Conversation
Bump actions/checkout v2 -> v7, actions/setup-dotnet v1 -> v6 and peaceiris/actions-gh-pages v3 -> v4. The old versions run on Node 12/16, which the runner force-migrates with a deprecation warning on every run. Add workflow_dispatch to the PR workflows so a branch can be built without opening a pull request. Mark build_macos continue-on-error: TorchSharp 0.100.3 resolves libtorch-cpu 2.0.1.1, which has no osx-arm64 native package, and macos-latest runs on arm64.
NUnit3TestAdapter 3.17.0 hangs and crashes the test host under the VSTest 18 runner shipped with the .NET 10 SDK. Update to the newest versions that still support net6.0: NUnit3TestAdapter 5.2.0, Microsoft.NET.Test.Sdk 17.13.0, NUnit 3.14.0 and coverlet.msbuild 6.0.4.
All projects now target net10.0 (previously netstandard2.1 and net6.0). Add global.json pinning SDK 10.0.100 with latestFeature roll-forward, and move CI, the Dockerfile and doc script references to .NET 10. Remove the CI steps that moved global.json aside. Update FSharp.Core 6.0.3 -> 10.0.100, the version shipped with the .NET 10 SDK. BinaryFormatter was removed from the runtime in .NET 9, which breaks save, load and Model.clone. Reference System.Runtime.Serialization.Formatters to restore it; applications must also set EnableUnsafeBinaryFormatterSerialization to true, as the test project now does. Files written by previous versions still load. Update NUnit3TestAdapter 6.3.0, Microsoft.NET.Test.Sdk 18.10.1 and coverlet.msbuild 10.0.1, and fsdocs-tool 15.0.0 -> 22.2.0, which cannot load projects with the .NET 10 SDK. Annotate the timeout in Plot.runScript as int: Process.WaitForExit gained a TimeSpan overload, making the call ambiguous.
1.0.1 has known high and moderate severity vulnerabilities. 2.1.13 is the latest release licensed under Apache-2.0; 3.0 and later use the Six Labors Split License. No API changes were needed.
torch.tensor now takes dimensions as ReadOnlySpan<int64> and has additional IList/Memory overloads; pass dimensions as a span to keep overload resolution unambiguous. avg_pool2d/avg_pool3d parameters strides/paddings are renamed stride/padding. max_unpool2d/max_unpool3d changed to PyTorch's (kernel_size, stride, padding, output_size) parameter order and validate output_size against kernel_size, which the unpool op does not receive. Implement max-unpooling as a scatter into a zero tensor of the output shape instead, for 1D, 2D and 3D. libtorch-cpu 2.10.0 includes an osx-arm64 native package, so build_macos no longer needs continue-on-error. The SkiaSharp dependency moves to 2.88.6, which has no known vulnerabilities.
libtorch now ships osx-arm64 natives. Forcing x64 prevents the test host and consuming applications from running on arm64, for example on Apple Silicon. DiffSharp.Benchmarks.Python keeps x64 because it binds a Windows x64 Python runtime.
Download to a temporary file and extract into a staging directory, moving results into place only when complete. Previously an interrupted download or extraction left partial data that later runs treated as complete. Replace the obsolete WebClient with HttpClient, and the hand-written tar reader with System.Formats.Tar.
Remove the Coveralls upload, its .NET 5 install, the coveralls.net tool, codecov.yml and the coverage badge. Stop enabling GPU tests by default on one specific machine; set DIFFSHARP_TESTGPU instead. Update the DEVGUIDE GPU instructions and remove the obsolete macOS arm64 section. Update the TorchSharp version used by the example scripts to 0.107.0. Add the unmanaged constraint required by Tensor.data<T>. Remove .DS_Store.
libtorch-cpu-osx-arm64 links libomp by absolute Homebrew path (/opt/homebrew/opt/libomp/lib/libomp.dylib) rather than using the copy shipped in the package, so libtorch fails to load unless Homebrew's libomp is installed.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.