Adam Pocock
Adam Pocock
Does it end up in the classpath that Maven/your IDE generates when you run the program?
Ok. Comparing the published artifacts for tensorflow-core-api it looks like the Windows one is missing tensorflow_framework, but I'm not sure if that's necessary on Windows as it might be compiled...
Fixed in `master`.
You can add a regularization term the way you usually would, put an l2 penalty on the weights being large and add that to the loss function that you backprop.
What versions of TF-Java and TF decision forests are you using? And what platform are you running on?
Looks like that TFDF version is pretty old. Try using the 1.0.1 release, that should be compatible with TF 2.10 which is the version built into TF-Java 0.5.0.
We've moved to using the latest binaries from tensorflow so this should work better now, without symbol mismatches.
Converting from a data frame to a byte array has a lot of options on how you represent the data, columns and encoding. Do you have something specific in mind...
We don't provide binaries for ARM64. You should be able to compile them from source yourself on Linux, but it will take a while.
I believe there are ARM64 binaries for JavaCPP, so it's mainly compiling TensorFlow. To do that check out this project, install the right version of bazel & maven, then run...