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

Apply suggestions from code review

Browse source
This commit is contained in:
Georgi Gerganov 2023-03-15 21:35:48 +02:00 • committed by GitHub
parent 9eb4598fa3
commit f056beb384
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 7 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

6
main.cpp
View file

@ -218,7 +218,7 @@ bool llama_model_load(const std::string & fname, llama_model & model, gpt_vocab
ctx_size += n_ctx*n_layer*n_embd*ggml_type_sizef(GGML_TYPE_F32); // memory_k
ctx_size += n_ctx*n_layer*n_embd*ggml_type_sizef(GGML_TYPE_F32); // memory_v
ctx_size += (5 + 10*n_layer)*hparams.n_ctx; // object overhead
ctx_size += (5 + 10*n_layer)*256; // object overhead
fprintf(stderr, "%s: ggml ctx size = %6.2f MB\n", __func__, ctx_size/(1024.0*1024.0));
}
@ -547,7 +547,9 @@ bool llama_eval(
const int d_key = n_embd/n_head;
static size_t buf_size = hparams.n_ctx*1024*1024;
// TODO: check if this size scales with n_ctx linearly and remove constant. somehow I feel it wasn't the case
// static size_t buf_size = hparams.n_ctx*1024*1024;
static size_t buf_size = 512u*1024*1024;
static void * buf = malloc(buf_size);
if (mem_per_token > 0 && mem_per_token*N > buf_size) {

1
utils.h
View file

@ -18,6 +18,7 @@ struct gpt_params {
int32_t n_predict = 128; // new tokens to predict
int32_t repeat_last_n = 64; // last n tokens to penalize
int32_t n_ctx = 512; //context size
// sampling parameters
int32_t top_k = 40;
float top_p = 0.95f;