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

ggml: add SWIGLU_CLAMP (#27930)

Browse Source 
* ggml: add SWIGLU_CLAMP

* add vulkan shader
This commit is contained in:
Aman Gupta
2026-08-30 23:00:02 +08:00
committed by GitHub
parent 2578138397
commit 0190529ec4
40 changed files with 768 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-opencl/ggml-opencl.cpp
+16 -3
View File
@@ -744,8 +744,9 @@ struct ggml_backend_opencl_context {
cl_kernel kernel_tri;
cl_kernel kernel_fill;
cl_kernel kernel_clamp;
cl_kernel kernel_geglu, kernel_reglu, kernel_swiglu, kernel_swiglu_oai, kernel_geglu_erf, kernel_geglu_quick,
kernel_geglu_f16, kernel_reglu_f16, kernel_swiglu_f16, kernel_geglu_erf_f16, kernel_geglu_quick_f16;
cl_kernel kernel_geglu, kernel_reglu, kernel_swiglu, kernel_swiglu_oai, kernel_swiglu_clamp, kernel_geglu_erf,
kernel_geglu_quick, kernel_geglu_f16, kernel_reglu_f16, kernel_swiglu_f16, kernel_swiglu_clamp_f16,
kernel_geglu_erf_f16, kernel_geglu_quick_f16;
cl_kernel kernel_norm, kernel_norm_mul_add;
cl_kernel kernel_rms_norm, kernel_rms_norm_mul;
cl_kernel kernel_l2_norm_f32;
@@ -1601,11 +1602,13 @@ static void load_cl_kernels(ggml_backend_opencl_context *backend_ctx) {
CL_CHECK((backend_ctx->kernel_reglu = clCreateKernel(backend_ctx->program_glu, "kernel_reglu", &err), err));
CL_CHECK((backend_ctx->kernel_swiglu = clCreateKernel(backend_ctx->program_glu, "kernel_swiglu", &err), err));
CL_CHECK((backend_ctx->kernel_swiglu_oai = clCreateKernel(backend_ctx->program_glu, "kernel_swiglu_oai", &err), err));
CL_CHECK((backend_ctx->kernel_swiglu_clamp = clCreateKernel(backend_ctx->program_glu, "kernel_swiglu_clamp", &err), err));
CL_CHECK((backend_ctx->kernel_geglu_erf = clCreateKernel(backend_ctx->program_glu, "kernel_geglu_erf", &err), err));
CL_CHECK((backend_ctx->kernel_geglu_quick = clCreateKernel(backend_ctx->program_glu, "kernel_geglu_quick", &err), err));
CL_CHECK((backend_ctx->kernel_geglu_f16 = clCreateKernel(backend_ctx->program_glu, "kernel_geglu_f16", &err), err));
CL_CHECK((backend_ctx->kernel_reglu_f16 = clCreateKernel(backend_ctx->program_glu, "kernel_reglu_f16", &err), err));
CL_CHECK((backend_ctx->kernel_swiglu_f16 = clCreateKernel(backend_ctx->program_glu, "kernel_swiglu_f16", &err), err));
CL_CHECK((backend_ctx->kernel_swiglu_clamp_f16 = clCreateKernel(backend_ctx->program_glu, "kernel_swiglu_clamp_f16", &err), err));
CL_CHECK((backend_ctx->kernel_geglu_erf_f16 = clCreateKernel(backend_ctx->program_glu, "kernel_geglu_erf_f16", &err), err));
CL_CHECK((backend_ctx->kernel_geglu_quick_f16 = clCreateKernel(backend_ctx->program_glu, "kernel_geglu_quick_f16", &err), err));
GGML_LOG_CONT(".");
@@ -7700,6 +7703,7 @@ static bool ggml_opencl_supports_op(ggml_backend_dev_t dev, const struct ggml_te
case GGML_GLU_OP_SWIGLU_OAI:
case GGML_GLU_OP_GEGLU_ERF:
case GGML_GLU_OP_GEGLU_QUICK:
case GGML_GLU_OP_SWIGLU_CLAMP:
return ggml_is_contiguous_1(op->src[0]) && (op->type == GGML_TYPE_F32 || op->type == GGML_TYPE_F16);
default:
return false;
@@ -24886,6 +24890,13 @@ static void ggml_cl_glu(ggml_backend_t backend, const ggml_tensor * src0, const
case GGML_GLU_OP_SWIGLU_OAI:
kernel = backend_ctx->kernel_swiglu_oai;
break;
case GGML_GLU_OP_SWIGLU_CLAMP:
if (dst->type == GGML_TYPE_F32) {
kernel = backend_ctx->kernel_swiglu_clamp;
} else {
kernel = backend_ctx->kernel_swiglu_clamp_f16;
}
break;
case GGML_GLU_OP_GEGLU_ERF:
if (dst->type == GGML_TYPE_F32) {
kernel = backend_ctx->kernel_geglu_erf;
@@ -24941,8 +24952,10 @@ static void ggml_cl_glu(ggml_backend_t backend, const ggml_tensor * src0, const
CL_CHECK(clSetKernelArg(kernel, 10, sizeof(int), &ne00_off));
CL_CHECK(clSetKernelArg(kernel, 11, sizeof(int), &ne10_off));
if (ggml_get_glu_op(dst) == GGML_GLU_OP_SWIGLU_OAI) {
if (ggml_get_glu_op(dst) == GGML_GLU_OP_SWIGLU_OAI || ggml_get_glu_op(dst) == GGML_GLU_OP_SWIGLU_CLAMP) {
CL_CHECK(clSetKernelArg(kernel, 12, sizeof(float), &limit));
}
if (ggml_get_glu_op(dst) == GGML_GLU_OP_SWIGLU_OAI) {
CL_CHECK(clSetKernelArg(kernel, 13, sizeof(float), &alpha));
}
ggml/src/ggml-opencl/kernels/glu.cl
+65
View File
@@ -243,6 +243,71 @@ kernel void kernel_swiglu_oai(
}
}
//------------------------------------------------------------------------------
// swiglu_clamp
//------------------------------------------------------------------------------
kernel void kernel_swiglu_clamp(
global char * src0,
ulong offset0,
global char * src1,
ulong offset1,
global char * dst,
ulong offsetd,
ulong nb01,
ulong nb11,
int ne0,
ulong nb1,
int ne00_off,
int ne10_off,
float limit
) {
src0 = (global char*)((global char*)src0 + offset0);
src1 = (global char*)((global char*)src1 + offset1);
dst = (global char*)((global char*)dst + offsetd);
global float * src0_row = (global float *) ((global char *) src0 + get_group_id(0)*nb01) + ne00_off;
global float * src1_row = (global float *) ((global char *) src1 + get_group_id(0)*nb11) + ne10_off;
global float * dst_row = (global float *) ((global char *) dst + get_group_id(0)*nb1);
for (int i0 = get_local_id(0); i0 < ne0; i0 += get_local_size(0)) {
const float gate = min(src0_row[i0], limit);
const float up = clamp(src1_row[i0], -limit, limit);
dst_row[i0] = gate / (1.0f + exp(-gate)) * up;
}
}
kernel void kernel_swiglu_clamp_f16(
global char * src0,
ulong offset0,
global char * src1,
ulong offset1,
global char * dst,
ulong offsetd,
ulong nb01,
ulong nb11,
int ne0,
ulong nb1,
int ne00_off,
int ne10_off,
float limit
) {
src0 = (global char*)((global char*)src0 + offset0);
src1 = (global char*)((global char*)src1 + offset1);
dst = (global char*)((global char*)dst + offsetd);
global half * src0_row = (global half *) ((global char *) src0 + get_group_id(0)*nb01) + ne00_off;
global half * src1_row = (global half *) ((global char *) src1 + get_group_id(0)*nb11) + ne10_off;
global half * dst_row = (global half *) ((global char *) dst + get_group_id(0)*nb1);
for (int i0 = get_local_id(0); i0 < ne0; i0 += get_local_size(0)) {
const float gate = min((float) src0_row[i0], limit);
const float up = clamp((float) src1_row[i0], -limit, limit);
dst_row[i0] = (half) (gate / (1.0f + exp(-gate)) * up);
}
}
//------------------------------------------------------------------------------
// geglu_erf
//------------------------------------------------------------------------------