public_git/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/whisper.cpp.git synced 2026-10-05 22:11:21 +02:00
Code Issues Packages Projects Releases Wiki Activity

ggml: add ggml_rope_set_offset (+ metal support) (llama/27120)

Browse Source 
* add params

* cpu kernel

* metal kernel

* add test backend ops

* gate other backends

* ggml: (cuda) support ggml_rope_set_offset (llama/27121)

* rm cuda supports_op guard, fix webgpu clang-format

* ggml: support ggml_rope_set_offset on vulkan (llama/27344)

* ggml: support ggml_rope_set_offset on vulkan

* remove inplace optimization
This commit is contained in:
Xuan-Son Nguyen
2026-08-21 19:23:26 +03:00
committed by Georgi Gerganov
parent d830bd220c
commit 8442c74f56
18 changed files with 244 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/include/ggml.h
+8
View File
@@ -1981,6 +1981,14 @@ extern "C" {
float beta_fast,
float beta_slow);
// set the offset dims for RoPE
// a must be GGML_OP_ROPE or GGML_OP_ROPE_BACK
// vision RoPE is not supported
// example: (marking: x = rotated, 0 = unrotated)
// n_embd = 10, n_dims = 4, offset = 2 --> [00xxxx0000]
GGML_API struct ggml_tensor * ggml_rope_set_offset(
struct ggml_tensor * a,
int n_offs);
// clamp
// in-place, returns view(a)