-
Notifications
You must be signed in to change notification settings - Fork 14.1k
Description
What happened?
When compiling the Minicpm model, during the process of image encoding, memory allocation operations are performed for the variables.
The first subgraph execution runs fine, but during the second malloc, it reports a core dump error. I have not been able to identify the cause. Please help me.
code file: llava.cpp
code: image_embd_v[i] = (float *)malloc(clip_embd_nbytes(ctx_clip)); ##clip_embd_nbytes(ctx_clip):917504
if (clip_is_minicpmv(ctx_clip)) {
std::vector<float *> image_embd_v;
image_embd_v.resize(img_res_v.size);
struct clip_image_size * load_image_size = clip_image_size_init();
for (size_t i = 0; i < img_res_v.size; i++) {
const int64_t t_img_enc_step_start_us = ggml_time_us();
image_embd_v[i] = (float *)malloc(clip_embd_nbytes(ctx_clip));
int patch_size=14;
load_image_size->width = img_res_v.data[i].nx;
load_image_size->height = img_res_v.data[i].ny;
clip_add_load_image_size(ctx_clip, load_image_size);
bool encoded = false;
Name and Version
./llama-llava-cli latest linux