custom-or-calibrated-trajectories).
 
-**Geometry gradients.** Set `requires_grad=True` on trajectory tensors before you build `Projector`. See [REFERENCE](docs/REFERENCE.md#geometry-gradients).
+**Geometry gradients.** Set `requires_grad=True` on trajectory tensors before you build `Projector`. See [REFERENCE](https://github.com/sypsyp97/diffct/blob/main/docs/REFERENCE.md#geometry-gradients).
 
-**Several GPUs and nodes.** Use `devices=[0, 1, 2, 3]` in one process, or one process per GPU with torchrun for one or more nodes. See [REFERENCE](docs/REFERENCE.md#several-gpus-and-nodes) and [DISTRIBUTED.md](docs/DISTRIBUTED.md).
+**Several GPUs and nodes.** Use `devices=[0, 1, 2, 3]` in one process, or one process per GPU with torchrun for one or more nodes. See [REFERENCE](https://github.com/sypsyp97/diffct/blob/main/docs/REFERENCE.md#several-gpus-and-nodes) and [DISTRIBUTED.md](https://github.com/sypsyp97/diffct/blob/main/docs/DISTRIBUTED.md).
 
 ```python
 A = Projector(trajectory, (128, 128, 128), (384, 256), detector_spacing=0.8, devices=[0, 1, 2, 3])
@@ -143,8 +148,8 @@ A = Projector(trajectory, (128, 128, 128), (384, 256), detector_spacing=0.8, dev
 
 <p align="center">
   <picture>
-    <source media="(prefers-color-scheme: dark)" srcset="docs/assets/scaling_dark.png">
-    <img src="docs/assets/scaling_light.png" width="100%" alt="Time per forward, adjoint and CGLS iteration on 1, 4 and 8 GPUs, 64^3 and 128^3">
+    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/scaling_dark.png">
+    <img src="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/scaling_light.png" width="100%" alt="Time per forward, adjoint and CGLS iteration on 1, 4 and 8 GPUs, 64^3 and 128^3">
   </picture>
 </p>
 
@@ -154,7 +159,7 @@ A = Projector(trajectory, (128, 128, 128), (384, 256), detector_spacing=0.8, dev
 | Adjoint (backprojection) | 20.66 ms | 5.89 ms (3.5×) | 4.62 ms (4.5×) |
 | CGLS iteration | 30.08 ms | 8.08 ms (3.7×) | 5.59 ms (5.4×) |
 
-Circular trajectory, 360 views, detector (2n, 1.5n) cells at pitch 1.25, A100 64 GB GPUs. Small volumes scale less: at 64³, one CGLS iteration takes 4.63 ms on 1 GPU, 1.57 ms on 4 and 1.56 ms on 8. Raw data: [docs/assets/scaling.json](docs/assets/scaling.json).
+Circular trajectory, 360 views, detector (2n, 1.5n) cells at pitch 1.25, A100 64 GB GPUs. Small volumes scale less: at 64³, one CGLS iteration takes 4.63 ms on 1 GPU, 1.57 ms on 4 and 1.56 ms on 8. Raw data: [docs/assets/scaling.json](https://github.com/sypsyp97/diffct/blob/main/docs/assets/scaling.json).
 
 To measure one against several GPUs on your machine (default scan of the script, not the table setup):
 
@@ -167,7 +172,7 @@ python examples/benchmark_projector.py --devices 0 1
 **Measured walnut.** 240 measured views, circular cone beam, 256³: FDK (Hann window), SIRT (200 iterations), CGLS (20), TV (300, weight 0.3); axial and coronal centre slices.
 
 <p align="center">
-  <img src="docs/assets/walnut_measured.png" width="100%" alt="Measured walnut: FDK, SIRT, CGLS and TV reconstructions">
+  <img src="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/walnut_measured.png" width="100%" alt="Measured walnut: FDK, SIRT, CGLS and TV reconstructions">
 </p>
 
 ```bash
@@ -177,7 +182,7 @@ python examples/walnut_reconstruction.py --figure out.png
 **Simulated helical scan of the walnut.** The FDK walnut volume as ground truth, 720 views, 1% noise, 256³. PSNR: FDK 26.67 dB, CGLS (30 iterations) 34.13 dB, SIRT (200) 34.52 dB, TV (200, weight 1.0) 37.77 dB.
 
 <p align="center">
-  <img src="docs/assets/walnut_helical.png" width="100%" alt="Simulated helical scan of the walnut: FDK, CGLS, SIRT and TV reconstructions">
+  <img src="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/walnut_helical.png" width="100%" alt="Simulated helical scan of the walnut: FDK, CGLS, SIRT and TV reconstructions">
 </p>
 
 ```bash
@@ -185,22 +190,22 @@ python examples/walnut_reconstruction.py --algorithms --save-volume walnut256.np
 python examples/iterative_reconstruction.py --size 256 --views 720 --trajectory helical --noise 0.01 --phantom walnut256.npy --figure out.png
 ```
 
-Walnut data: Meaney 2022, Zenodo 6986012, CC BY 4.0; see [examples/data/NOTICE](examples/data/NOTICE).
+Walnut data: Meaney 2022, Zenodo 6986012, CC BY 4.0; see [examples/data/NOTICE](https://github.com/sypsyp97/diffct/blob/main/examples/data/NOTICE).
 
 ## Examples
 
-`quickstart.py`, `analytical_reconstruction.py`, `iterative_reconstruction.py`, `walnut_reconstruction.py`, `geometry_calibration.py`, `benchmark_projector.py`, `plot_trajectory.py`, and the Slurm template `slurm/multi_node.sbatch`. Launch modes, distributed-loss rules and measured results are in [examples/README.md](examples/README.md).
+`quickstart.py`, `analytical_reconstruction.py`, `iterative_reconstruction.py`, `walnut_reconstruction.py`, `geometry_calibration.py`, `benchmark_projector.py`, `plot_trajectory.py`, and the Slurm template `slurm/multi_node.sbatch`. Launch modes and distributed-loss rules are in [examples/README.md](https://github.com/sypsyp97/diffct/blob/main/examples/README.md).
 
 ## Documentation
 
 | Document | Contents |
 |---|---|
-| [Trajectory guide](docs/source/trajectories.rst) | Trajectory tuples and `Projector` usage |
-| [docs/REFERENCE.md](docs/REFERENCE.md) | Capabilities and limits, custom trajectories, geometry gradients, multi-GPU details, geometry and units |
-| [docs/DISTRIBUTED.md](docs/DISTRIBUTED.md) | Execution/memory choices, distributed loss rules, Slurm and cross-node checks |
-| [docs/MIGRATION.md](docs/MIGRATION.md) | Moving from the circular-only API |
-| [docs/video/README.md](docs/video/README.md) | How the intro video is rendered |
-| [CHANGELOG.md](CHANGELOG.md) | Release history |
+| [Trajectory guide](https://github.com/sypsyp97/diffct/blob/main/docs/source/trajectories.rst) | Trajectory tuples and `Projector` usage |
+| [docs/REFERENCE.md](https://github.com/sypsyp97/diffct/blob/main/docs/REFERENCE.md) | Capabilities and limits, custom trajectories, geometry gradients, multi-GPU details, geometry and units |
+| [docs/DISTRIBUTED.md](https://github.com/sypsyp97/diffct/blob/main/docs/DISTRIBUTED.md) | Execution/memory choices, distributed loss rules, Slurm and cross-node checks |
+| [docs/MIGRATION.md](https://github.com/sypsyp97/diffct/blob/main/docs/MIGRATION.md) | Moving from the circular-only API |
+| [docs/video/README.md](https://github.com/sypsyp97/diffct/blob/main/docs/video/README.md) | How the intro video is rendered |
+| [CHANGELOG.md](https://github.com/sypsyp97/diffct/blob/main/CHANGELOG.md) | Release history |
 
 Run the test suite on a CUDA host with `python -m pytest tests/ -q`.
 
@@ -232,7 +237,7 @@ If you use this library in your research, please cite the software and the techn
 
 ## License and acknowledgements
 
-Apache 2.0, see [LICENSE](LICENSE). The project draws on
+Apache 2.0, see [LICENSE](https://github.com/sypsyp97/diffct/blob/main/LICENSE). The project draws on
 [PYRO-NN](https://github.com/csyben/PYRO-NN) and
 [geometry_gradients_CT](https://github.com/mareikethies/geometry_gradients_CT).
-Issues and pull requests are welcome. The walnut data are CC BY 4.0 (Meaney 2022); see [NOTICE](NOTICE).
+Issues and pull requests are welcome. The walnut data are CC BY 4.0 (Meaney 2022); see [NOTICE](https://github.com/sypsyp97/diffct/blob/main/NOTICE).

{diffct-2.0.1 → diffct-2.0.2}/README.md

@@ -1,6 +1,6 @@
 <h1 align="center">diffct</h1>
 
-<p align="center">English · <a href="README.zh.md">简体中文</a></p>
+<p align="center">English · <a href="https://github.com/sypsyp97/diffct/blob/main/README.zh.md">简体中文</a></p>
 
 <p align="center">
   Differentiable CUDA projectors for CT: arbitrary trajectories, multi-GPU, multi-node, geometry gradients.
@@ -12,15 +12,15 @@
   <a href="https://pypi.org/project/diffct/"><img src="https://img.shields.io/pypi/v/diffct.svg?style=flat-square&logo=pypi&logoColor=white" alt="PyPI version"></a>
   <a href="https://sypsyp97.github.io/diffct/"><img src="https://img.shields.io/github/actions/workflow/status/sypsyp97/diffct/docs.yml?branch=main&label=docs&style=flat-square" alt="Documentation"></a>
   <a href="https://github.com/sypsyp97/diffct/actions"><img src="https://img.shields.io/github/actions/workflow/status/sypsyp97/diffct/ci.yml?branch=main&label=CI&style=flat-square" alt="CI/CD"></a>
-  <a href="https://deepwiki.com/sypsyp97/diffct"><img src="docs/assets/deepwiki-badge.svg" alt="Ask DeepWiki"></a>
+  <a href="https://deepwiki.com/sypsyp97/diffct"><img src="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/deepwiki-badge.svg" alt="Ask DeepWiki"></a>
 </p>
 
 <p align="center">
-  <img src="docs/assets/diffct_intro.gif" width="100%" alt="diffct intro: projection, sinogram, trajectories, multi-GPU split, reconstruction and geometry calibration">
+  <img src="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/diffct_intro.gif" width="100%" alt="diffct intro: projection, sinogram, trajectories, multi-GPU split, reconstruction and geometry calibration">
 </p>
 
 <p align="center">
-  <a href="docs/assets/diffct_intro.mp4">Intro video (MP4)</a> ·
+  <a href="https://github.com/sypsyp97/diffct/blob/main/docs/assets/diffct_intro.mp4">Intro video (MP4)</a> ·
   <a href="https://www.preprints.org/manuscript/202605.1446/v1">Technical report</a> ·
   <a href="https://doi.org/10.20944/preprints202605.1446.v1">DOI</a>
 </p>
@@ -34,7 +34,9 @@
   <a href="#citation">Citation</a>
 </p>
 
-> **Version 2.0** replaces the circular-orbit API of 1.x with per-view trajectories. Code written for 1.x needs changes; see the [migration notes](docs/MIGRATION.md).
+> **Install** with the CUDA extra, `pip install "diffct[cu12]"` or `pip install "diffct[cu13]"`, after PyTorch. A plain `pip install diffct` needs a system CUDA Toolkit.
+>
+> **Version 2.0** replaces the circular-orbit API of 1.x with per-view trajectories. Code written for 1.x needs changes; see the [migration notes](https://github.com/sypsyp97/diffct/blob/main/docs/MIGRATION.md).
 > For Apple Silicon, use the [MLX port by Linda-Sophie Schneider](https://github.com/Linda-SophieSchneider/DiffCT-MLX).
 
 ## Why diffct
@@ -46,16 +48,19 @@ To our knowledge, diffct is the only open-source GPU CT library that combines ar
 - **Multi-GPU and multi-node.** Use `devices=[0, 1, 2, 3]` in one process, or one process per GPU with torchrun and NCCL. Views are partitioned and the volume is replicated. Speedup depends on workload and communication costs.
 - **Analytical helpers.** `diffct.analytical` provides ramp filters (ram-lak, shepp-logan, cosine, hamming, hann), fan, cone and Parker weights, and FBP and FDK backprojection.
 
-Capabilities, limits and isocenter rules: [docs/REFERENCE.md](docs/REFERENCE.md#capabilities-and-limits).
+Capabilities, limits and isocenter rules: [docs/REFERENCE.md](https://github.com/sypsyp97/diffct/blob/main/docs/REFERENCE.md#capabilities-and-limits).
 
 ## Install
 
-You need a CUDA GPU and PyTorch. Install PyTorch for your CUDA version first.
+You need an NVIDIA GPU. Install PyTorch for your CUDA version first. Then install diffct with the extra for the same CUDA major version:
 
 ```bash
-pip install "diffct[cu12]"    # [cu13] for CUDA 13; install PyTorch for your CUDA first
+pip install "diffct[cu12]"    # PyTorch built for CUDA 12
+pip install "diffct[cu13]"    # PyTorch built for CUDA 13
 ```
 
+The extra installs the CUDA compiler libraries that Numba CUDA needs to compile the kernels. A plain `pip install diffct` does not install them. Use it only if a CUDA Toolkit is already installed on the system.
+
 From source, with the examples:
 
 ```bash
@@ -98,11 +103,11 @@ loss = 0.5 * (A.project(x) - sinogram).square().sum()
 loss.backward()                     # x.grad = A^T (A x - y)
 ```
 
-**Custom or calibrated trajectories.** Pass calibrated tensors directly instead of a generator; no circular fit is needed. See [REFERENCE](docs/REFERENCE.md#custom-or-calibrated-trajectories).
+**Custom or calibrated trajectories.** Pass calibrated tensors directly instead of a generator; no circular fit is needed. See [REFERENCE](https://github.com/sypsyp97/diffct/blob/main/docs/REFERENCE.md#custom-or-calibrated-trajectories).
 
-**Geometry gradients.** Set `requires_grad=True` on trajectory tensors before you build `Projector`. See [REFERENCE](docs/REFERENCE.md#geometry-gradients).
+**Geometry gradients.** Set `requires_grad=True` on trajectory tensors before you build `Projector`. See [REFERENCE](https://github.com/sypsyp97/diffct/blob/main/docs/REFERENCE.md#geometry-gradients).
 
-**Several GPUs and nodes.** Use `devices=[0, 1, 2, 3]` in one process, or one process per GPU with torchrun for one or more nodes. See [REFERENCE](docs/REFERENCE.md#several-gpus-and-nodes) and [DISTRIBUTED.md](docs/DISTRIBUTED.md).
+**Several GPUs and nodes.** Use `devices=[0, 1, 2, 3]` in one process, or one process per GPU with torchrun for one or more nodes. See [REFERENCE](https://github.com/sypsyp97/diffct/blob/main/docs/REFERENCE.md#several-gpus-and-nodes) and [DISTRIBUTED.md](https://github.com/sypsyp97/diffct/blob/main/docs/DISTRIBUTED.md).
 
 ```python
 A = Projector(trajectory, (128, 128, 128), (384, 256), detector_spacing=0.8, devices=[0, 1, 2, 3])
@@ -112,8 +117,8 @@ A = Projector(trajectory, (128, 128, 128), (384, 256), detector_spacing=0.8, dev
 
 <p align="center">
   <picture>
-    <source media="(prefers-color-scheme: dark)" srcset="docs/assets/scaling_dark.png">
-    <img src="docs/assets/scaling_light.png" width="100%" alt="Time per forward, adjoint and CGLS iteration on 1, 4 and 8 GPUs, 64^3 and 128^3">
+    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/scaling_dark.png">
+    <img src="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/scaling_light.png" width="100%" alt="Time per forward, adjoint and CGLS iteration on 1, 4 and 8 GPUs, 64^3 and 128^3">
   </picture>
 </p>
 
@@ -123,7 +128,7 @@ A = Projector(trajectory, (128, 128, 128), (384, 256), detector_spacing=0.8, dev
 | Adjoint (backprojection) | 20.66 ms | 5.89 ms (3.5×) | 4.62 ms (4.5×) |
 | CGLS iteration | 30.08 ms | 8.08 ms (3.7×) | 5.59 ms (5.4×) |
 
-Circular trajectory, 360 views, detector (2n, 1.5n) cells at pitch 1.25, A100 64 GB GPUs. Small volumes scale less: at 64³, one CGLS iteration takes 4.63 ms on 1 GPU, 1.57 ms on 4 and 1.56 ms on 8. Raw data: [docs/assets/scaling.json](docs/assets/scaling.json).
+Circular trajectory, 360 views, detector (2n, 1.5n) cells at pitch 1.25, A100 64 GB GPUs. Small volumes scale less: at 64³, one CGLS iteration takes 4.63 ms on 1 GPU, 1.57 ms on 4 and 1.56 ms on 8. Raw data: [docs/assets/scaling.json](https://github.com/sypsyp97/diffct/blob/main/docs/assets/scaling.json).
 
 To measure one against several GPUs on your machine (default scan of the script, not the table setup):
 
@@ -136,7 +141,7 @@ python examples/benchmark_projector.py --devices 0 1
 **Measured walnut.** 240 measured views, circular cone beam, 256³: FDK (Hann window), SIRT (200 iterations), CGLS (20), TV (300, weight 0.3); axial and coronal centre slices.
 
 <p align="center">
-  <img src="docs/assets/walnut_measured.png" width="100%" alt="Measured walnut: FDK, SIRT, CGLS and TV reconstructions">
+  <img src="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/walnut_measured.png" width="100%" alt="Measured walnut: FDK, SIRT, CGLS and TV reconstructions">
 </p>
 
 ```bash
@@ -146,7 +151,7 @@ python examples/walnut_reconstruction.py --figure out.png
 **Simulated helical scan of the walnut.** The FDK walnut volume as ground truth, 720 views, 1% noise, 256³. PSNR: FDK 26.67 dB, CGLS (30 iterations) 34.13 dB, SIRT (200) 34.52 dB, TV (200, weight 1.0) 37.77 dB.
 
 <p align="center">
-  <img src="docs/assets/walnut_helical.png" width="100%" alt="Simulated helical scan of the walnut: FDK, CGLS, SIRT and TV reconstructions">
+  <img src="https://raw.githubusercontent.com/sypsyp97/diffct/main/docs/assets/walnut_helical.png" width="100%" alt="Simulated helical scan of the walnut: FDK, CGLS, SIRT and TV reconstructions">
 </p>
 
 ```bash
@@ -154,22 +159,22 @@ python examples/walnut_reconstruction.py --algorithms --save-volume walnut256.np
 python examples/iterative_reconstruction.py --size 256 --views 720 --trajectory helical --noise 0.01 --phantom walnut256.npy --figure out.png
 ```
 
-Walnut data: Meaney 2022, Zenodo 6986012, CC BY 4.0; see [examples/data/NOTICE](examples/data/NOTICE).
+Walnut data: Meaney 2022, Zenodo 6986012, CC BY 4.0; see [examples/data/NOTICE](https://github.com/sypsyp97/diffct/blob/main/examples/data/NOTICE).
 
 ## Examples
 
-`quickstart.py`, `analytical_reconstruction.py`, `iterative_reconstruction.py`, `walnut_reconstruction.py`, `geometry_calibration.py`, `benchmark_projector.py`, `plot_trajectory.py`, and the Slurm template `slurm/multi_node.sbatch`. Launch modes, distributed-loss rules and measured results are in [examples/README.md](examples/README.md).
+`quickstart.py`, `analytical_reconstruction.py`, `iterative_reconstruction.py`, `walnut_reconstruction.py`, `geometry_calibration.py`, `benchmark_projector.py`, `plot_trajectory.py`, and the Slurm template `slurm/multi_node.sbatch`. Launch modes and distributed-loss rules are in [examples/README.md](https://github.com/sypsyp97/diffct/blob/main/examples/README.md).
 
 ## Documentation
 
 | Document | Contents |
 |---|---|
-| [Trajectory guide](docs/source/trajectories.rst) | Trajectory tuples and `Projector` usage |
-| [docs/REFERENCE.md](docs/REFERENCE.md) | Capabilities and limits, custom trajectories, geometry gradients, multi-GPU details, geometry and units |
-| [docs/DISTRIBUTED.md](docs/DISTRIBUTED.md) | Execution/memory choices, distributed loss rules, Slurm and cross-node checks |
-| [docs/MIGRATION.md](docs/MIGRATION.md) | Moving from the circular-only API |
-| [docs/video/README.md](docs/video/README.md) | How the intro video is rendered |
-| [CHANGELOG.md](CHANGELOG.md) | Release history |
+| [Trajectory guide](https://github.com/sypsyp97/diffct/blob/main/docs/source/trajectories.rst) | Trajectory tuples and `Projector` usage |
+| [docs/REFERENCE.md](https://github.com/sypsyp97/diffct/blob/main/docs/REFERENCE.md) | Capabilities and limits, custom trajectories, geometry gradients, multi-GPU details, geometry and units |
+| [docs/DISTRIBUTED.md](https://github.com/sypsyp97/diffct/blob/main/docs/DISTRIBUTED.md) | Execution/memory choices, distributed loss rules, Slurm and cross-node checks |
+| [docs/MIGRATION.md](https://github.com/sypsyp97/diffct/blob/main/docs/MIGRATION.md) | Moving from the circular-only API |
+| [docs/video/README.md](https://github.com/sypsyp97/diffct/blob/main/docs/video/README.md) | How the intro video is rendered |
+| [CHANGELOG.md](https://github.com/sypsyp97/diffct/blob/main/CHANGELOG.md) | Release history |
 
 Run the test suite on a CUDA host with `python -m pytest tests/ -q`.
 
@@ -201,7 +206,7 @@ If you use this library in your research, please cite the software and the techn
 
 ## License and acknowledgements
 
-Apache 2.0, see [LICENSE](LICENSE). The project draws on
+Apache 2.0, see [LICENSE](https://github.com/sypsyp97/diffct/blob/main/LICENSE). The project draws on
 [PYRO-NN](https://github.com/csyben/PYRO-NN) and
 [geometry_gradients_CT](https://github.com/mareikethies/geometry_gradients_CT).
-Issues and pull requests are welcome. The walnut data are CC BY 4.0 (Meaney 2022); see [NOTICE](NOTICE).
+Issues and pull requests are welcome. The walnut data are CC BY 4.0 (Meaney 2022); see [NOTICE](https://github.com/sypsyp97/diffct/blob/main/NOTICE).

{diffct-2.0.1 → diffct-2.0.2}/diffct/__init__.py

@@ -49,7 +49,7 @@ try:
 except ImportError:
     pass
 
-__version__ = '2.0.1'
+__version__ = '2.0.2'
 
 __all__ = [
     'Projector',

{diffct-2.0.1 → diffct-2.0.2}/diffct/utils.py

@@ -10,7 +10,6 @@ import contextlib
 import functools
 import inspect
 import math
-import sys
 import weakref
 import numpy as np
 import torch
@@ -191,16 +190,20 @@ def _on_device_of(name):
 def _keep_cuda_modules_at_exit():
     """Stop numba-cuda from unloading CUDA modules while the interpreter exits.
 
-    numba-cuda queues module unloads and flushes the queue once it holds more
-    than ``CUDA_DEALLOCS_COUNT`` items. After kernels were loaded on several
-    GPUs, a flush during interpreter shutdown can unload the same library twice
-    and corrupt the heap, so the process crashes after all work is done. The
-    process exit releases these resources anyway.
+    numba-cuda queues module unloads and other deallocations, and flushes a
+    queue once it passes a count or byte limit. After kernels were loaded on
+    several GPUs, a flush during interpreter shutdown can unload the same
+    library twice and corrupt the heap, so the process crashes after all work
+    is done. The process exit releases these resources anyway.
+
+    Raising the limits is not enough: the byte limit is
+    ``int(capacity * CUDA_DEALLOCS_RATIO)``, and queues with no capacity set
+    have a limit of 0 bytes for any finite ratio. numba-cuda skips the flush
+    while a queue is disabled, so disable every queue for the rest of the exit.
     """
-    from numba.cuda.core import config as cuda_config
+    from numba.cuda.cudadrv.driver import _PendingDeallocs
 
-    cuda_config.CUDA_DEALLOCS_COUNT = sys.maxsize
-    cuda_config.CUDA_DEALLOCS_RATIO = float("inf")
+    _PendingDeallocs.is_disabled = property(lambda self: True)
 
 
 class _ExitOrderAnchor:

diffct-2.0.2/docs/video/README.md

@@ -0,0 +1,47 @@
+# Intro video
+
+`diffct_intro.py` is the Manim Community (v0.21) source of `docs/assets/diffct_intro.gif`
+and `docs/assets/diffct_intro.mp4`. Rendering needs Manim, LaTeX and FFmpeg.
+The scenes read three files in `docs/video/`:
+
+- `walnut_measured.npz`: measured projections and reconstruction slices.
+- `data2d.npz`: a walnut slice and a display sinogram made with SciPy.
+- `calib_history.json`: saved results from a separate geometry calibration run.
+
+These inputs are included in the checkout. Rendering them does not need CUDA.
+The scenes do not read `recon_slices.npz` or `recon_psnr.json`.
+
+To rebuild the inputs, install SciPy in the environment that runs `make_inputs.py`.
+SciPy is not a core diffct dependency.
+Run these commands from the repository root:
+
+```bash
+python -m pip install scipy
+python docs/video/make_inputs.py --gpu
+```
+
+`--gpu` needs a CUDA-enabled diffct installation. It writes `walnut_measured.npz`,
+`recon_slices.npz`, `recon_psnr.json` and `data2d.npz`.
+Without `--gpu`, the script writes only `data2d.npz` and reads the existing `walnut_measured.npz`.
+Neither mode writes `calib_history.json`. The calibration example does not export this JSON file.
+
+Render and join the scenes:
+
+```bash
+cd docs/video
+rm -f scenes.txt
+for s in S00Intro S01Trajectory S02Autograd S03Geometry S04MultiGPU S05Measured S06End; do
+  manim -qh diffct_intro.py $s
+  echo "file 'media/videos/diffct_intro/1080p60/$s.mp4'" >> scenes.txt
+done
+ffmpeg -f concat -safe 0 -i scenes.txt -c:v libx264 -preset slow -crf 26 -pix_fmt yuv420p -r 30 \
+  -movflags +faststart -an ../assets/diffct_intro.mp4
+ffmpeg -i ../assets/diffct_intro.mp4 -vf "fps=8,scale=800:-1:flags=lanczos,palettegen=max_colors=64:stats_mode=full" palette.png
+ffmpeg -i ../assets/diffct_intro.mp4 -i palette.png \
+  -lavfi "fps=8,scale=800:-1:flags=lanczos[x];[x][1:v]paletteuse=dither=none" ../assets/diffct_intro.gif
+```
+
+The reconstructions come from the measured walnut in `examples/data/walnut_cone.npz`
+(Meaney 2022, CC BY 4.0; see `examples/data/NOTICE`). The timings in the multi-GPU scene
+are measured diffct results on A100 64 GB GPUs; see `docs/assets/scaling.json`.
+The scene stores these timings as constants. Rebuilding the inputs does not update them.

{diffct-2.0.1 → diffct-2.0.2}/examples/README.md

@@ -21,10 +21,12 @@ python examples/iterative_reconstruction.py --size 32 --views 32 \
     --trajectory helical --algorithms cgls --iterations 5
 ```
 
-This still computes an FDK baseline; five CGLS steps are a usage example, not
-the reconstruction-quality configuration in the results below. The default
-iterative run uses a 128³ volume, 360 views, and CGLS/SIRT/TV with 30/200/200
-iterations respectively.
+This still computes an FDK baseline. Five CGLS steps show how to use the script.
+The default iterative run uses a 128³ volume, 360 views, and CGLS/SIRT/TV with 30/200/200
+iterations respectively. Its TV weight is 1.0.
+The measured walnut script uses SIRT/CGLS/TV with 200/20/300 iterations and TV weight 0.3.
+These weights are example settings. Choose the TV weight for your geometry,
+voxel spacing, resolution and data scale.
 
 To inspect a helical trajectory without a GPU (PyTorch and matplotlib required):
 

{diffct-2.0.1 → diffct-2.0.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "diffct"
-version = "2.0.1"
+version = "2.0.2"
 description = "Differentiable CUDA projectors for CT: arbitrary trajectories, multi-GPU and multi-node execution, geometry gradients"
 readme = "README.md"
 requires-python = ">=3.10"

diffct-2.0.2/tests/test_cuda_exit_deallocations.py

@@ -0,0 +1,83 @@
+"""CPU regression checks for CUDA deallocation queues during shutdown."""
+
+import logging
+
+import pytest
+
+
+@pytest.mark.parametrize("capacity", [None, 80 * 1024**3], ids=["unset", "80-GiB"])
+@pytest.mark.parametrize("with_size", [False, True], ids=["unsized", "sized"])
+def test_exit_hook_keeps_cuda_deallocations_pending(monkeypatch, capacity, with_size):
+    try:
+        from numba.cuda.cudadrv import driver
+        from numba.cuda.cudadrv.driver import _PendingDeallocs
+    except ImportError as error:
+        pytest.skip(f"CUDA deallocation queue cannot be imported: {error}")
+
+    from numba.core import config
+    from diffct.utils import _keep_cuda_modules_at_exit
+
+    # Queue logging normally starts when a CUDA driver is initialized.
+    monkeypatch.setattr(driver, "_logger", logging.getLogger(__name__), raising=False)
+    # Capture all shared state before the hook can replace or add attributes.
+    class_attributes = dict(vars(_PendingDeallocs))
+    configurations = {id(configuration): configuration for configuration in (config, driver.config)}
+    configuration_attributes = {
+        identifier: {
+            name: value
+            for name, value in vars(configuration).items()
+            if name.startswith("CUDA_DEALLOCS_")
+        }
+        for identifier, configuration in configurations.items()
+    }
+
+    queue = _PendingDeallocs() if capacity is None else _PendingDeallocs(capacity)
+    destroyed = []
+
+    def destructor(handle):
+        destroyed.append(handle)
+
+    try:
+        _keep_cuda_modules_at_exit()
+
+        # Sizes are bookkeeping values; this test does not allocate CUDA memory.
+        size = (80 * 1024**3) // 32
+        for handle in range(64):
+            if with_size:
+                queue.add_item(destructor, handle, size)
+            else:
+                queue.add_item(destructor, handle)
+            assert destroyed == []
+            assert len(queue) == handle + 1
+    finally:
+        for name in set(vars(_PendingDeallocs)) - class_attributes.keys():
+            delattr(_PendingDeallocs, name)
+        for name, value in class_attributes.items():
+            if name not in vars(_PendingDeallocs) or vars(_PendingDeallocs)[name] is not value:
+                setattr(_PendingDeallocs, name, value)
+        for identifier, configuration in configurations.items():
+            saved = configuration_attributes[identifier]
+            for name in tuple(vars(configuration)):
+                if name.startswith("CUDA_DEALLOCS_") and name not in saved:
+                    delattr(configuration, name)
+            for name, value in saved.items():
+                setattr(configuration, name, value)
+
+    assert dict(vars(_PendingDeallocs)) == class_attributes
+    for identifier, configuration in configurations.items():
+        assert {
+            name: value
+            for name, value in vars(configuration).items()
+            if name.startswith("CUDA_DEALLOCS_")
+        } == configuration_attributes[identifier]
+
+    # A fresh queue must regain the stock count-triggered flush after restoration.
+    stock_queue = _PendingDeallocs()
+    stock_destroyed = []
+    count_limit = driver.config.CUDA_DEALLOCS_COUNT
+    for handle in range(count_limit):
+        stock_queue.add_item(stock_destroyed.append, handle)
+        assert stock_destroyed == []
+    stock_queue.add_item(stock_destroyed.append, count_limit)
+    assert stock_destroyed == list(range(count_limit + 1))
+    assert len(stock_queue) == 0

{diffct-2.0.1 → diffct-2.0.2}/tests/test_multi_gpu_exit.py

@@ -81,3 +81,8 @@ def test_multi_gpu_exit():
         f"CUDA subprocess exited with return code {child.returncode}\n"
         f"Child stderr tail:\n{child.stderr[-8000:]}"
     )
+    for marker in ("Exception ignored", "Traceback", "ValueError", "OverflowError"):
+        assert marker not in child.stderr, (
+            f"CUDA subprocess reported {marker!r} during cleanup\n"
+            f"Child stderr tail:\n{child.stderr[-8000:]}"
+        )

diffct-2.0.1/docs/video/README.md

@@ -1,23 +0,0 @@
-# Intro video
-
-`diffct_intro.py` is the Manim Community (v0.21) source of `docs/assets/diffct_intro.gif`
-and `docs/assets/diffct_intro.mp4`. Make the inputs on a CUDA GPU, then render and join the scenes:
-
-```bash
-python docs/video/make_inputs.py --gpu      # walnut reconstructions; without --gpu only data2d.npz
-cd docs/video
-rm -f scenes.txt
-for s in S00Intro S01Trajectory S02Autograd S03Geometry S04MultiGPU S05Measured S06End; do
-  manim -qh diffct_intro.py $s
-  echo "file 'media/videos/diffct_intro/1080p60/$s.mp4'" >> scenes.txt
-done
-ffmpeg -f concat -safe 0 -i scenes.txt -c:v libx264 -preset slow -crf 26 -pix_fmt yuv420p -r 30 \
-  -movflags +faststart -an ../assets/diffct_intro.mp4
-ffmpeg -i ../assets/diffct_intro.mp4 -vf "fps=8,scale=800:-1:flags=lanczos,palettegen=max_colors=64:stats_mode=full" palette.png
-ffmpeg -i ../assets/diffct_intro.mp4 -i palette.png \
-  -lavfi "fps=8,scale=800:-1:flags=lanczos[x];[x][1:v]paletteuse=dither=none" ../assets/diffct_intro.gif
-```
-
-The reconstructions come from the measured walnut in `examples/data/walnut_cone.npz`
-(Meaney 2022, CC BY 4.0; see `examples/data/NOTICE`). The timings in the multi-GPU scene
-are measured diffct results on A100 64 GB GPUs; see `docs/assets/scaling.json`.