vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-11 09:34:37 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'master' into concedo_experimental

Browse source 
# Conflicts:
#	.github/workflows/build.yml
#	Makefile
#	README.md
This commit is contained in:
Concedo 2023-09-18 23:20:02 +08:00
commit 0142760fc3
8 changed files with 129 additions and 62 deletions
Download patch file Download diff file Expand all files Collapse all files

18
llama.cpp
View file

@ -932,6 +932,7 @@ enum e_model {
static const size_t kB = 1024;
static const size_t MB = kB*kB;
static const size_t GB = kB*kB*kB;
// default hparams (LLaMA 7B)
struct llama_hparams {
@ -1285,6 +1286,7 @@ struct llama_model_loader {
int n_created = 0;
int64_t n_elements = 0;
size_t n_bytes = 0;
bool use_mmap = false;
@ -1317,6 +1319,7 @@ struct llama_model_loader {
const char * name = gguf_get_tensor_name(ctx_gguf, i);
struct ggml_tensor * t = ggml_get_tensor(ctx_meta, name);
n_elements += ggml_nelements(t);
n_bytes += ggml_nbytes(t);
}
LLAMA_LOG_INFO("%s: loaded meta data with %d key-value pairs and %d tensors from %s (version %s)\n",
@ -1915,7 +1918,12 @@ static void llm_load_print_meta(llama_model_loader & ml, llama_model & model) {
LLAMA_LOG_INFO("%s: freq_scale = %g\n", __func__, hparams.rope_freq_scale);
LLAMA_LOG_INFO("%s: model type = %s\n", __func__, llama_model_type_name(model.type));
LLAMA_LOG_INFO("%s: model ftype = %s\n", __func__, llama_model_ftype_name(model.ftype).c_str());
LLAMA_LOG_INFO("%s: model size = %.2f B\n", __func__, ml.n_elements*1e-9);
LLAMA_LOG_INFO("%s: model params = %.2f B\n", __func__, ml.n_elements*1e-9);
if (ml.n_bytes < GB) {
LLAMA_LOG_INFO("%s: model size = %.2f MiB (%.2f BPW) \n", __func__, ml.n_bytes/1024.0/1024.0, ml.n_bytes*8.0/ml.n_elements);
} else {
LLAMA_LOG_INFO("%s: model size = %.2f GiB (%.2f BPW) \n", __func__, ml.n_bytes/1024.0/1024.0/1024.0, ml.n_bytes*8.0/ml.n_elements);
}
// general kv
LLAMA_LOG_INFO("%s: general.name = %s\n", __func__, model.name.c_str());
@ -3505,7 +3513,7 @@ static struct ggml_cgraph * llm_build_starcoder(
ggml_allocr_alloc(lctx.alloc, token);
if (!ggml_allocr_is_measure(lctx.alloc)) {
memcpy(token->data, embd, N * n_embd * ggml_element_size(inpL));
memcpy(token->data, embd, N * n_embd * ggml_element_size(token));
}
}
@ -7045,19 +7053,21 @@ llama_token llama_token_nl(const struct llama_context * ctx) {
int llama_tokenize(
struct llama_context * ctx,
const char * text,
int text_len,
llama_token * tokens,
int n_max_tokens,
bool add_bos) {
return llama_tokenize_with_model(&ctx->model, text, tokens, n_max_tokens, add_bos);
return llama_tokenize_with_model(&ctx->model, text, text_len, tokens, n_max_tokens, add_bos);
}
int llama_tokenize_with_model(
const struct llama_model * model,
const char * text,
int text_len,
llama_token * tokens,
int n_max_tokens,
bool add_bos) {
auto res = llama_tokenize_internal(model->vocab, text, add_bos);
auto res = llama_tokenize_internal(model->vocab, std::string(text, text_len), add_bos);
if (n_max_tokens < (int) res.size()) {
// LLAMA_LOG_ERROR("%s: too many tokens\n", __func__);