vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

replace variable size array with vector

Browse source
This commit is contained in:
HimariO 2024-10-21 21:34:24 +08:00
parent 3d19dd44b6
commit 53480d2bdb
1 changed files with 13 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

17
examples/llava/qwen2vl-cli.cpp
View file

@ -40,14 +40,18 @@ static bool qwen2vl_eval_image_embed(llama_context * ctx_llama, const struct lla
*st_pos_id += std::max(pw, ph); *st_pos_id += std::max(pw, ph);
int processed = 0; int processed = 0;
std::vector<llama_pos> batch_mrope_pos;
batch_mrope_pos.resize(img_tokens * 4);
for (int i = 0; i < img_tokens; i += n_batch) { for (int i = 0; i < img_tokens; i += n_batch) {
int n_eval = img_tokens - i; int n_eval = img_tokens - i;
if (n_eval > n_batch) { if (n_eval > n_batch) {
n_eval = n_batch; n_eval = n_batch;
} }
llama_pos batch_mrope_pos[n_eval * 4]; // llama_pos batch_mrope_pos[n_eval * 4];
memcpy(batch_mrope_pos, &mrope_pos[processed], n_eval * sizeof(llama_pos)); std::fill(batch_mrope_pos.begin(), batch_mrope_pos.end(), 0);
memcpy(batch_mrope_pos.data(), &mrope_pos[processed], n_eval * sizeof(llama_pos));
memcpy(&batch_mrope_pos[n_eval * 1], &mrope_pos[img_tokens * 1 + processed], n_eval * sizeof(llama_pos)); memcpy(&batch_mrope_pos[n_eval * 1], &mrope_pos[img_tokens * 1 + processed], n_eval * sizeof(llama_pos));
memcpy(&batch_mrope_pos[n_eval * 2], &mrope_pos[img_tokens * 2 + processed], n_eval * sizeof(llama_pos)); memcpy(&batch_mrope_pos[n_eval * 2], &mrope_pos[img_tokens * 2 + processed], n_eval * sizeof(llama_pos));
memcpy(&batch_mrope_pos[n_eval * 3], &mrope_pos[img_tokens * 3 + processed], n_eval * sizeof(llama_pos)); memcpy(&batch_mrope_pos[n_eval * 3], &mrope_pos[img_tokens * 3 + processed], n_eval * sizeof(llama_pos));
@ -56,7 +60,7 @@ static bool qwen2vl_eval_image_embed(llama_context * ctx_llama, const struct lla
int32_t(n_eval), // n_tokens int32_t(n_eval), // n_tokens
nullptr, // token nullptr, // token
(image_embed->embed+i*n_embd), // embed (image_embed->embed+i*n_embd), // embed
batch_mrope_pos, // pos batch_mrope_pos.data(), // pos
nullptr, // n_seq_id nullptr, // n_seq_id
nullptr, // seq_id nullptr, // seq_id
nullptr, // logits nullptr, // logits
@ -346,6 +350,8 @@ static void llava_free(struct llava_context * ctx_llava) {
llama_backend_free(); llama_backend_free();
} }
#ifndef NDEBUG
static void tmp_test_conv2d_reshape(struct llava_context * ctx_llava, gpt_params * params) { static void tmp_test_conv2d_reshape(struct llava_context * ctx_llava, gpt_params * params) {
int image_size_width = 256; int image_size_width = 256;
int image_size_height = 256; int image_size_height = 256;
@ -859,6 +865,7 @@ static llava_image_embed * tmp_load_img_embed() {
return result; return result;
} }
#endif
/* /*
----------------------------------------------------------------------------------------------------------------- -----------------------------------------------------------------------------------------------------------------
@ -902,6 +909,7 @@ int main(int argc, char ** argv) {
// This section is for testing LLM parts of the model during development phase! // This section is for testing LLM parts of the model during development phase!
auto ctx_llava = llava_init_context(&params, model); auto ctx_llava = llava_init_context(&params, model);
#ifndef NDEBUG
// { // {
// auto img_embed = tmp_load_img_embed(); // auto img_embed = tmp_load_img_embed();
// struct clip_image_size * load_image_size = clip_image_size_init(); // struct clip_image_size * load_image_size = clip_image_size_init();
@ -922,7 +930,8 @@ int main(int argc, char ** argv) {
// tmp_test_rope(ctx_llava, &params); // tmp_test_rope(ctx_llava, &params);
// tmp_test_mrope(ctx_llava, &params); // tmp_test_mrope(ctx_llava, &params);
// tmp_test_mrope_2d(ctx_llava, &params); // tmp_test_mrope_2d(ctx_llava, &params);
#endif
// process_prompt(ctx_llava, nullptr, &params, params.prompt); // process_prompt(ctx_llava, nullptr, &params, params.prompt);
llama_perf_context_print(ctx_llava->ctx_llama); llama_perf_context_print(ctx_llava->ctx_llama);