upstream-archive/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-10-03 19:37:29 -05:00
Code Issues Packages Projects Releases Wiki Activity

Small cleanup

Browse Source
This commit is contained in:
Pedro Cuenca
2026-08-06 13:57:31 +02:00
parent 26ed5924e7
commit ebac3dd7ea
1 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/mtmd/models/onyx.cpp
+5 -5
View File
@@ -2,9 +2,7 @@
// Onyx vision encoder: 50-layer ViT with 2D RoPE, sparse block-diagonal
// window attention (every 4th + last layer global), pixel-shuffle downsample, then
// adapter MLP + LLM's vision_projection. Output dim = 6656 (onyx n_embd),
// injected via llama_batch.embd; the onyx LLM graph applies the scaleless rms_norm
// (== reference perception_emb_norm).
// adapter MLP + LLM's vision_projection.
//
// Several quantities are precomputed on host and fed as named graph inputs (filled in
// clip.cpp set_input, PROJECTOR_TYPE_ONYX branch):
@@ -23,7 +21,8 @@ ggml_cgraph * clip_graph_onyx::build() {
auto inp_i32 = [&](const char * name, int64_t n) {
ggml_tensor * t = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n);
ggml_set_name(t, name); ggml_set_input(t);
ggml_set_name(t, name);
ggml_set_input(t);
return t;
};
@@ -34,7 +33,8 @@ ggml_cgraph * clip_graph_onyx::build() {
ggml_tensor * ds_perm = inp_i32("onyx_ds_perm", n_tok);
ggml_tensor * sp_mask = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_tok, n_tok);
ggml_set_name(sp_mask, "onyx_sp_mask"); ggml_set_input(sp_mask);
ggml_set_name(sp_mask, "onyx_sp_mask");
ggml_set_input(sp_mask);
// patchify via build_inp (conv2d over raw pixels) + bilinear-resized learned pos-emb
ggml_tensor * x = build_inp(); // [n_embd, n_tok, 1]