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

ggml : online attention (CPU)

Browse source
This commit is contained in:
Georgi Gerganov 2024-01-20 12:26:49 +02:00
parent c3cdfffa88
commit a9681febd6
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
6 changed files with 231 additions and 198 deletions
Download patch file Download diff file Expand all files Collapse all files

5
ggml.h
View file

@ -1620,6 +1620,11 @@ extern "C" {
struct ggml_tensor * v,
bool masked);
// q: [n_embd, n_batch, n_head, 1]
// k: [n_embd, n_kv, n_head_kv, 1]
// v: [n_embd, n_kv, n_head_kv, 1] !! not transposed !!
// mask: [n_kv, n_batch, 1, 1]
// res: [n_embd, n_head, n_batch, 1] !! permuted !!
GGML_API struct ggml_tensor * ggml_flash_attn_ext(
struct ggml_context * ctx,
struct ggml_tensor * q,