Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-20 01:31:42 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge commit '2948e6049a4ad0f96a4ab15246db2d2086b80703' into concedo_experimental

Browse Source 
# Conflicts:
#	.github/workflows/build.yml
#	CONTRIBUTING.md
#	docs/backend/VirtGPU/development.md
#	docs/ops.md
#	docs/ops/WebGPU.csv
#	embd_res/templates/GigaChat3-10B-A1.8B.jinja
#	embd_res/templates/GigaChat3.1-10B-A1.8B.jinja
#	ggml/src/ggml-hip/CMakeLists.txt
#	ggml/src/ggml-opencl/CMakeLists.txt
#	ggml/src/ggml-opencl/ggml-opencl.cpp
#	ggml/src/ggml-webgpu/ggml-webgpu-shader-lib.hpp
#	ggml/src/ggml-webgpu/ggml-webgpu.cpp
#	scripts/sync_vendor.py
#	tests/CMakeLists.txt
#	tests/test-backend-ops.cpp
#	tests/test-chat.cpp
#	tests/test-grammar-integration.cpp
#	tests/test-quantize-fns.cpp
This commit is contained in:
Concedo
2026-03-15 11:21:24 +08:00
parent 67c9798d0b 2948e6049a
commit b1c500ae2b
72 changed files with 2338 additions and 430 deletions
Download Patch File Download Diff File Expand all files Collapse all files
common/arg.cpp
+11 -5
View File
@@ -735,23 +735,28 @@ static void common_params_print_completion(common_params_context & ctx_arg) {
"llama-completion",
"llama-convert-llama2c-to-ggml",
"llama-cvector-generator",
"llama-debug",
"llama-diffusion-cli",
"llama-embedding",
"llama-eval-callback",
"llama-export-lora",
"llama-finetune",
"llama-fit-params",
"llama-gemma3-cli",
"llama-gen-docs",
"llama-gguf",
"llama-gguf-hash",
"llama-gguf-split",
"llama-gritlm",
"llama-idle",
"llama-imatrix",
"llama-infill",
"llama-mtmd-cli",
"llama-llava-clip-quantize-cli",
"llama-llava-cli",
"llama-lookahead",
"llama-lookup",
"llama-lookup-create",
"llama-lookup-merge",
"llama-lookup-stats",
"llama-minicpmv-cli",
"llama-mtmd-cli",
"llama-parallel",
"llama-passkey",
"llama-perplexity",
@@ -2669,7 +2674,8 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
[](common_params & params, const std::string & value) {
params.out_file = value;
}
).set_examples({LLAMA_EXAMPLE_IMATRIX, LLAMA_EXAMPLE_CVECTOR_GENERATOR, LLAMA_EXAMPLE_EXPORT_LORA, LLAMA_EXAMPLE_TTS, LLAMA_EXAMPLE_FINETUNE, LLAMA_EXAMPLE_RESULTS}));
).set_examples({LLAMA_EXAMPLE_IMATRIX, LLAMA_EXAMPLE_CVECTOR_GENERATOR, LLAMA_EXAMPLE_EXPORT_LORA, LLAMA_EXAMPLE_TTS, LLAMA_EXAMPLE_FINETUNE,
LLAMA_EXAMPLE_RESULTS, LLAMA_EXAMPLE_EXPORT_GRAPH_OPS}));
add_opt(common_arg(
{"-ofreq", "--output-frequency"}, "N",
string_format("output the imatrix every N iterations (default: %d)", params.n_out_freq),
common/chat.cpp
+80
View File
@@ -1369,6 +1369,77 @@ static common_chat_params common_chat_params_init_lfm2(const common_chat_templat
return data;
}
static common_chat_params common_chat_params_init_gigachat_v3(
const common_chat_template & tmpl,
const autoparser::templates_params & inputs) {
common_chat_params data;
data.prompt = common_chat_template_direct_apply(tmpl, inputs);
data.format = COMMON_CHAT_FORMAT_PEG_NATIVE;
data.supports_thinking = false;
data.preserved_tokens = {
"<|message_sep|>\n\n",
"<|role_sep|>\n",
};
auto has_tools = inputs.tools.is_array() && !inputs.tools.empty();
auto include_grammar = has_tools && inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE;
auto tool_call_start_prefix = "<|message_sep|>\n\nfunction call<|role_sep|>\n";
auto parser = build_chat_peg_parser([&](common_chat_peg_builder & p) {
if (has_tools && inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE) {
// Build a choice of all available tools
auto tool_choice = p.choice();
for (const auto & tool : inputs.tools) {
const auto & function = tool.at("function");
std::string name = function.at("name");
const auto & schema = function.at("parameters");
auto tool_name = p.json_member("name", "\"" + p.tool_name(p.literal(name)) + "\"");
auto tool_args = p.json_member("arguments", p.tool_args(p.schema(p.json(), "tool-" + name + "-schema", schema)));
auto tool_open = p.tool_open(p.literal("{") << tool_name);
tool_choice |= p.rule("tool-" + name, tool_open << "," << tool_args << "}");
}
// Define the tool call structure
auto min_calls = inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_REQUIRED ? 1 : 0;
auto max_calls = 1; // parallel toolcalls are not supported
auto tool_call = p.rule("tool-call", p.literal(tool_call_start_prefix) + tool_choice);
auto tool_calls = p.trigger_rule("tool-call-root", p.repeat(tool_call, /* min = */ min_calls, /* max = */ max_calls));
return p.content(p.until("<|message_sep|>\n\n")) << tool_calls;
}
// Content only parser
include_grammar = false;
return p.content(p.rest());
});
data.parser = parser.save();
if (include_grammar) {
data.grammar_lazy = has_tools && inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_AUTO;
data.grammar = build_grammar([&](const common_grammar_builder & builder) {
foreach_function(inputs.tools, [&](const json & tool) {
const auto & function = tool.at("function");
auto schema = function.at("parameters");
builder.resolve_refs(schema);
});
parser.build_grammar(builder, data.grammar_lazy);
});
data.grammar_triggers = {
{COMMON_GRAMMAR_TRIGGER_TYPE_WORD, tool_call_start_prefix}
};
}
return data;
}
namespace workaround {
static void map_developer_role_to_system(json & messages) {
@@ -1540,6 +1611,15 @@ static common_chat_params common_chat_templates_apply_jinja(const struct common_
return common_chat_params_init_lfm2(tmpl, params);
}
// GigaChatV3 format detection
if (src.find("<|role_sep|>") != std::string::npos &&
src.find("<|message_sep|>") != std::string::npos &&
src.find("<|function_call|>") == std::string::npos
) {
LOG_DBG("Using specialized template: GigaChatV3\n");
return common_chat_params_init_gigachat_v3(tmpl, params);
}
try {
LOG_DBG("Using differential autoparser\n");
struct autoparser::autoparser autoparser;
common/common.h
+2 -1
View File
@@ -102,6 +102,7 @@ enum llama_example {
LLAMA_EXAMPLE_FINETUNE,
LLAMA_EXAMPLE_FIT_PARAMS,
LLAMA_EXAMPLE_RESULTS,
LLAMA_EXAMPLE_EXPORT_GRAPH_OPS,
LLAMA_EXAMPLE_COUNT,
};
@@ -923,7 +924,7 @@ const char * const LLM_KV_SPLIT_TENSORS_COUNT = "split.tensors.count";
// MoE utils
//
const char * const LLM_FFN_EXPS_REGEX = "\\.ffn_(up|down|gate)_(ch|)exps";
const char * const LLM_FFN_EXPS_REGEX = "\\.ffn_(up|down|gate|gate_up)_(ch|)exps";
inline std::string llm_ffn_exps_block_regex(int idx) {
return string_format("blk\\.%d%s", idx, LLM_FFN_EXPS_REGEX);