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

Merge branch 'master' into gg/flash-attn

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2024-04-24 14:00:32 +03:00
commit 8937ec5307
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
57 changed files with 1576 additions and 8245 deletions
Download patch file Download diff file Expand all files Collapse all files

18
llama.h
View file

@ -784,6 +784,9 @@ extern "C" {
LLAMA_API enum llama_token_type llama_token_get_type(const struct llama_model * model, llama_token token);
// Check if the token is supposed to end generation (end-of-generation, eg. EOS, EOT, etc.)
LLAMA_API bool llama_token_is_eog(const struct llama_model * model, llama_token token);
// Special tokens
LLAMA_API llama_token llama_token_bos(const struct llama_model * model); // beginning-of-sentence
LLAMA_API llama_token llama_token_eos(const struct llama_model * model); // end-of-sentence
@ -797,7 +800,7 @@ extern "C" {
// Returns -1 if unknown, 1 for true or 0 for false.
LLAMA_API int32_t llama_add_eos_token(const struct llama_model * model);
// codellama infill tokens
// Codellama infill tokens
LLAMA_API llama_token llama_token_prefix(const struct llama_model * model); // Beginning of infill prefix
LLAMA_API llama_token llama_token_middle(const struct llama_model * model); // Beginning of infill middle
LLAMA_API llama_token llama_token_suffix(const struct llama_model * model); // Beginning of infill suffix
@ -826,11 +829,13 @@ extern "C" {
// Uses the vocabulary in the provided context.
// Does not write null terminator to the buffer.
// User code is responsible to remove the leading whitespace of the first non-BOS token when decoding multiple tokens.
// @param special If true, special tokens are rendered in the output.
LLAMA_API int32_t llama_token_to_piece(
const struct llama_model * model,
llama_token token,
char * buf,
int32_t length);
int32_t length,
bool special);
/// Apply chat template. Inspired by hf apply_chat_template() on python.
/// Both "model" and "custom_template" are optional, but at least one is required. "custom_template" has higher precedence than "model"
@ -983,7 +988,7 @@ extern "C" {
struct llama_context * ctx,
llama_token_data_array * candidates);
/// @details Randomly selects a token from the candidates based on their probabilities.
/// @details Randomly selects a token from the candidates based on their probabilities using the RNG of ctx.
LLAMA_API llama_token llama_sample_token(
struct llama_context * ctx,
llama_token_data_array * candidates);
@ -1070,8 +1075,9 @@ extern "C" {
// Internal API to be implemented by llama.cpp and used by tests/benchmarks only
#ifdef LLAMA_API_INTERNAL
#include <vector>
#include <random>
#include <string>
#include <vector>
struct ggml_tensor;
@ -1108,6 +1114,10 @@ std::pair<std::vector<uint32_t>, llama_partial_utf8> decode_utf8(
const std::string & src,
llama_partial_utf8 partial_start);
// Randomly selects a token from the candidates based on their probabilities using given std::mt19937.
// This is a temporary workaround in order to fix race conditions when sampling with multiple sequences.
llama_token llama_sample_token_with_rng(struct llama_context * ctx, llama_token_data_array * candidates, std::mt19937 & rng);
#endif // LLAMA_API_INTERNAL
#endif // LLAMA_H