TorchSharp
TorchSharp copied to clipboard
A .NET library that provides access to the library that powers PyTorch.
Improves debugging of native library loading exceptions by printing any unhandled exceptions to the trace log. Fixes #1310
It seems extremely error-prone to presume to replace the OS-specific library search strategies. It doesn't feel like library developers should assume they have to rewrite how native libraries are found...
This is happening here: https://github.com/dotnet/TorchSharp/blob/3d01e74707e3058cf82eeca7c6fe99ba13729e2f/src/TorchSharp/Torch.cs#L63 Instead, it should follow the pattern from `TryLoadNativeLibraryByName` just below: https://github.com/dotnet/TorchSharp/blob/3d01e74707e3058cf82eeca7c6fe99ba13729e2f/src/TorchSharp/Torch.cs#L78 The current implementation makes it really hard to debug what is happening when undocumented...
For example, the following codes can produce a warning in the console ```csharp using TorchSharp; var loss = torch.nn.MSELoss(); var x = torch.zeros([]); var y = torch.zeros([1]); loss.call(x, y); ```...
closes #1266 --- **It includes some aggressive changes. Careful checks are required.**
A System.AccessViolationException Error occurs when executing a simple code model = jit.load(modelPath, DeviceType.CUDA); how can I fit that the torchsharp version is TorchSharp-cuda-windows 0.100.7 **the details of error information:** System.AccessViolationException...
Apparently TorchSharp doesn't support quantized models. Am I missing something?
Hello, Hope this not an error with my code... So I really like to use ranges. Lets assume tensor of 3 dimensions. I can do x[some tensor] .... I can...
@NiklasGustafsson @ericstj @GeorgeS2019 Probably people want to have a place to download pre-trained model files so that they can fine-tune pre-made models by TorchVision and TorchAudio without converting files themselves....
Autocast
Soon i will try make AMP (Automatic Mixed Precision) with GradScaler.