Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 05:48:47 +01:00
Code Issues Packages Projects Releases Wiki Activity

baby-llama : fix build after ggml_rope change (#2016)

Browse Source
This commit is contained in:
Howard Su 2023-06-27 13:07:13 +08:00 committed by GitHub
parent 181e8d9755
commit 0be54f75a6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
examples/baby-llama/baby-llama.cpp
View File

@ -566,8 +566,8 @@ struct ggml_tensor * forward(
// wk shape [n_embd, n_embd, 1, 1] // wk shape [n_embd, n_embd, 1, 1]
// Qcur shape [n_embd/n_head, n_head, N, 1] // Qcur shape [n_embd/n_head, n_head, N, 1]
// Kcur shape [n_embd/n_head, n_head, N, 1] // Kcur shape [n_embd/n_head, n_head, N, 1]
struct ggml_tensor * Qcur = ggml_rope(ctx0, ggml_reshape_3d(ctx0, ggml_mul_mat(ctx0, model->layers[il].wq, cur), n_embd/n_head, n_head, N), n_past, n_rot, 0); struct ggml_tensor * Qcur = ggml_rope(ctx0, ggml_reshape_3d(ctx0, ggml_mul_mat(ctx0, model->layers[il].wq, cur), n_embd/n_head, n_head, N), n_past, n_rot, 0, 0);
struct ggml_tensor * Kcur = ggml_rope(ctx0, ggml_reshape_3d(ctx0, ggml_mul_mat(ctx0, model->layers[il].wk, cur), n_embd/n_head, n_head, N), n_past, n_rot, 0); struct ggml_tensor * Kcur = ggml_rope(ctx0, ggml_reshape_3d(ctx0, ggml_mul_mat(ctx0, model->layers[il].wk, cur), n_embd/n_head, n_head, N), n_past, n_rot, 0, 0);
// store key and value to memory // store key and value to memory
{ {
@ -823,8 +823,8 @@ struct ggml_tensor * forward_batch(
// wk shape [n_embd, n_embd, 1, 1] // wk shape [n_embd, n_embd, 1, 1]
// Qcur shape [n_embd/n_head, n_head, N, n_batch] // Qcur shape [n_embd/n_head, n_head, N, n_batch]
// Kcur shape [n_embd/n_head, n_head, N, n_batch] // Kcur shape [n_embd/n_head, n_head, N, n_batch]
struct ggml_tensor * Qcur = ggml_rope(ctx0, ggml_reshape_4d(ctx0, ggml_mul_mat(ctx0, model->layers[il].wq, cur), n_embd/n_head, n_head, N, n_batch), n_past, n_rot, 0); struct ggml_tensor * Qcur = ggml_rope(ctx0, ggml_reshape_4d(ctx0, ggml_mul_mat(ctx0, model->layers[il].wq, cur), n_embd/n_head, n_head, N, n_batch), n_past, n_rot, 0, 0);
struct ggml_tensor * Kcur = ggml_rope(ctx0, ggml_reshape_4d(ctx0, ggml_mul_mat(ctx0, model->layers[il].wk, cur), n_embd/n_head, n_head, N, n_batch), n_past, n_rot, 0); struct ggml_tensor * Kcur = ggml_rope(ctx0, ggml_reshape_4d(ctx0, ggml_mul_mat(ctx0, model->layers[il].wk, cur), n_embd/n_head, n_head, N, n_batch), n_past, n_rot, 0, 0);
assert_shape_4d(Qcur, n_embd/n_head, n_head, N, n_batch); assert_shape_4d(Qcur, n_embd/n_head, n_head, N, n_batch);
assert_shape_4d(Kcur, n_embd/n_head, n_head, N, n_batch); assert_shape_4d(Kcur, n_embd/n_head, n_head, N, n_batch);
@ -1116,7 +1116,7 @@ struct ggml_tensor * forward_lora(
model->layers[il].wqb, model->layers[il].wqb,
cur)), cur)),
n_embd/n_head, n_head, N), n_embd/n_head, n_head, N),
n_past, n_rot, 0); n_past, n_rot, 0, 0);
struct ggml_tensor * Kcur = ggml_rope(ctx0, struct ggml_tensor * Kcur = ggml_rope(ctx0,
ggml_reshape_3d(ctx0, ggml_reshape_3d(ctx0,
ggml_mul_mat(ctx0, ggml_mul_mat(ctx0,
@ -1125,7 +1125,7 @@ struct ggml_tensor * forward_lora(
model->layers[il].wkb, model->layers[il].wkb,
cur)), cur)),
n_embd/n_head, n_head, N), n_embd/n_head, n_head, N),
n_past, n_rot, 0); n_past, n_rot, 0, 0);
// store key and value to memory // store key and value to memory
{ {