Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-02-07 08:53:16 +01:00
Code Issues Packages Projects Releases Wiki Activity

Remove changes not related to softmax

Browse Source
This commit is contained in:
Akarshan Biswas 2025-01-20 17:41:55 +05:30
parent 495e7ea48d
commit 82d5c0dd80
No known key found for this signature in database
GPG Key ID: 52A578A14B32134D
3 changed files with 8 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-sycl/common.cpp
View File

@ -79,7 +79,7 @@ void ggml_sycl_op_flatten(ggml_backend_sycl_context & ctx, const ggml_tensor *sr
float * src0_ddf = (float *) src0->data; float * src0_ddf = (float *) src0->data;
float * src1_ddf = use_src1 ? (float *) src1->data : nullptr; float * src1_ddf = use_src1 ? (float *) src1->data : nullptr;
float * dst_ddf = (float *) dst->data; float * dst_ddf = (float *) dst->data;
/* These are never used */
ggml_sycl_pool_alloc<float> src0_f(ctx.pool()); ggml_sycl_pool_alloc<float> src0_f(ctx.pool());
ggml_sycl_pool_alloc<float> src1_f(ctx.pool()); ggml_sycl_pool_alloc<float> src1_f(ctx.pool());
ggml_sycl_pool_alloc<float> dst_f(ctx.pool()); ggml_sycl_pool_alloc<float> dst_f(ctx.pool());

4
ggml/src/ggml-sycl/dmmv.cpp
View File

@ -2,7 +2,6 @@
#include "dmmv.hpp" #include "dmmv.hpp"
#include "dequantize.hpp" #include "dequantize.hpp"
#include "presets.hpp" #include "presets.hpp"
#include "ggml-impl.h"
static void convert_f16(const void * vx, const int64_t ib, const int iqs, dfloat2 & v){ static void convert_f16(const void * vx, const int64_t ib, const int iqs, dfloat2 & v){
@ -974,7 +973,6 @@ void ggml_sycl_op_dequantize_mul_mat_vec(
} }
#else #else
const dfloat * src1_dfloat = (const dfloat *) src1_ddf_i; // dfloat == float, no conversion const dfloat * src1_dfloat = (const dfloat *) src1_ddf_i; // dfloat == float, no conversion
GGML_UNUSED(ctx);
#endif // GGML_SYCL_F16 #endif // GGML_SYCL_F16
switch (src0->type) { switch (src0->type) {
@ -1012,7 +1010,7 @@ void ggml_sycl_op_dequantize_mul_mat_vec(
convert_mul_mat_vec_f16_sycl(src0_dd_i, src1_dfloat, dst_dd_i, ne00, row_diff, stream); convert_mul_mat_vec_f16_sycl(src0_dd_i, src1_dfloat, dst_dd_i, ne00, row_diff, stream);
break; break;
default: default:
GGML_LOG_ERROR("ggml_sycl_op_dequantize_mul_mat_vec unsupported GGML_TYPE %d\n", src0->type); printf("ggml_sycl_op_dequantize_mul_mat_vec unsupported GGML_TYPE %d\n", src0->type);
GGML_ABORT("fatal error"); GGML_ABORT("fatal error");
break; break;
} }

6
ggml/src/ggml-sycl/wkv6.cpp
View File

@ -97,6 +97,9 @@ static void rwkv_wkv_f32_kernel(
void ggml_sycl_op_rwkv_wkv6(ggml_backend_sycl_context& ctx, ggml_tensor* dst) { void ggml_sycl_op_rwkv_wkv6(ggml_backend_sycl_context& ctx, ggml_tensor* dst) {
const ggml_tensor *src0 = dst->src[0];
const ggml_tensor *src1 = dst->src[1];
const float* k_d = (const float*)dst->src[0]->data; const float* k_d = (const float*)dst->src[0]->data;
const float* v_d = (const float*)dst->src[1]->data; const float* v_d = (const float*)dst->src[1]->data;
const float* r_d = (const float*)dst->src[2]->data; const float* r_d = (const float*)dst->src[2]->data;
@ -134,4 +137,7 @@ void ggml_sycl_op_rwkv_wkv6(ggml_backend_sycl_context& ctx, ggml_tensor* dst) {
); );
}); });
}); });
GGML_UNUSED(src0);
GGML_UNUSED(src1);
} }