vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

remove unused ggml_rope_xpos_back

Browse source 
it is better to have only one `ggml_rope_back` function that accepts all rope parameters, so that `ggml_compute_backward` can propagate all parameters without having to switch between different rope_back variants.
This commit is contained in:
xaedes 2023-11-06 23:48:28 +01:00
parent 93e65d32a4
commit 926930ba17
No known key found for this signature in database
GPG key ID: 30030EDD817EA2B1
1 changed files with 0 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

8
ggml.h
View file

@ -1380,14 +1380,6 @@ extern "C" {
float beta_fast, float beta_fast,
bool beta_slow); bool beta_slow);
GGML_API struct ggml_tensor * ggml_rope_xpos_back(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
int n_dims,
float base,
bool down);
// alibi position embedding // alibi position embedding
// in-place, returns view(a) // in-place, returns view(a)
GGML_API struct ggml_tensor * ggml_alibi( GGML_API struct ggml_tensor * ggml_alibi(