@John - In case it wasn't clear from earlier messages, "GPU" has nothing to do with the user interface, it originally means "Graphics processing unit" and refers to specific dedicated hardware accelerators for graphics typically for video games, but such processors are now commonly used for high-performance numerical computations that have nothing to do with graphics and video games. Neural nets in Go are an example of a thing that has nothing to do with graphics that requires high-performance numerical computations to run efficiently.John Fairbairn wrote:I tried this and, while it didn't work, the information outputted may help someone:KataGo tries to make it plain that it's doing this tuning, by outputting a bunch of stuff to the command line as it's doing so showing progress in the various phases of the tuning,
Uncaught exception: OpenCL error at C:\Data\Data\Coding\Python\Katago\cpp\neuralnet\openclhelpers.cpp, func err, line 188, error CL_PLATFORM_NOT_FOUND_KHR
"CPU" just refers to the ordinary kind of processor that every modern computer has.
With a dedicated graphics card such as from Nvidia or AMD, the GPU version of a program will typically run much faster (get way more playouts per second, search much deeper) and play much stronger than the CPU version, which is why the GPU is preferable if you have such a card. They are identical in terms of interface, the only difference is things like the playouts/s that you will see.
OpenCL is generally used as a cross-platform means for programs to access a GPU. KataGo has no pure-CPU version right now, so it wants you to have OpenCL and *some* form of graphics card or acceleration. The error you got means that KataGo did NOT find such an OpenCL platform on your computer, so you probably don't have a dedicated graphics card such as from Nvidia or AMD on your computer - is that right? (Or if you do, the possibly their drivers are not up to date).
If you indeed don't have those, then you probably are not going to be able to use KataGo for now. A pure CPU version is on the list of things to do, but will not be available for quite some time. There *are* pure-CPU implementations of OpenCL as well that you could try installing - for example if you have a recent Intel CPU the following in theory could work: https://software.intel.com/en-us/opencl-sdk . But that's probably getting pretty far down the rabbit hole, and at that point it would probably be easier to just stick with the CPU version of LZ.
If on the other hand you do think you have a graphics/GPU accelerator or such on your computer, then the error is not that you don't have one, but that it is not being detected for some reason. In that case, yes, it would be worth digging further as Tryss has suggested.
Does that mostly make sense?