Steward Garcia

Great job!! I'm struggling to achieve performance improvement in CUDA because I'm having issues where 90% of the kernel execution time is spent on memory I/O, and the remaining 10%...

@ggerganov I'm trying to port the kernel you already have in Metal to CUDA, but I'm not completely clear on how it works yet. So, I would appreciate it if...

@ggerganov in the function `simdgroup_load`, what is the last parameter passed? I assume it's the stride of the data or am I wrong? ```c++ // load the queries from shared...

@ggerganov Is this behaivor expected? Did you think that if it was like this, all the elements of the array would be negative infinity in Metal? 

@ggerganov I have been examining the kernel I created in CUDA, but it produces incorrect values despite all the operations being exactly the same. I really want to ask for...

@compilade ust out of curiosity, is any convolution operation performed? I see some tensors with the name `conv`, but I never see `ggml_conv_1d` or `ggml_conv_2d` being used at any point.

the model file doesn't exist

@ggerganov I have created a web application to stress-test the server and see how it handles multiple clients sending random questions and documents simultaneously. I tested it with four clients...

@Areshkew I'm getting this on windows: ```bash wave.c wave.vcxproj -> C:\Users\stewa\Downloads\VkGalaxy-main\glfw-master\build\examples\Debug\wave.exe 1>Building Custom Rule C:/Users/stewa/Downloads/VkGalaxy-main/glfw-master/tests/CMakeLists.txt window.c window.vcxproj -> C:\Users\stewa\Downloads\VkGalaxy-main\glfw-master\build\tests\Debug\window.exe 1>Building Custom Rule C:/Users/stewa/Downloads/VkGalaxy-main/glfw-master/examples/CMakeLists.txt windows.c windows.vcxproj -> C:\Users\stewa\Downloads\VkGalaxy-main\glfw-master\build\examples\Debug\windows.exe 1>Building Custom Rule...

> @FSSRepo For windows, you gotta make the build files, where you will find the cmake-gui to be easiest way. I realized that I had to build it in Release...