Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-01-30 13:53:03 +01:00
Code Issues Packages Projects Releases Wiki Activity

Review update: Use GGML_SYCL_DEBUG

Browse Source
This commit is contained in:
Akarshan Biswas 2025-01-20 20:26:02 +05:30
parent b913e83c81
commit 53847e4a55
No known key found for this signature in database
GPG Key ID: 52A578A14B32134D
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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

@ -249,12 +249,12 @@ void ggml_sycl_op_soft_max(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
dpct::queue_ptr main_stream = ctx.stream();
if (dst->src[1] && dst->src[1]->type == GGML_TYPE_F16) {
//printf("%s: fp16 mask\n", __func__);
GGML_SYCL_DEBUG("%s: fp16 mask\n", __func__);
const sycl::half * src1_dd = static_cast<sycl::half *>(dst->src[1]->data);
soft_max_f32_sycl<sycl::half>(src0_dd, src1_dd, dst_dd, ne00, nrows_x, nrows_y, scale, max_bias,
main_stream, ctx.device);
} else if (dst->src[1] && dst->src[1]->type == GGML_TYPE_F32) {
//printf("%s: fp32 mask\n", __func__);
GGML_SYCL_DEBUG("%s: fp32 mask\n", __func__);
const float * src1_dd = static_cast<const float *>(dst->src[1]->data);
soft_max_f32_sycl<float>(src0_dd, src1_dd, dst_dd, ne00, nrows_x, nrows_y, scale, max_bias, main_stream, ctx.device);
} else {