tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-19 22:36:13 -04:00
Code Issues Packages Projects Releases Wiki Activity

ggml : add support for ChatGLM RoPE

Browse Source
This commit is contained in:
Georgi Gerganov
2023-06-27 00:06:51 +03:00
parent d38e451578
commit d9779021bd
2 changed files with 76 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
ggml.h
View File

@@ -1036,13 +1036,15 @@ extern "C" {
// rotary position embedding
// if mode & 1 == 1, skip n_past elements
// if mode & 2 == 1, GPT-NeoX style
// if mode & 4 == 1, ChatGLM style
// TODO: avoid creating a new tensor every time
GGML_API struct ggml_tensor * ggml_rope(
struct ggml_context * ctx,
struct ggml_tensor * a,
int n_past,
int n_dims,
int mode);
int mode,
int n_ctx);
// in-place, returns view(a)
GGML_API struct ggml_tensor * ggml_rope_inplace(
@@ -1050,7 +1052,8 @@ extern "C" {
struct ggml_tensor * a,
int n_past,
int n_dims,
int mode);
int mode,
int n_ctx);
// rotary position embedding backward, i.e compute dx from dy
// a - dy