tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-14 20:29:41 -04:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into cisc/jina-embeddings-v3

Browse Source
This commit is contained in:
Sigbjørn Skjæret
2025-07-20 20:16:05 +02:00
committed by GitHub
parent e092776f9b 2be60cbc27
commit 80094033fa
154 changed files with 41058 additions and 3977 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
common/CMakeLists.txt
View File

@@ -86,8 +86,7 @@ if (LLAMA_CURL)
endif()
target_compile_definitions(${TARGET} PUBLIC LLAMA_USE_CURL)
include_directories(${CURL_INCLUDE_DIRS})
find_library(CURL_LIBRARY curl REQUIRED)
set(LLAMA_COMMON_EXTRA_LIBS ${LLAMA_COMMON_EXTRA_LIBS} ${CURL_LIBRARY})
set(LLAMA_COMMON_EXTRA_LIBS ${LLAMA_COMMON_EXTRA_LIBS} ${CURL_LIBRARIES})
endif ()
if (LLAMA_LLGUIDANCE)
@@ -112,13 +111,13 @@ if (LLAMA_LLGUIDANCE)
ExternalProject_Add(llguidance_ext
GIT_REPOSITORY https://github.com/guidance-ai/llguidance
# v0.7.20 (+ fix to build on GCC 15):
GIT_TAG b5b8b64dba11c4e4ee6b1d1450d3a3ae279891e8
# v1.0.1:
GIT_TAG d795912fedc7d393de740177ea9ea761e7905774
PREFIX ${CMAKE_BINARY_DIR}/llguidance
SOURCE_DIR ${LLGUIDANCE_SRC}
BUILD_IN_SOURCE TRUE
CONFIGURE_COMMAND ""
BUILD_COMMAND cargo build --release
BUILD_COMMAND cargo build --release --package llguidance
INSTALL_COMMAND ""
BUILD_BYPRODUCTS ${LLGUIDANCE_PATH}/${LLGUIDANCE_LIB_NAME} ${LLGUIDANCE_PATH}/llguidance.h
UPDATE_COMMAND ""

44
common/arg.cpp
View File

@@ -1464,6 +1464,14 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
params.swa_full = true;
}
).set_env("LLAMA_ARG_SWA_FULL"));
add_opt(common_arg(
{"--kv-unified", "-kvu"},
string_format("use single unified KV buffer for the KV cache of all sequences (default: %s)\n"
"[(more info)](https://github.com/ggml-org/llama.cpp/pull/14363)", params.kv_unified ? "true" : "false"),
[](common_params & params) {
params.kv_unified = true;
}
).set_env("LLAMA_ARG_KV_SPLIT"));
add_opt(common_arg(
{"--no-context-shift"},
string_format("disables context shift on infinite text generation (default: %s)", params.ctx_shift ? "disabled" : "enabled"),
@@ -2734,6 +2742,13 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
params.public_path = value;
}
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_STATIC_PATH"));
add_opt(common_arg(
{"--api-prefix"}, "PREFIX",
string_format("prefix path the server serves from, without the trailing slash (default: %s)", params.api_prefix.c_str()),
[](common_params & params, const std::string & value) {
params.api_prefix = value;
}
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_API_PREFIX"));
add_opt(common_arg(
{"--no-webui"},
string_format("Disable the Web UI (default: %s)", params.webui ? "enabled" : "disabled"),
@@ -3416,5 +3431,34 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
}
).set_examples({LLAMA_EXAMPLE_SERVER}));
// diffusion parameters
add_opt(common_arg(
{ "--diffusion-steps" }, "N",
string_format("number of diffusion steps (default: %d)", params.diffusion.steps),
[](common_params & params, int value) { params.diffusion.steps = value; }
).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
add_opt(common_arg(
{ "--diffusion-eps" }, "F",
string_format("epsilon for timesteps (default: %.6f)", (double) params.diffusion.eps),
[](common_params & params, const std::string & value) { params.diffusion.eps = std::stof(value); }
).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
add_opt(common_arg(
{ "--diffusion-algorithm" }, "N",
string_format("diffusion algorithm: 0=ORIGIN, 1=MASKGIT_PLUS, 2=TOPK_MARGIN, 3=ENTROPY (default: %d)",
params.diffusion.algorithm),
[](common_params & params, int value) { params.diffusion.algorithm = value; }
).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
add_opt(common_arg(
{ "--diffusion-alg-temp" }, "F",
string_format("algorithm temperature (default: %.3f)", (double) params.diffusion.alg_temp),
[](common_params & params, const std::string & value) { params.diffusion.alg_temp = std::stof(value); }
).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
add_opt(common_arg(
{ "--diffusion-visual" },
string_format("enable visual diffusion mode (show progressive generation) (default: %s)",
params.diffusion.visual_mode ? "true" : "false"),
[](common_params & params) { params.diffusion.visual_mode = true; }
).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
return ctx_arg;
}

28
common/common.cpp
View File

@@ -448,6 +448,15 @@ void string_replace_all(std::string & s, const std::string & search, const std::
bool string_ends_with(const std::string_view & str, const std::string_view & suffix) {
return str.size() >= suffix.size() && str.compare(str.size()-suffix.size(), suffix.size(), suffix) == 0;
}
bool string_remove_suffix(std::string & str, const std::string_view & suffix) {
bool has_suffix = string_ends_with(str, suffix);
if (has_suffix) {
str = str.substr(0, str.size() - suffix.size());
}
return has_suffix;
}
size_t string_find_partial_stop(const std::string_view & str, const std::string_view & stop) {
if (!str.empty() && !stop.empty()) {
const char text_last_char = str.back();
@@ -1011,15 +1020,21 @@ struct common_init_result common_init_from_params(common_params & params) {
params.sampling.ignore_eos = false;
}
if (params.sampling.ignore_eos) {
for (llama_token i = 0; i < llama_vocab_n_tokens(vocab); i++) {
if (llama_vocab_is_eog(vocab, i)) {
LOG_INF("%s: added %s logit bias = %f\n", __func__, common_token_to_piece(lctx, i).c_str(), -INFINITY);
params.sampling.logit_bias.push_back({i, -INFINITY});
}
// initialize once
for (llama_token i = 0; i < llama_vocab_n_tokens(vocab); i++) {
if (llama_vocab_is_eog(vocab, i)) {
LOG_INF("%s: added %s logit bias = %f\n", __func__, common_token_to_piece(lctx, i).c_str(), -INFINITY);
params.sampling.logit_bias_eog.push_back({i, -INFINITY});
}
}
if (params.sampling.ignore_eos) {
// add EOG biases to the active set of logit biases
params.sampling.logit_bias.insert(
params.sampling.logit_bias.end(),
params.sampling.logit_bias_eog.begin(), params.sampling.logit_bias_eog.end());
}
if (params.sampling.penalty_last_n == -1) {
LOG_INF("%s: setting penalty_last_n to ctx_size = %d\n", __func__, llama_n_ctx(lctx));
params.sampling.penalty_last_n = llama_n_ctx(lctx);
@@ -1163,6 +1178,7 @@ struct llama_context_params common_context_params_to_llama(const common_params &
cparams.no_perf = params.no_perf;
cparams.op_offload = !params.no_op_offload;
cparams.swa_full = params.swa_full;
cparams.kv_unified = params.kv_unified;
cparams.type_k = params.cache_type_k;
cparams.type_v = params.cache_type_v;

16
common/common.h
View File

@@ -84,6 +84,7 @@ enum llama_example {
LLAMA_EXAMPLE_LOOKUP,
LLAMA_EXAMPLE_PARALLEL,
LLAMA_EXAMPLE_TTS,
LLAMA_EXAMPLE_DIFFUSION,
LLAMA_EXAMPLE_COUNT,
};
@@ -180,7 +181,8 @@ struct common_params_sampling {
std::vector<common_grammar_trigger> grammar_triggers; // optional triggers (for lazy grammars)
std::set<llama_token> preserved_tokens;
std::vector<llama_logit_bias> logit_bias; // logit biases to apply
std::vector<llama_logit_bias> logit_bias; // logit biases to apply
std::vector<llama_logit_bias> logit_bias_eog; // pre-calculated logit biases for EOG tokens
// print the parameters into a string
std::string print() const;
@@ -220,6 +222,14 @@ struct common_params_vocoder {
bool use_guide_tokens = false; // enable guide tokens to improve TTS accuracy // NOLINT
};
struct common_params_diffusion {
int32_t steps = 64; // number of diffusion steps
float eps = 1e-3f; // epsilon for timesteps
int32_t algorithm = 0; // diffusion algorithm (0=ORIGIN, 1=MASKGIT_PLUS, 2=TOPK_MARGIN, 3=ENTROPY)
float alg_temp = 0.0f; // algorithm temperature
bool visual_mode = false; // show progressive diffusion on screen
};
enum common_reasoning_format {
COMMON_REASONING_FORMAT_NONE,
COMMON_REASONING_FORMAT_DEEPSEEK_LEGACY, // Extract thinking tag contents and return as `message.reasoning_content`, or leave inline in <think> tags in stream mode
@@ -271,6 +281,7 @@ struct common_params {
struct common_params_sampling sampling;
struct common_params_speculative speculative;
struct common_params_vocoder vocoder;
struct common_params_diffusion diffusion;
struct common_params_model model;
@@ -333,6 +344,7 @@ struct common_params {
bool no_perf = false; // disable performance metrics
bool ctx_shift = true; // context shift on inifinite text generation
bool swa_full = false; // use full-size SWA cache (https://github.com/ggml-org/llama.cpp/pull/13194#issuecomment-2868343055)
bool kv_unified = false; // enable unified KV cache
bool input_prefix_bos = false; // prefix BOS to user inputs, preceding input_prefix
bool use_mmap = true; // use mmap for faster loads
@@ -373,6 +385,7 @@ struct common_params {
std::string hostname = "127.0.0.1";
std::string public_path = ""; // NOLINT
std::string api_prefix = ""; // NOLINT
std::string chat_template = ""; // NOLINT
bool use_jinja = false; // NOLINT
bool enable_chat_template = true;
@@ -524,6 +537,7 @@ static bool string_starts_with(const std::string & str,
// While we wait for C++20's std::string::ends_with...
bool string_ends_with(const std::string_view & str, const std::string_view & suffix);
bool string_remove_suffix(std::string & str, const std::string_view & suffix);
size_t string_find_partial_stop(const std::string_view & str, const std::string_view & stop);
bool string_parse_kv_override(const char * data, std::vector<llama_model_kv_override> & overrides);