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

Merge branch 'json-fixes-cpp' into json-fixes

Browse source
This commit is contained in:
ochafik 2024-03-12 03:48:38 +00:00
commit 7e1440cc01
10 changed files with 1060 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

2
CMakeLists.txt
View file

@ -135,7 +135,7 @@ include(${CMAKE_CURRENT_SOURCE_DIR}/scripts/build-info.cmake)
if (LLAMA_SYCL)
set(CMAKE_CXX_STANDARD 17)
else()
set(CMAKE_CXX_STANDARD 11)
set(CMAKE_CXX_STANDARD 17)
endif()
set(CMAKE_CXX_STANDARD_REQUIRED true)

16
Makefile
View file

@ -9,7 +9,8 @@ TEST_TARGETS = \
tests/test-llama-grammar tests/test-grammar-parser tests/test-double-float tests/test-grad0 tests/test-opt \
tests/test-quantize-fns tests/test-quantize-perf tests/test-sampling tests/test-tokenizer-0-llama \
tests/test-tokenizer-0-falcon tests/test-tokenizer-1-llama tests/test-tokenizer-1-bpe tests/test-rope \
tests/test-backend-ops tests/test-model-load-cancel tests/test-autorelease
tests/test-backend-ops tests/test-model-load-cancel tests/test-autorelease \
tests/test-json-schema-to-grammar
# Code coverage output files
COV_TARGETS = *.gcno tests/*.gcno *.gcda tests/*.gcda *.gcov tests/*.gcov lcov-report gcovr-report
@ -99,7 +100,7 @@ endif
# keep standard at C11 and C++11
MK_CPPFLAGS = -I. -Icommon
MK_CFLAGS = -std=c11 -fPIC
MK_CXXFLAGS = -std=c++11 -fPIC
MK_CXXFLAGS = -std=c++17 -fPIC
MK_NVCCFLAGS = -std=c++11
# -Ofast tends to produce faster code, but may not be available for some compilers.
@ -653,6 +654,11 @@ console.o: common/console.cpp common/console.h
grammar-parser.o: common/grammar-parser.cpp common/grammar-parser.h
$(CXX) $(CXXFLAGS) -c $< -o $@
json-schema-to-grammar.o: examples/server/json-schema-to-grammar.cpp examples/server/json-schema-to-grammar.h
$(CXX) $(CXXFLAGS) -c $< -o $@
# $(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<) -DLLAMA_BUILD_JSON_SCHEMA_CONVERTER=1
# $(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
train.o: common/train.cpp common/train.h
$(CXX) $(CXXFLAGS) -c $< -o $@
@ -728,7 +734,7 @@ save-load-state: examples/save-load-state/save-load-state.cpp ggml.o llama.o $(C
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
server: examples/server/server.cpp examples/server/utils.hpp examples/server/httplib.h examples/server/json.hpp examples/server/index.html.hpp examples/server/index.js.hpp examples/server/completion.js.hpp common/stb_image.h ggml.o llama.o $(COMMON_DEPS) grammar-parser.o $(OBJS)
server: examples/server/server.cpp examples/server/utils.hpp examples/server/httplib.h examples/server/json.hpp examples/server/index.html.hpp examples/server/index.js.hpp examples/server/completion.js.hpp json-schema-to-grammar.o common/stb_image.h ggml.o llama.o $(COMMON_DEPS) grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h %.hpp $<,$^) -Iexamples/server $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS) $(LWINSOCK2)
@ -844,6 +850,10 @@ tests/test-double-float: tests/test-double-float.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-json-schema-to-grammar: tests/test-json-schema-to-grammar.cpp json-schema-to-grammar.o
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-grad0: tests/test-grad0.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)

9
examples/server/CMakeLists.txt
View file

@ -2,7 +2,14 @@ set(TARGET server)
option(LLAMA_SERVER_VERBOSE "Build verbose logging option for Server" ON)
option(LLAMA_SERVER_SSL "Build SSL support for the server" OFF)
include_directories(${CMAKE_CURRENT_SOURCE_DIR})
add_executable(${TARGET} server.cpp utils.hpp json.hpp httplib.h)
add_executable(${TARGET}
server.cpp
utils.hpp
json.hpp
httplib.h
json-schema-to-grammar.cpp
json-schema-to-grammar.h
)
install(TARGETS ${TARGET} RUNTIME)
target_compile_definitions(${TARGET} PRIVATE
SERVER_VERBOSE=$<BOOL:${LLAMA_SERVER_VERBOSE}>

698
examples/server/json-schema-to-grammar.cpp Normal file
View file

@ -0,0 +1,698 @@
#include "json-schema-to-grammar.h"
#include <algorithm>
#include <fstream>
#include <iostream>
#include <map>
#include <regex>
#include <sstream>
#include <string>
#include <unordered_map>
#include <unordered_set>
#include <vector>
using json = nlohmann::json;
using namespace std;
const string SPACE_RULE = "\" \"?";
unordered_map<string, string> PRIMITIVE_RULES = {
{"boolean", "(\"true\" | \"false\") space"},
{"number", "(\"-\"? ([0-9] | [1-9] [0-9]*)) (\".\" [0-9]+)? ([eE] [-+]? [0-9]+)? space"},
{"integer", "(\"-\"? ([0-9] | [1-9] [0-9]*)) space"},
{"value", "object | array | string | number | boolean"},
{"object", "\"{\" space ( string \":\" space value (\",\" space string \":\" space value)* )? \"}\" space"},
{"array", "\"[\" space ( value (\",\" space value)* )? \"]\" space"},
{"uuid", "\"\\\"\" [0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F] "
"\"-\" [0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F] "
"\"-\" [0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F] "
"\"-\" [0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F] "
"\"-\" [0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F] \"\\\"\" space"},
{"string", " \"\\\"\" (\n"
" [^\"\\\\] |\n"
" \"\\\\\" ([\"\\\\/bfnrt] | \"u\" [0-9a-fA-F] [0-9a-fA-F] [0-9a-fA-F] [0-9a-fA-F])\n"
" )* \"\\\"\" space"},
{"null", "\"null\" space"}
};
unordered_map<string, string> DATE_RULES = {
{"date", "[0-9] [0-9] [0-9] [0-9] \"-\" ( \"0\" [1-9] | \"1\" [0-2] ) \"-\" ( [0-2] [0-9] | \"3\" [0-1] )"},
{"time", "([01] [0-9] | \"2\" [0-3]) \":\" [0-5] [0-9] \":\" [0-5] [0-9] ( \".\" [0-9] [0-9] [0-9] )? ( \"Z\" | ( \"+\" | \"-\" ) ( [01] [0-9] | \"2\" [0-3] ) \":\" [0-5] [0-9] )"},
{"date-time", "date \"T\" time"},
{"date-string", "\"\\\"\" date \"\\\"\" space"},
{"time-string", "\"\\\"\" time \"\\\"\" space"},
{"date-time-string", "\"\\\"\" date-time \"\\\"\" space"}
};
regex INVALID_RULE_CHARS_RE("[^a-zA-Z0-9-]+");
regex GRAMMAR_LITERAL_ESCAPE_RE("[\r\n\"]");
regex GRAMMAR_RANGE_LITERAL_ESCAPE_RE("[\r\n\"\\]\\-\\\\]");
unordered_map<char, string> GRAMMAR_LITERAL_ESCAPES = {
{'\r', "\\r"}, {'\n', "\\n"}, {'"', "\\\""}, {'-', "\\-"}, {']', "\\]"}
};
unordered_set<char> NON_LITERAL_SET = {'|', '.', '(', ')', '[', ']', '{', '}', '*', '+', '?'};
unordered_set<char> ESCAPED_IN_REGEXPS_BUT_NOT_IN_LITERALS = {'{', '*', '+', '?'};
template <typename Iterator>
string join(Iterator begin, Iterator end, const string& separator) {
ostringstream result;
if (begin != end) {
result << *begin;
for (Iterator it = begin + 1; it != end; ++it) {
result << separator << *it;
}
}
return result.str();
}
static std::vector<std::string> split(const std::string& str, const std::string& delimiter) {
std::vector<std::string> tokens;
size_t start = 0;
size_t end = str.find(delimiter);
while (end != std::string::npos) {
tokens.push_back(str.substr(start, end - start));
start = end + delimiter.length();
end = str.find(delimiter, start);
}
tokens.push_back(str.substr(start));
return tokens;
}
static string repeat(const string& str, size_t n) {
if (n == 0) {
return "";
}
string result;
result.reserve(str.length() * n);
for (size_t i = 0; i < n; ++i) {
result += str;
}
return result;
}
static std::string replacePattern(const std::string& input, const regex& regex, const function<string(const smatch &)>& replacement) {
std::smatch match;
std::string result;
std::string::const_iterator searchStart(input.cbegin());
std::string::const_iterator searchEnd(input.cend());
while (std::regex_search(searchStart, searchEnd, match, regex)) {
result.append(searchStart, searchStart + match.position());
result.append(replacement(match));
searchStart = match.suffix().first;
}
result.append(searchStart, searchEnd);
return result;
}
static string _format_literal(const string& literal) {
string escaped = replacePattern(json(literal).dump(), GRAMMAR_LITERAL_ESCAPE_RE, [&](const auto& match) {
char c = match.str()[0];
return GRAMMAR_LITERAL_ESCAPES.at(c);
});
return "\"" + escaped + "\"";
}
static string _format_range_char(const string& ch) {
return replacePattern(ch, GRAMMAR_RANGE_LITERAL_ESCAPE_RE, [&](const auto& match) {
char c = match.str()[0];
return GRAMMAR_LITERAL_ESCAPES.at(c);
});
}
class SchemaConverter {
private:
std::optional<std::function<json(const string&)>> _fetch_json;
bool _dotall;
map<string, string> _rules;
unordered_map<string, nlohmann::json> _refs;
unordered_set<string> _refs_being_resolved;
vector<string> _errors;
vector<string> _warnings;
string _add_rule(const string& name, const string& rule) {
string esc_name = regex_replace(name, INVALID_RULE_CHARS_RE, "-");
if (_rules.find(esc_name) == _rules.end() || _rules[esc_name] == rule) {
_rules[esc_name] = rule;
return esc_name;
} else {
int i = 0;
while (_rules.find(esc_name + to_string(i)) != _rules.end() && _rules[esc_name + to_string(i)] != rule) {
i++;
}
string key = esc_name + to_string(i);
_rules[key] = rule;
return key;
}
}
string _generate_union_rule(const string& name, const vector<json>& alt_schemas) {
vector<string> rules;
for (size_t i = 0; i < alt_schemas.size(); i++) {
rules.push_back(visit(alt_schemas[i], name + (name.empty() ? "" : "-") + to_string(i)));
}
return join(rules.begin(), rules.end(), " | ");
}
string _visit_pattern(const string& pattern, const string& name) {
if (!(pattern.front() == '^' && pattern.back() == '$')) {
_errors.push_back("Pattern must start with '^' and end with '$'");
return "";
}
string sub_pattern = pattern.substr(1, pattern.length() - 2);
unordered_map<string, string> sub_rule_ids;
size_t i = 0;
size_t length = sub_pattern.length();
std::function<pair<string, bool>()> transform = [&]() -> pair<string, bool> {
size_t start = i;
vector<pair<string, bool>> seq;
auto get_dot = [&]() {
string rule;
if (_dotall) {
rule = "[\\U00000000-\\U0010FFFF]";
} else {
rule = "[\\U00000000-\\x09\\x0B\\x0C\\x0E-\\U0010FFFF]";
}
return _add_rule("dot", rule);
};
// Joins the sequence, merging consecutive literals together.
auto join_seq = [&]() {
vector<string> results;
string literal;
auto flush_literal = [&]() {
if (literal.empty()) {
return false;
}
results.push_back("\"" + literal + "\"");
literal.clear();
return true;
};
for (const auto& item : seq) {
if (item.second) {
literal += item.first.substr(1, item.first.length() - 2);
} else {
flush_literal();
results.push_back(item.first);
}
}
if (flush_literal() && results.size() == 1) {
return make_pair(results[0], true);
}
return make_pair(join(results.begin(), results.end(), " "), false);
};
while (i < length) {
char c = sub_pattern[i];
if (c == '.') {
seq.push_back(make_pair(get_dot(), false));
i++;
} else if (c == '(') {
i++;
if (i < length) {
if (sub_pattern[i] == '?') {
_warnings.push_back("Unsupported pattern syntax");
}
}
auto sub_result = transform();
seq.push_back(make_pair("(" + sub_result.first + ")", false));
} else if (c == ')') {
i++;
if (start > 0 && sub_pattern[start - 1] != '(') {
_errors.push_back("Unbalanced parentheses");
}
return join_seq();
} else if (c == '[') {
string square_brackets = string(1, c);
i++;
while (i < length && sub_pattern[i] != ']') {
if (sub_pattern[i] == '\\') {
square_brackets += sub_pattern.substr(i, 2);
i += 2;
} else {
square_brackets += sub_pattern[i];
i++;
}
}
if (i >= length) {
_errors.push_back("Unbalanced square brackets");
}
square_brackets += ']';
i++;
seq.push_back(make_pair(square_brackets, false));
} else if (c == '|') {
seq.push_back(make_pair("|", false));
i++;
} else if (c == '*' || c == '+' || c == '?') {
seq.back().first += c;
i++;
} else if (c == '{') {
string curly_brackets = string(1, c);
i++;
while (i < length && sub_pattern[i] != '}') {
curly_brackets += sub_pattern[i];
i++;
}
if (i >= length) {
_errors.push_back("Unbalanced curly brackets");
}
curly_brackets += '}';
i++;
auto nums = split(curly_brackets.substr(1, curly_brackets.length() - 2), ",");
int min_times = 0;
int max_times = numeric_limits<int>::max();
if (nums.size() == 1) {
min_times = max_times = stoi(nums[0]);
} else if (nums.size() != 2) {
_errors.push_back("Wrong number of values in curly brackets");
} else {
if (!nums[0].empty()) {
min_times = stoi(nums[0]);
}
if (!nums[1].empty()) {
max_times = stoi(nums[1]);
}
}
auto [sub, sub_is_literal] = seq.back();
if (min_times == 0 && max_times == numeric_limits<int>::max()) {
seq.back().first = sub + "*";
} else if (min_times == 0 && max_times == 1) {
seq.back().first = sub + "?";
} else if (min_times == 1 && max_times == numeric_limits<int>::max()) {
seq.back().first = sub + "+";
} else {
if (!sub_is_literal) {
string& sub_id = sub_rule_ids[sub];
if (sub_id.empty()) {
sub_id = _add_rule(name + "-" + to_string(sub_rule_ids.size()), sub);
}
sub = sub_id;
}
string result;
if (sub_is_literal && min_times > 0) {
result = "\"" + repeat(sub.substr(1, sub.length() - 2), min_times) + "\"";
} else {
for (int j = 0; j < min_times; j++) {
if (j > 0) {
result += " ";
}
result += sub;
}
}
if (min_times > 0 && min_times < max_times) {
result += " ";
}
if (max_times == numeric_limits<int>::max()) {
result += sub + "*";
} else {
for (int j = min_times; j < max_times; j++) {
if (j > min_times) {
result += " ";
}
result += sub + "?";
}
}
seq.back().first = result;
seq.back().second = false;
}
} else {
string literal;
while (i < length && NON_LITERAL_SET.find(sub_pattern[i]) == NON_LITERAL_SET.end() &&
(i == length - 1 || ESCAPED_IN_REGEXPS_BUT_NOT_IN_LITERALS.find(sub_pattern[i + 1]) == ESCAPED_IN_REGEXPS_BUT_NOT_IN_LITERALS.end())) {
if (sub_pattern[i] == '\\' && i < length - 1) {
i++;
if (NON_LITERAL_SET.find(sub_pattern[i]) != NON_LITERAL_SET.end()) {
literal += sub_pattern[i];
} else {
literal += "\\" + string(1, sub_pattern[i]);
}
i++;
} else {
if (sub_pattern[i] == '"') {
literal += "\\";
}
literal += sub_pattern[i];
i++;
}
}
if (!literal.empty()) {
seq.push_back(make_pair("\"" + literal + "\"", true));
}
if (i < length && NON_LITERAL_SET.find(sub_pattern[i]) == NON_LITERAL_SET.end()) {
seq.push_back(make_pair("\"" + string(1, sub_pattern[i]) + "\"", true));
i++;
}
}
}
return join_seq();
};
return _add_rule(name, transform().first);
}
string _resolve_ref(const string& ref) {
string ref_name = ref.substr(ref.find_last_of('/') + 1);
if (_rules.find(ref_name) == _rules.end() && _refs_being_resolved.find(ref) == _refs_being_resolved.end()) {
_refs_being_resolved.insert(ref);
json resolved = _refs[ref];
ref_name = visit(resolved, ref_name);
_refs_being_resolved.erase(ref);
}
return ref_name;
}
string _build_object_rule(const vector<pair<string, json>>& properties, const unordered_set<string>& required, const string& name) {
vector<string> required_props;
vector<string> optional_props;
unordered_map<string, string> prop_kv_rule_names;
for (const auto& [prop_name, prop_schema] : properties) {
string prop_rule_name = visit(prop_schema, name + (name.empty() ? "" : "-") + prop_name);
prop_kv_rule_names[prop_name] = _add_rule(
name + (name.empty() ? "" : "-") + prop_name + "-kv",
_format_literal(prop_name) + " space \":\" space " + prop_rule_name
);
if (required.find(prop_name) != required.end()) {
required_props.push_back(prop_name);
} else {
optional_props.push_back(prop_name);
}
}
string rule = "\"{\" space ";
for (size_t i = 0; i < required_props.size(); i++) {
if (i > 0) {
rule += " \",\" space ";
}
rule += prop_kv_rule_names[required_props[i]];
}
if (!optional_props.empty()) {
rule += " (";
if (!required_props.empty()) {
rule += " \",\" space ( ";
}
function<string(const vector<string>&, bool)> get_recursive_refs = [&](const vector<string>& ks, bool first_is_optional) {
string res;
if (ks.empty()) {
return res;
}
string k = ks[0];
string kv_rule_name = prop_kv_rule_names[k];
if (first_is_optional) {
res = "( \",\" space " + kv_rule_name + " )?";
} else {
res = kv_rule_name;
}
if (ks.size() > 1) {
res += " " + _add_rule(
name + (name.empty() ? "" : "-") + k + "-rest",
get_recursive_refs(vector<string>(ks.begin() + 1, ks.end()), true)
);
}
return res;
};
for (size_t i = 0; i < optional_props.size(); i++) {
if (i > 0) {
rule += " | ";
}
rule += get_recursive_refs(vector<string>(optional_props.begin() + i, optional_props.end()), false);
}
if (!required_props.empty()) {
rule += " )";
}
rule += " )?";
}
rule += " \"}\" space ";
return rule;
}
public:
SchemaConverter(
const std::optional<std::function<json(const string&)>>& fetch_json,
bool dotall)
: _fetch_json(fetch_json), _dotall(dotall)
{
_rules["space"] = SPACE_RULE;
}
void resolve_refs(nlohmann::json& schema, const std::string& url) {
/*
* Resolves all $ref fields in the given schema, fetching any remote schemas,
* replacing each $ref with absolute reference URL and populates _refs with the
* respective referenced (sub)schema dictionaries.
*/
function<void(json&)> visit_refs = [&](json& n) {
if (n.is_array()) {
for (auto& x : n) {
visit_refs(x);
}
} else if (n.is_object()) {
if (n.contains("$ref")) {
string ref = n["$ref"];
if (_refs.find(ref) == _refs.end()) {
json target;
if (ref.find("https://") == 0 && _fetch_json) {
string base_url = ref.substr(0, ref.find('#'));
auto it = _refs.find(base_url);
if (it != _refs.end()) {
target = it->second;
} else {
// Fetch the referenced schema and resolve its refs
auto referenced = _fetch_json.value()(ref);
resolve_refs(referenced, base_url);
_refs[base_url] = referenced;
}
if (ref.find('#') == string::npos || ref.substr(ref.find('#') + 1).empty()) {
return;
}
} else if (ref.find("#/") == 0) {
target = schema;
n["$ref"] = url + ref;
ref = url + ref;
} else {
_errors.push_back("Unsupported ref: " + ref);
return;
}
string pointer = ref.substr(ref.find('#') + 1);
vector<string> tokens = split(pointer, "/");
for (size_t i = 1; i < tokens.size(); ++i) {
string sel = tokens[i];
if (target.is_null() || !target.contains(sel)) {
_errors.push_back("Error resolving ref " + ref + ": " + sel + " not in " + target.dump());
return;
}
target = target[sel];
}
_refs[ref] = target;
}
} else {
for (auto& [key, value] : n.items()) {
visit_refs(value);
}
}
}
};
visit_refs(schema);
}
string visit(const json& schema, const string& name) {
json schema_type = schema.contains("type") ? schema["type"] : json();
string schema_format = schema.contains("format") ? schema["format"].get<string>() : "";
string rule_name = name.empty() ? "root" : name;
if (schema.contains("$ref")) {
return _resolve_ref(schema["$ref"]);
} else if (schema.contains("oneOf") || schema.contains("anyOf")) {
vector<json> alt_schemas = schema.contains("oneOf") ? schema["oneOf"].get<vector<json>>() : schema["anyOf"].get<vector<json>>();
return _add_rule(rule_name, _generate_union_rule(name, alt_schemas));
} else if (schema_type.is_array()) {
vector<json> schema_types;
for (const auto& t : schema_type) {
schema_types.push_back({{"type", t}});
}
return _add_rule(rule_name, _generate_union_rule(name, schema_types));
} else if (schema.contains("const")) {
return _add_rule(rule_name, _format_literal(schema["const"].dump()));
} else if (schema.contains("enum")) {
vector<string> enum_values;
for (const auto& v : schema["enum"]) {
enum_values.push_back(_format_literal(v.dump()));
}
return _add_rule(rule_name, join(enum_values.begin(), enum_values.end(), " | "));
} else if ((schema_type.is_null() || schema_type == "object") && schema.contains("properties")) {
unordered_set<string> required;
if (schema.contains("required") && schema["required"].is_array()) {
for (const auto& item : schema["required"]) {
if (item.is_string()) {
required.insert(item.get<string>());
}
}
}
vector<pair<string, json>> properties;
for (const auto& prop : schema["properties"].items()) {
properties.emplace_back(prop.key(), prop.value());
}
return _add_rule(rule_name, _build_object_rule(properties, required, name));
} else if ((schema_type.is_null() || schema_type == "object") && schema.contains("allOf")) {
unordered_set<string> required;
vector<pair<string, json>> properties;
string hybrid_name = name;
std::function<void(const json&, bool)> add_component = [&](const json& comp_schema, bool is_required) {
if (comp_schema.contains("$ref")) {
add_component(_refs[comp_schema["$ref"]], is_required);
} else if (comp_schema.contains("properties")) {
for (const auto& prop : comp_schema["properties"].items()) {
properties.emplace_back(prop.key(), prop.value());
if (is_required) {
required.insert(prop.key());
}
}
} else {
// todo warning
}
};
for (auto& t : schema["allOf"]) {
if (t.contains("anyOf")) {
for (auto& tt : t["anyOf"]) {
add_component(tt, false);
}
} else {
add_component(t, true);
}
}
return _add_rule(rule_name, _build_object_rule(properties, required, hybrid_name));
} else if ((schema_type.is_null() || schema_type == "object") && schema.contains("additionalProperties")) {
json additional_properties = schema["additionalProperties"];
if (!additional_properties.is_object()) {
additional_properties = json::object();
}
string sub_name = name + (name.empty() ? "" : "-") + "additionalProperties";
string value_rule = visit(additional_properties, sub_name + "-value");
string kv_rule = _add_rule(sub_name + "-kv", "string \":\" space " + value_rule);
return _add_rule(rule_name, "( " + kv_rule + " ( \",\" space " + kv_rule + " )* )*");
} else if ((schema_type.is_null() || schema_type == "array") && (schema.contains("items") || schema.contains("prefixItems"))) {
json items = schema.contains("items") ? schema["items"] : schema["prefixItems"];
if (items.is_array()) {
string rule = "\"[\" space ";
for (size_t i = 0; i < items.size(); i++) {
if (i > 0) {
rule += " \",\" space ";
}
rule += visit(items[i], name + (name.empty() ? "" : "-") + "tuple-" + to_string(i));
}
rule += " \"]\" space";
return _add_rule(rule_name, rule);
} else {
string item_rule_name = visit(items, name + (name.empty() ? "" : "-") + "item");
string list_item_operator = "( \",\" space " + item_rule_name + " )";
string successive_items;
int min_items = schema.contains("minItems") ? schema["minItems"].get<int>() : 0;
json max_items_json = schema.contains("maxItems") ? schema["maxItems"] : json();
int max_items = max_items_json.is_number_integer() ? max_items_json.get<int>() : -1;
if (min_items > 0) {
successive_items += repeat(list_item_operator, min_items - 1);
min_items--;
}
if (max_items >= 0 && max_items > min_items) {
successive_items += repeat(list_item_operator + "?", max_items - min_items - 1);
} else {
successive_items += list_item_operator + "*";
}
string rule;
if (min_items == 0) {
rule = "\"[\" space ( " + item_rule_name + " " + successive_items + " )? \"]\" space";
} else {
rule = "\"[\" space " + item_rule_name + " " + successive_items + " \"]\" space";
}
return _add_rule(rule_name, rule);
}
} else if ((schema_type.is_null() || schema_type == "string") && schema.contains("pattern")) {
return _visit_pattern(schema["pattern"], rule_name);
} else if (schema.empty() || (schema.size() == 1 && schema_type == "object")) {
for (const auto& [t, r] : PRIMITIVE_RULES) {
_add_rule(t, r);
}
return "object";
} else if ((schema_type.is_null() || schema_type == "string") && regex_match(schema_format, regex("^uuid[1-5]?$"))) {
return _add_rule(rule_name == "root" ? "root" : schema_format, PRIMITIVE_RULES.at("uuid"));
} else if ((schema_type.is_null() || schema_type == "string") && DATE_RULES.find(schema_format) != DATE_RULES.end()) {
for (const auto& [t, r] : DATE_RULES) {
_add_rule(t, r);
}
return schema_format + "-string";
} else {
if (PRIMITIVE_RULES.find(schema_type.get<string>()) == PRIMITIVE_RULES.end()) {
_errors.push_back("Unrecognized schema: " + schema.dump());
return "";
}
// TODO: support minimum, maximum, exclusiveMinimum, exclusiveMaximum at least for zero
return _add_rule(rule_name == "root" ? "root" : schema_type.get<string>(), PRIMITIVE_RULES.at(schema_type.get<string>()));
}
}
void check_errors() {
if (!_errors.empty()) {
throw std::runtime_error("JSON schema conversion failed:\n" + join(_errors.begin(), _errors.end(), "\n"));
}
if (!_warnings.empty()) {
std::cerr << "WARNING: JSON schema conversion was incomplete: " + join(_warnings.begin(), _warnings.end(), "; ") << std::endl;
}
}
string format_grammar() {
stringstream ss;
for (const auto& [name, rule] : _rules) {
ss << name << " ::= " << rule << endl;
}
return ss.str();
}
};
string json_schema_to_grammar(const json& schema) {
SchemaConverter converter(/* fetch_json= */ std::nullopt, /* dotall= */ false);
auto copy = schema;
converter.resolve_refs(copy, "input");
converter.visit(copy, "");
converter.check_errors();
return converter.format_grammar();
}
#ifdef LLAMA_BUILD_JSON_SCHEMA_CONVERTER
int main(int argc, const char** argv) {
if (argc != 2) {
cerr << "Expected only one argument" << endl;
return -1;
}
string file(argv[1]);
string schema;
if (file == "-") {
schema.append(istreambuf_iterator<char>(cin), istreambuf_iterator<char>());
} else {
ifstream in(argv[1]);
schema.append(istreambuf_iterator<char>(in), istreambuf_iterator<char>());
}
cout << json_schema_to_grammar(json::parse(schema)).c_str() << endl;
}
#endif

4
examples/server/json-schema-to-grammar.h Normal file
View file

@ -0,0 +1,4 @@
#pragma once
#include "json.hpp"
std::string json_schema_to_grammar(const nlohmann::json& schema);

9
examples/server/server.cpp
View file

@ -1,6 +1,7 @@
#include "utils.hpp"
#include "common.h"
#include "json-schema-to-grammar.h"
#include "llama.h"
#include "grammar-parser.h"
@ -178,6 +179,7 @@ struct server_slot {
llama_token sampled;
struct llama_sampling_params sparams;
llama_sampling_context * ctx_sampling = nullptr;
json json_schema;
int32_t ga_i = 0; // group-attention state
int32_t ga_n = 1; // group-attention factor
@ -838,7 +840,12 @@ struct server_context {
slot.sparams.penalize_nl = json_value(data, "penalize_nl", default_sparams.penalize_nl);
slot.params.n_keep = json_value(data, "n_keep", slot.params.n_keep);
slot.params.seed = json_value(data, "seed", default_params.seed);
slot.sparams.grammar = json_value(data, "grammar", default_sparams.grammar);
if (data.contains("json_schema") && !data.contains("grammar")) {
auto schema = json_value(data, "json_schema", json::object());
slot.sparams.grammar = json_schema_to_grammar(schema);
} else {
slot.sparams.grammar = json_value(data, "grammar", default_sparams.grammar);
}
slot.sparams.n_probs = json_value(data, "n_probs", default_sparams.n_probs);
slot.sparams.min_keep = json_value(data, "min_keep", default_sparams.min_keep);

10
examples/server/utils.hpp
View file

@ -361,10 +361,18 @@ static json oaicompat_completion_params_parse(
llama_params["ignore_eos"] = json_value(body, "ignore_eos", false);
llama_params["tfs_z"] = json_value(body, "tfs_z", default_sparams.tfs_z);
if (body.count("grammar") != 0) {
if (body.contains("grammar")) {
llama_params["grammar"] = json_value(body, "grammar", json::object());
}
if (body.contains("response_format")) {
auto response_format = json_value(body, "response_format", json::object());
if (response_format.contains("schema") && response_format["type"] == "json_object") {
llama_params["json_schema"] = json_value(response_format, "schema", json::object());
std::cerr << "GOT " << llama_params["json_schema"] << std::endl;
}
}
// Handle 'stop' field
if (body.contains("stop") && body["stop"].is_string()) {
llama_params["stop"] = json::array({body["stop"].get<std::string>()});

2
tests/CMakeLists.txt
View file

@ -59,6 +59,8 @@ llama_build_and_test_executable(test-backend-ops.cpp)
llama_build_and_test_executable(test-rope.cpp)
llama_build_and_test_executable(test-json-schema-to-grammar.cpp)
llama_build_and_test_executable_with_label(test-model-load-cancel.cpp "model")
llama_build_and_test_executable_with_label(test-autorelease.cpp "model")

10
tests/run-json-schema-to-grammar.mjs Normal file
View file

@ -0,0 +1,10 @@
import { readFileSync } from "fs"
import { SchemaConverter } from "../examples/server/public/json-schema-to-grammar.mjs"
const [, , file] = process.argv
const url = `file://${file}`
let schema = JSON.parse(readFileSync(file, "utf8"));
const converter = new SchemaConverter({})
schema = await converter.resolveRefs(schema, url)
converter.visit(schema, '')
console.log(converter.formatGrammar())

307
tests/test-json-schema-to-grammar.cpp Executable file
View file

@ -0,0 +1,307 @@
#ifdef NDEBUG
#undef NDEBUG
#endif
#include "llama.h"
#include "grammar-parser.h"
#include <cstdio>
#include <cstdlib>
#include <filesystem>
#include <iostream>
#include <istream>
#include <fstream>
#include <sstream>
#include "../examples/server/json-schema-to-grammar.h"
#include <cassert>
#include <regex>
using namespace std;
namespace fs = std::filesystem;
string INPUT_NAME(tmpnam(nullptr));
string OUT_NAME(tmpnam(nullptr));
static std::string trim(const std::string & source) {
std::string s(source);
s.erase(0,s.find_first_not_of(" \n\r\t"));
s.erase(s.find_last_not_of(" \n\r\t")+1);
return regex_replace(s, regex("(^|\n)[ \t]+"), "$1");
}
struct TestCase {
string name;
string schema;
string expected;
void prepare() const {
ofstream f;
f.open(INPUT_NAME);
f << schema.c_str();
f.close();
}
void read_and_verify(const string& series) const {
ostringstream actuals;
actuals << ifstream(OUT_NAME).rdbuf();
auto actual = actuals.str();
verify(series, actual);
}
void verify(const string& series, const string& actual) const {
if (trim(actual) != trim(expected)) {
cerr << "#" << endl;
cerr << "# Test " << series.c_str() << " / " << name.c_str() << " failed." << endl;
cerr << "#" << endl;
cerr << schema.c_str() << endl;
cerr << "# EXPECTED:\n" << expected.c_str() << endl;
cerr << "# ACTUAL:\n" << actual.c_str() << endl;
assert(false);
}
}
};
static void run_py(const TestCase& tc) {
cerr << "# Running Python " << tc.name.c_str() << endl;
tc.prepare();
std::system(("python ./examples/json-schema-to-grammar.py " + INPUT_NAME + " > " + OUT_NAME).c_str());
tc.read_and_verify("Python");
}
static void run_mjs(const TestCase& tc) {
cerr << "# Running MJS " << tc.name.c_str() << endl;
tc.prepare();
std::system(("node ./tests/run-json-schema-to-grammar.mjs " + INPUT_NAME + " > " + OUT_NAME).c_str());
tc.read_and_verify("JavaScript");
}
static void run_cpp(const TestCase& tc) {
cerr << "# Running C++ " << tc.name.c_str() << endl;
auto actual = json_schema_to_grammar(nlohmann::json::parse(tc.schema));
tc.verify("C++", actual);
}
static void run_all(const TestCase& tc) {
run_py(tc);
run_mjs(tc);
run_cpp(tc);
}
int main() {
run_all({
.name = "exotic formats",
.schema = R"""({
"prefixItems": [
{ "format": "date" },
{ "format": "uuid" },
{ "format": "time" },
{ "format": "date-time" }
]
})""",
.expected = R"""(
date ::= [0-9] [0-9] [0-9] [0-9] "-" ( "0" [1-9] | "1" [0-2] ) "-" ( [0-2] [0-9] | "3" [0-1] )
date-string ::= "\"" date "\"" space
date-time ::= date "T" time
date-time-string ::= "\"" date-time "\"" space
root ::= "[" space date-string "," space uuid "," space time-string "," space date-time-string "]" space
space ::= " "?
time ::= ([01] [0-9] | "2" [0-3]) ":" [0-5] [0-9] ":" [0-5] [0-9] ( "." [0-9] [0-9] [0-9] )? ( "Z" | ( "+" | "-" ) ( [01] [0-9] | "2" [0-3] ) ":" [0-5] [0-9] )
time-string ::= "\"" time "\"" space
uuid ::= "\"" [0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F] "-" [0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F] "-" [0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F] "-" [0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F] "-" [0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F][0-9a-fA-F] "\"" space
)"""
});
run_all({
.name = "string",
.schema = R"""({
"type": "string"
})""",
.expected = R"""(
root ::= "\"" (
[^"\\] |
"\\" (["\\/bfnrt] | "u" [0-9a-fA-F] [0-9a-fA-F] [0-9a-fA-F] [0-9a-fA-F])
)* "\"" space
space ::= " "?
)"""
});
run_all({
.name = "boolean",
.schema = R"""({
"type": "boolean"
})""",
.expected = R"""(
root ::= ("true" | "false") space
space ::= " "?
)"""
});
run_all({
.name = "integer",
.schema = R"""({
"type": "integer"
})""",
.expected = R"""(
root ::= ("-"? ([0-9] | [1-9] [0-9]*)) space
space ::= " "?
)"""
});
run_all({
.name = "number",
.schema = R"""({
"type": "number"
})""",
.expected = R"""(
root ::= ("-"? ([0-9] | [1-9] [0-9]*)) ("." [0-9]+)? ([eE] [-+]? [0-9]+)? space
space ::= " "?
)"""
});
run_all({
.name = "minItems",
.schema = R"""({
"items": {
"type": "boolean"
},
"minItems": 2
})""",
.expected = R"""(
boolean ::= ("true" | "false") space
root ::= "[" space boolean ( "," space boolean )( "," space boolean )* "]" space
space ::= " "?
)"""
});
run_all({
.name = "maxItems 1",
.schema = R"""({
"items": {
"type": "boolean"
},
"maxItems": 1
})""",
.expected = R"""(
boolean ::= ("true" | "false") space
root ::= "[" space ( boolean )? "]" space
space ::= " "?
)"""
});
run_all({
.name = "maxItems 2",
.schema = R"""({
"items": {
"type": "boolean"
},
"maxItems": 2
})""",
.expected = R"""(
boolean ::= ("true" | "false") space
root ::= "[" space ( boolean ( "," space boolean )? )? "]" space
space ::= " "?
)"""
});
run_all({
.name = "min + maxItems",
.schema = R"""({
"items": {
"type": ["number", "integer"]
},
"minItems": 3,
"maxItems": 5
})""",
.expected = R"""(
integer ::= ("-"? ([0-9] | [1-9] [0-9]*)) space
item ::= number | integer
number ::= ("-"? ([0-9] | [1-9] [0-9]*)) ("." [0-9]+)? ([eE] [-+]? [0-9]+)? space
root ::= "[" space item ( "," space item )( "," space item )( "," space item )?( "," space item )? "]" space
space ::= " "?
)"""
});
run_all({
.name = "regexp",
.schema = R"""({
"type": "string",
"pattern": "^(\\([0-9]{1,3}\\))?[0-9]{3}-[0-9]{4} and...$"
})""",
.expected = R"""(
dot ::= [\U00000000-\x09\x0B\x0C\x0E-\U0010FFFF]
root ::= ("(" root-1 root-1? root-1? ")")? root-1 root-1 root-1 "-" root-1 root-1 root-1 root-1 " and" dot dot dot
root-1 ::= [0-9]
space ::= " "?
)"""
});
run_all({
.name = "optionals",
.schema = R"""({
"$schema": "http://json-schema.org/draft-07/schema#",
"type": "object",
"properties": {
"a": {
"type": "string"
},
"b": {
"type": "string"
},
"c": {
"type": [
"number",
"string"
]
},
"d": {
"type": "string"
},
"e": {
"type": "object",
"additionalProperties": {
"type": "array",
"items": {
"type": "array",
"minItems": 2,
"items": [
{
"type": "string"
},
{
"type": "number"
}
],
"maxItems": 2
}
}
}
},
"required": [
"a",
"b"
],
"additionalProperties": false,
"definitions": {}
})""",
.expected = R"""(
a-kv ::= "\"a\"" space ":" space string
b-kv ::= "\"b\"" space ":" space string
c ::= number | string
c-kv ::= "\"c\"" space ":" space c
c-rest ::= ( "," space d-kv )? d-rest
d-kv ::= "\"d\"" space ":" space string
d-rest ::= ( "," space e-kv )?
e ::= ( e-additionalProperties-kv ( "," space e-additionalProperties-kv )* )*
e-additionalProperties-kv ::= string ":" space e-additionalProperties-value
e-additionalProperties-value ::= "[" space ( e-additionalProperties-value-item ( "," space e-additionalProperties-value-item )* )? "]" space
e-additionalProperties-value-item ::= "[" space string "," space number "]" space
e-kv ::= "\"e\"" space ":" space e
number ::= ("-"? ([0-9] | [1-9] [0-9]*)) ("." [0-9]+)? ([eE] [-+]? [0-9]+)? space
root ::= "{" space a-kv "," space b-kv ( "," space ( c-kv c-rest | d-kv d-rest | e-kv ) )? "}" space
space ::= " "?
string ::= "\"" (
[^"\\] |
"\\" (["\\/bfnrt] | "u" [0-9a-fA-F] [0-9a-fA-F] [0-9a-fA-F] [0-9a-fA-F])
)* "\"" space
)"""
});
}