ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-04-24 16:39:45 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add mtmd: clip.cpp compiles

Browse Source
This commit is contained in:
Iwan Kawrakow
2025-09-25 13:23:13 +03:00
parent 59133173fa
commit 6b0c8e02a8
3 changed files with 585 additions and 75 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
examples/mtmd/clip.cpp
View File

@@ -4283,15 +4283,18 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
GGML_ABORT("Unknown projector type");
}
// ggml_backend_cpu_set_n_threads(ctx->backend_cpu, n_threads);
ggml_backend_dev_t dev = ggml_backend_get_device(ctx->backend_cpu);
ggml_backend_reg_t reg = dev ? ggml_backend_dev_backend_reg(dev) : nullptr;
if (reg) {
auto ggml_backend_set_n_threads_fn = (ggml_backend_set_n_threads_t) ggml_backend_reg_get_proc_address(reg, "ggml_backend_set_n_threads");
if (ggml_backend_set_n_threads_fn) {
ggml_backend_set_n_threads_fn(ctx->backend_cpu, n_threads);
}
if (ctx->backend_cpu) {
ggml_backend_cpu_set_n_threads(ctx->backend_cpu, n_threads);
}
//// ggml_backend_cpu_set_n_threads(ctx->backend_cpu, n_threads);
//ggml_backend_dev_t dev = ggml_backend_get_device(ctx->backend_cpu);
//ggml_backend_reg_t reg = dev ? ggml_backend_dev_backend_reg(dev) : nullptr;
//if (reg) {
// auto ggml_backend_set_n_threads_fn = (ggml_backend_set_n_threads_t) ggml_backend_reg_get_proc_address(reg, "ggml_backend_set_n_threads");
// if (ggml_backend_set_n_threads_fn) {
// ggml_backend_set_n_threads_fn(ctx->backend_cpu, n_threads);
// }
//}
auto status = ggml_backend_sched_graph_compute(ctx->sched.get(), gf);
if (status != GGML_STATUS_SUCCESS) {
@@ -4312,7 +4315,9 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
}
// the last node is the embedding tensor
ggml_tensor * embeddings = ggml_graph_node(gf, -1);
//ggml_tensor * embeddings = ggml_graph_node(gf, -1);
GGML_ASSERT(gf->n_nodes > 0);
ggml_tensor * embeddings = gf->nodes[gf->n_nodes-1];
// sanity check (only support batch size of 1 for now)
const int n_tokens_out = embeddings->ne[1];