Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-24 13:28:50 +01:00
Code Issues Packages Projects Releases Wiki Activity

llava : remove extra cont (#5587)

Browse Source
This commit is contained in:
Georgi Gerganov 2024-02-19 15:23:17 +02:00
parent 6fd413791a
commit 1387cf60f7
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
examples/llava/clip.cpp
View File

@ -616,7 +616,7 @@ static ggml_cgraph * clip_image_build_graph(clip_ctx * ctx, const clip_image_f32
KQ = ggml_soft_max_inplace(ctx0, KQ); KQ = ggml_soft_max_inplace(ctx0, KQ);
struct ggml_tensor * KQV = ggml_mul_mat(ctx0, V, KQ); struct ggml_tensor * KQV = ggml_mul_mat(ctx0, V, KQ);
KQV = ggml_reshape_4d(ctx0, KQV, d_head, num_positions, n_head, batch_size); KQV = ggml_reshape_4d(ctx0, KQV, d_head, num_positions, n_head, batch_size);
KQV = ggml_cont(ctx0, ggml_permute(ctx0, KQV, 0, 2, 1, 3)); KQV = ggml_permute(ctx0, KQV, 0, 2, 1, 3);
cur = ggml_cont_3d(ctx0, KQV, hidden_size, num_positions, batch_size); cur = ggml_cont_3d(ctx0, KQV, hidden_size, num_positions, batch_size);
} }