EngineX-Ascend/enginex-ascend-910-llama.cpp
10
0
Fork 0
Code Issues Pull Requests Actions 4 Projects Releases Wiki Activity

vulkan: support multi/vision rope, and noncontiguous rope (#11902)

Browse Source
This commit is contained in:
Jeff Bolz
2025-02-16 01:52:23 -06:00
committed by GitHub
parent c2ea16f260
commit bf42a23d0a
7 changed files with 204 additions and 41 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
ggml/src/ggml-vulkan/vulkan-shaders/rope_norm.comp
View File

@@ -3,15 +3,18 @@
#include "rope_head.comp"
void main() {
const uint col = gl_GlobalInvocationID.y * 2;
const uint row = gl_GlobalInvocationID.x;
const uint i0 = 2*gl_GlobalInvocationID.y;
uint ne0 = p.ncols;
uint ne1 = p.p_delta_rows;
if (col >= p.ncols) {
if (i0 >= ne0) {
return;
}
if (col >= p.n_dims) {
const uint i = row*p.ncols + col;
const uint row_dst = gl_GlobalInvocationID.x;
if (i0 >= p.n_dims) {
const uint i = row_dst*ne0 + i0;
data_d[i + 0] = data_a[i + 0];
data_d[i + 1] = data_a[i + 1];
@@ -19,19 +22,22 @@ void main() {
return;
}
const uint i = row*p.ncols + col;
const uint i2 = row/p.p_delta_rows;
const uint row_x = row_dst % ne1;
const uint channel_x = row_dst / ne1;
const float theta_base = data_pos[i2] * pow(p.theta_scale, col/2.0f);
const uint idst = row_dst*ne0 + i0;
const uint ix = channel_x*p.s2 + row_x*p.s1 + i0;
const float freq_factor = p.has_ff != 0 ? data_ff[col/2] : 1.0f;
const float theta_base = data_pos[channel_x] * pow(p.theta_scale, i0/2.0f);
const float freq_factor = p.has_ff != 0 ? data_ff[i0/2] : 1.0f;
float cos_theta, sin_theta;
rope_yarn(theta_base / freq_factor, col, cos_theta, sin_theta);
rope_yarn(theta_base / freq_factor, i0, cos_theta, sin_theta);
const float x0 = float(data_a[i + 0]);
const float x1 = float(data_a[i + 1]);
const float x0 = float(data_a[ix + 0]);
const float x1 = float(data_a[ix + 1]);
data_d[i + 0] = D_TYPE(x0*cos_theta - x1*sin_theta);
data_d[i + 1] = D_TYPE(x0*sin_theta + x1*cos_theta);
data_d[idst + 0] = D_TYPE(x0*cos_theta - x1*sin_theta);
data_d[idst + 1] = D_TYPE(x0*sin_theta + x1*cos_theta);
}