jxtps

Unfortunately still not working, with seemingly the same failure mode? CPU:   GPU:   _Previous report masked the missing cudart64_12.dll since I had a lingering Cuda 12.6 installation,...

I downloaded the latest version listed on https://learn.microsoft.com/en-us/cpp/windows/latest-supported-vc-redist?view=msvc-170 i.e. `X86 https://aka.ms/vs/17/release/vc_redist.x86.exe Permalink for latest supported x86 version` After installing and rebooting I tried again, and got the same result as...

Ah, sorry, yes, that was the one I actually installed, when writing the comment I copy-pasted the wrong line from the table.

Tried that but unfortunately still the same result as before?!

The full app is a complex beast, the way I ran it was in a static block before any javacpp / pytorch is touched - I’ve been using the same...

Sorry, I’m quite confused here - are you saying my c:/windows/system32 DLLs are the wrong version? And that is somehow impacting the unpacking of the bytedeco JAR contents, so the...

I have two modes of use: 1. Create a new Tensor from scratch, and 2. Fill an existing tensor with new data (to avoid memory churn when batch processing a...

Also, it seems like the non-snapshot, published versions of `cuda-platform-redist` is lagging behind `cuda-platform` a bit? https://mvnrepository.com/artifact/org.bytedeco/cuda-platform has 12.9-9.10-1.5.12 https://mvnrepository.com/artifact/org.bytedeco/cuda-platform-redist ends on 12.6-9.5-1.5.11

Awesome, thank you! Separately, if binary sizes are becoming a problem, maybe it's time to let sm_50 (Maxwell, 2014) & sm_60 (Pascal, 2016) go? (I don't know how much space...