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

gpt j use scratch buffers

Browse source
This commit is contained in:
Concedo 2023-06-21 16:10:31 +08:00
parent 266d47a4b9
commit dfdd20240c
1 changed files with 22 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

28
otherarch/gptj_v3.cpp
View file

@ -150,7 +150,7 @@ ModelLoadResult gptj_model_load(const std::string & fname, gptj_model & model, g
params.mem_size = ctx_size; params.mem_size = ctx_size;
params.mem_buffer = NULL; params.mem_buffer = NULL;
params.no_alloc = false; params.no_alloc = false;
model.ctx = ggml_init(params); model.ctx = ggml_init(params);
if (!model.ctx) { if (!model.ctx) {
@ -281,7 +281,7 @@ ModelLoadResult gptj_model_load(const std::string & fname, gptj_model & model, g
fprintf(stderr, "%s: tensor '%s' has wrong size in model file\n", __func__, name.data()); fprintf(stderr, "%s: tensor '%s' has wrong size in model file\n", __func__, name.data());
return ModelLoadResult::FAIL; return ModelLoadResult::FAIL;
} }
if (tensor->ne[0] != ne[0] || tensor->ne[1] != ne[1]) { if (tensor->ne[0] != ne[0] || tensor->ne[1] != ne[1]) {
@ -298,7 +298,7 @@ ModelLoadResult gptj_model_load(const std::string & fname, gptj_model & model, g
__func__, name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]); __func__, name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]);
return ModelLoadResult::FAIL; return ModelLoadResult::FAIL;
} }
} }
// for debugging // for debugging
@ -367,8 +367,16 @@ bool gptj_eval(
static size_t buf_size = 256u*1024*1024; static size_t buf_size = 256u*1024*1024;
static void * buf = malloc(buf_size); static void * buf = malloc(buf_size);
if (mem_per_token > 0 && (mem_per_token*N*2 + 64u*1024*1024) > buf_size) { // use 2 scratch buffers
const size_t buf_size_new = 320u*1024*1024 + 1.6*(mem_per_token*N); // add 10% to account for ggml object overhead // TODO: very hacky solution - reimplement in a more elegant way
static size_t scr0_size = (n_ctx>1024?512u:256u)*1024*1024;
static void * scr0 = malloc(scr0_size);
static size_t scr1_size = (n_ctx>1024?512u:256u)*1024*1024;
static void * scr1 = malloc(scr1_size);
if (mem_per_token > 0 && mem_per_token*N*1.05 > buf_size) {
const size_t buf_size_new = 64u*1024*1024 + 1.15*(mem_per_token*N); // add 10% to account for ggml object overhead
//printf("\n%s: reallocating buffer from %zu to %zu bytes\n", __func__, buf_size, buf_size_new); //printf("\n%s: reallocating buffer from %zu to %zu bytes\n", __func__, buf_size, buf_size_new);
// reallocate // reallocate
@ -388,7 +396,7 @@ bool gptj_eval(
params.mem_size = buf_size; params.mem_size = buf_size;
params.mem_buffer = buf; params.mem_buffer = buf;
params.no_alloc = false; params.no_alloc = false;
struct ggml_context * ctx0 = ggml_init(params); struct ggml_context * ctx0 = ggml_init(params);
struct ggml_cgraph gf = {}; struct ggml_cgraph gf = {};
@ -403,6 +411,8 @@ bool gptj_eval(
for (int il = 0; il < n_layer; ++il) { for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * cur; struct ggml_tensor * cur;
ggml_set_scratch(ctx0, { 0, scr0_size, scr0, });
// norm // norm
{ {
cur = ggml_norm(ctx0, inpL); cur = ggml_norm(ctx0, inpL);
@ -490,6 +500,8 @@ bool gptj_eval(
cur); cur);
} }
ggml_set_scratch(ctx0, { 0, scr1_size, scr1, });
struct ggml_tensor * inpFF = cur; struct ggml_tensor * inpFF = cur;
// feed-forward network // feed-forward network
@ -525,6 +537,8 @@ bool gptj_eval(
inpL = ggml_add(ctx0, cur, inpL); inpL = ggml_add(ctx0, cur, inpL);
} }
ggml_set_scratch(ctx0, { 0, scr0_size, scr0, });
// norm // norm
{ {
inpL = ggml_norm(ctx0, inpL); inpL = ggml_norm(ctx0, inpL);
@ -537,6 +551,8 @@ bool gptj_eval(
ggml_repeat(ctx0, model.ln_f_b, inpL)); ggml_repeat(ctx0, model.ln_f_b, inpL));
} }
ggml_set_scratch(ctx0, { 0, 0, nullptr, });
// lm_head // lm_head
{ {
inpL = ggml_mul_mat(ctx0, model.lmh_g, inpL); inpL = ggml_mul_mat(ctx0, model.lmh_g, inpL);