mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2025-07-30 06:03:37 -04:00
Vulkan Shader Refactor, Memory Debugging Option (#7947)
* Refactor shaders, extract GLSL code from ggml_vk_generate_shaders.py into vulkan-shaders directory * Improve debug log code * Add memory debug output option * Fix flake8 * Fix unnecessary high llama-3 VRAM use
This commit is contained in:
25
vulkan-shaders/gelu.comp
Normal file
25
vulkan-shaders/gelu.comp
Normal file
@@ -0,0 +1,25 @@
|
||||
#version 450
|
||||
|
||||
#include "generic_head.comp"
|
||||
#include "types.comp"
|
||||
|
||||
#extension GL_EXT_control_flow_attributes : enable
|
||||
|
||||
layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
|
||||
|
||||
layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
|
||||
layout (binding = 1) writeonly buffer D {D_TYPE data_d[];};
|
||||
|
||||
void main() {
|
||||
const float GELU_COEF_A = 0.044715f;
|
||||
const float SQRT_2_OVER_PI = 0.79788456080286535587989211986876f;
|
||||
const uint i = gl_GlobalInvocationID.x;
|
||||
|
||||
if (i >= p.KX) {
|
||||
return;
|
||||
}
|
||||
|
||||
const float xi = float(data_a[i]);
|
||||
const float val = SQRT_2_OVER_PI*xi*(1.0f + GELU_COEF_A*xi*xi);
|
||||
data_d[i] = D_TYPE(0.5f*xi*(2.0f - 2.0f / (exp(2 * val) + 1)));
|
||||
}
|
Reference in New Issue
Block a user