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

ggml : sync ggml (add GPT-NeoX RoPE implementation)

Browse source
This commit is contained in:
Georgi Gerganov 2023-04-20 23:32:59 +03:00
parent 9ff334f3c9
commit 12b5900dbc
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
3 changed files with 49 additions and 17 deletions
Download patch file Download diff file Expand all files Collapse all files

3
ggml.h
View file

@ -630,7 +630,8 @@ struct ggml_tensor * ggml_soft_max(
// rotary position embedding
// in-place, returns view(a)
// if mode == 1, skip n_past elements
// if mode & 1 == 1, skip n_past elements
// if mode & 2 == 1, GPT-NeoX style
// TODO: avoid creating a new tensor every time
struct ggml_tensor * ggml_rope(
struct ggml_context * ctx,