ExternalVendorCode/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2024-12-20 05:07:52 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

talk-llama : sync llama.cpp

Browse Source
This commit is contained in:
Georgi Gerganov 2024-01-14 18:08:20 +02:00
parent 2a5874441d
commit 6ebba525f1
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 20 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
examples/talk-llama/llama.cpp
View File

@ -2190,6 +2190,11 @@ struct llama_model_loader {
LLM_KV llm_kv = LLM_KV(LLM_ARCH_UNKNOWN);
llama_model_loader(const std::string & fname, bool use_mmap, const struct llama_model_kv_override * param_overrides_p) : file(fname.c_str(), "rb") {
int trace = 0;
if (getenv("LLAMA_TRACE")) {
trace = atoi(getenv("LLAMA_TRACE"));
}
struct gguf_init_params params = {
/*.no_alloc = */ true,
/*.ctx = */ &ctx_meta,
@ -2242,11 +2247,10 @@ struct llama_model_loader {
type_max = type;
}
// TODO: make runtime configurable
#if 0
if (trace > 0) {
struct ggml_tensor * meta = ggml_get_tensor(ctx_meta, gguf_get_tensor_name(ctx_gguf, i));
LLAMA_LOG_INFO("%s: - tensor %4d: %32s %-8s [ %s ]\n", __func__, i, ggml_get_name(meta), ggml_type_name(type), llama_format_tensor_shape(meta).c_str());
#endif
}
}
switch (type_max) {
@ -7095,7 +7099,7 @@ static std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab &
}
#ifdef PRETOKENIZERDEBUG
LLAMA_LOG_WARN(TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
LLAMA_LOG_WARN("TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
#endif
llm_tokenizer_spm tokenizer(vocab);
llama_escape_whitespace(raw_text);
@ -7116,7 +7120,7 @@ static std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab &
auto raw_text = fragment.raw_text.substr(fragment.offset, fragment.length);
#ifdef PRETOKENIZERDEBUG
LLAMA_LOG_WARN(TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
LLAMA_LOG_WARN("TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
#endif
llm_tokenizer_bpe tokenizer(vocab);
tokenizer.tokenize(raw_text, output);