DebugSlicer
A VS Code extension for visually inspecting PyTorch tensors, NumPy arrays, PIL images and MONAI MetaTensors within the debugger. It supports linked views and chunked loading of large arrays.
Timeline (last 90 days)
DebugSlicer is a VS Code extension that lets developers right-click on a tensor directly in the debugger and scroll through the data in three linked views.
DebugSlicer supports PyTorch tensors, MONAI MetaTensor, NumPy arrays, and PIL images.
When handling PyTorch tensors, DebugSlicer automatically moves them to the CPU, with no need to handle device issues manually.
DebugSlicer loads arrays of several GB in pieces to support debugging of large medical imaging or remote sensing data.
DebugSlicer is published on the VS Code Marketplace under the extension name CH4LLENG3R.debugslicer-python.
DebugSlicer's source code is hosted in the CH4LLENG3R/debugslicer-python repository on GitHub.
DebugSlicer is currently maintained by the author individually.
DebugSlicer's limitations include: it supports only the several data structure types mentioned above and does not cover TensorFlow or other frameworks; the interaction depth of its linked views is not on par with professional medical imaging software; and as an individually maintained open-source project, its long-term update cadence remains to be seen.
DebugSlicer is positioned as "a quick look at the data at a breakpoint" rather than a replacement for full visualization analysis tools, and its advantage lies in zero configuration and zero scripts, embedding directly into the existing VS Code debugging workflow.
All Facts (9)
DebugSlicer is a VS Code extension that lets developers right-click on a tensor directly in the debugger and scroll through the data in three linked views.
50%UnverifiedDebugSlicer supports PyTorch tensors, MONAI MetaTensor, NumPy arrays, and PIL images.
50%UnverifiedWhen handling PyTorch tensors, DebugSlicer automatically moves them to the CPU, with no need to handle device issues manually.
50%UnverifiedDebugSlicer loads arrays of several GB in pieces to support debugging of large medical imaging or remote sensing data.
50%UnverifiedDebugSlicer is positioned as "a quick look at the data at a breakpoint" rather than a replacement for full visualization analysis tools, and its advantage lies in zero configuration and zero scripts, embedding directly into the existing VS Code debugging workflow.
50%UnverifiedDebugSlicer's source code is hosted in the CH4LLENG3R/debugslicer-python repository on GitHub.
50%UnverifiedDebugSlicer is currently maintained by the author individually.
50%UnverifiedDebugSlicer's limitations include: it supports only the several data structure types mentioned above and does not cover TensorFlow or other frameworks; the interaction depth of its linked views is not on par with professional medical imaging software; and as an individually maintained open-source project, its long-term update cadence remains to be seen.
50%UnverifiedDebugSlicer is published on the VS Code Marketplace under the extension name CH4LLENG3R.debugslicer-python.
50%