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

llama : minor simplifications

Browse source
This commit is contained in:
Georgi Gerganov 2023-08-15 22:41:55 +03:00
parent f477fb069b
commit 23248d7d32
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
1 changed files with 11 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

18
gguf-llama.cpp
View file

@ -1098,11 +1098,11 @@ struct llama_model_loader {
this->use_mmap = use_mmap; this->use_mmap = use_mmap;
} }
void calc_sizes(size_t * ctx_size_p, size_t * mmapped_size_p) const { void calc_sizes(size_t & ctx_size_p, size_t & mmapped_size_p) const {
*ctx_size_p = *mmapped_size_p = 0; ctx_size_p = mmapped_size_p = 0;
for (const llama_load_tensor & lt : tensors_map.tensors) { for (const llama_load_tensor & lt : tensors_map.tensors) {
*ctx_size_p += sizeof(struct ggml_tensor) + GGML_OBJECT_SIZE; ctx_size_p += sizeof(struct ggml_tensor) + GGML_OBJECT_SIZE;
*(use_mmap ? mmapped_size_p : ctx_size_p) += lt.size + 16; (use_mmap ? mmapped_size_p : ctx_size_p) += ggml_nbytes_pad(lt.ggml_tensor);
} }
} }
@ -1160,18 +1160,18 @@ struct llama_model_loader {
void load_all_data(llama_progress_callback progress_callback, void * progress_callback_user_data, llama_mlock * lmlock) { void load_all_data(llama_progress_callback progress_callback, void * progress_callback_user_data, llama_mlock * lmlock) {
size_t data_size = 0; size_t data_size = 0;
size_t prefetch_size = 0;
size_t lock_size = 0; size_t lock_size = 0;
size_t pref_size = 0; // prefetch
for (const llama_load_tensor & lt : tensors_map.tensors) { for (const llama_load_tensor & lt : tensors_map.tensors) {
data_size += lt.size; data_size += lt.size;
if (lt.ggml_tensor->backend == GGML_BACKEND_CPU) { if (lt.ggml_tensor->backend == GGML_BACKEND_CPU) {
prefetch_size += lt.size; pref_size += lt.size;
} }
} }
if (use_mmap) { if (use_mmap) {
mapping.reset(new llama_mmap(&file_loader->file, prefetch_size, ggml_is_numa())); mapping.reset(new llama_mmap(&file_loader->file, pref_size, ggml_is_numa()));
if (lmlock) { if (lmlock) {
lmlock->init(mapping->addr); lmlock->init(mapping->addr);
} }
@ -1404,7 +1404,7 @@ static void llama_model_load_internal(
size_t ctx_size; size_t ctx_size;
size_t mmapped_size; size_t mmapped_size;
ml->calc_sizes(&ctx_size, &mmapped_size); ml->calc_sizes(ctx_size, mmapped_size);
LLAMA_LOG_INFO("%s: ggml ctx size = %7.2f MB\n", __func__, ctx_size/1024.0/1024.0); LLAMA_LOG_INFO("%s: ggml ctx size = %7.2f MB\n", __func__, ctx_size/1024.0/1024.0);
// create the ggml context // create the ggml context
@ -3688,7 +3688,7 @@ int llama_apply_lora_from_file_internal(const struct llama_model & model, const
size_t ctx_size; size_t ctx_size;
size_t mmapped_size; size_t mmapped_size;
model_loader->calc_sizes(&ctx_size, &mmapped_size); model_loader->calc_sizes(ctx_size, mmapped_size);
base_buf.resize(ctx_size); base_buf.resize(ctx_size);
ggml_init_params base_params; ggml_init_params base_params;