Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

OpenCL: Add concat, tsembd, upscale, tanh, pad and repeat (#13840)

Browse Source 
* add concat, pad, repeat, tsembd, tanh, upscale

* small fixes
This commit is contained in:
rmatif
2025-06-02 16:53:36 -07:00
committed by GitHub
parent 3637576288
commit bfb1e012a0
8 changed files with 1156 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-opencl/kernels/tsembd.cl
+48
View File
@@ -0,0 +1,48 @@
kernel void kernel_timestep_embedding(
global const void * p_timesteps,
ulong off_timesteps,
global void * p_dst,
ulong off_dst,
int dst_nb1_bytes,
int logical_dim,
int max_period
) {
int local_i;
int local_j;
int local_half_dim;
float local_timestep_val;
float local_freq;
float local_arg;
global float * local_embed_data_ptr;
global const float * local_timesteps_input_ptr;
global float * local_dst_output_base_ptr;
local_timesteps_input_ptr = (global const float *)((global char *)p_timesteps + off_timesteps);
local_dst_output_base_ptr = (global float *)((global char *)p_dst + off_dst);
local_i = get_global_id(1);
local_j = get_global_id(0);
local_half_dim = logical_dim / 2;
local_embed_data_ptr = (global float *)((global char *)local_dst_output_base_ptr + local_i * dst_nb1_bytes);
if (logical_dim % 2 != 0 && local_j == ((logical_dim + 1) / 2)) {
local_embed_data_ptr[logical_dim] = 0.0f;
}
if (local_j >= local_half_dim) {
return;
}
local_timestep_val = local_timesteps_input_ptr[local_i];
if (local_half_dim == 0) {
local_freq = 1.0f;
} else {
local_freq = exp(-log((float)max_period) * (float)local_j / (float)local_half_dim);
}
local_arg = local_timestep_val * local_freq;
local_embed_data_ptr[local_j] = cos(local_arg);
local_embed_data_ptr[local_j + local_half_dim] = sin(local_arg);
}