Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 01:05:09 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix indentation (+1 squashed commits)

Browse Source 
Squashed commits:

[4d0fc028] testing a simple workflow for windows full build
This commit is contained in:
Concedo
2024-04-11 21:28:25 +08:00
parent dd10a51ab6
commit d0e40f9233
5 changed files with 61 additions and 352 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/kcpp-build-release-win-full.yaml
+60
View File
@@ -0,0 +1,60 @@
name: Koboldcpp Builder Windows Full Binaries
on: workflow_dispatch
env:
BRANCH_NAME: ${{ github.head_ref || github.ref_name }}
jobs:
windows:
runs-on: windows-2019
steps:
- name: Clone
id: checkout
uses: actions/checkout@v3
with:
ref: concedo_experimental
- name: Get Python
uses: actions/setup-python@v2
with:
python-version: 3.8.10
- name: Install python dependencies
run: |
python -m pip install --upgrade pip
pip install customtkinter==5.2.0 pyinstaller==5.11.0 psutil==5.9.5
- name: Download and install win64devkit
run: |
curl -L https://github.com/skeeto/w64devkit/releases/download/v1.22.0/w64devkit-1.22.0.zip --output w64devkit.zip
Expand-Archive w64devkit.zip -DestinationPath .
- name: Add w64devkit to PATH
run: |
$env:Path += ";$(Get-Location)\w64devkit-1.22.0\w64devkit\bin"
echo $env:Path
- name: Build Non-CUDA
id: make_build
run: |
make -j ${env:NUMBER_OF_PROCESSORS}
ls
# - uses: Jimver/cuda-toolkit@v0.2.11
# id: cuda-toolkit
# with:
# cuda: '11.4.4'
# - name: Build CUDA
# id: cmake_build
# run: |
# mkdir build
# cd build
# cmake .. -DLLAMA_CUBLAS=ON -DCMAKE_SYSTEM_VERSION="10.0.19041.0"
# cmake --build . --config Release -j ${env:NUMBER_OF_PROCESSORS}
# - name: Save artifact
# uses: actions/upload-artifact@v3
# with:
# name: kcpp_windows_cuda_binary
# path: build/bin/Release/
cmake/FindSIMD.cmake
-100
View File
@@ -1,100 +0,0 @@
include(CheckCSourceRuns)
set(AVX_CODE "
#include <immintrin.h>
int main()
{
__m256 a;
a = _mm256_set1_ps(0);
return 0;
}
")
set(AVX512_CODE "
#include <immintrin.h>
int main()
{
__m512i a = _mm512_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0);
__m512i b = a;
__mmask64 equality_mask = _mm512_cmp_epi8_mask(a, b, _MM_CMPINT_EQ);
return 0;
}
")
set(AVX2_CODE "
#include <immintrin.h>
int main()
{
__m256i a = {0};
a = _mm256_abs_epi16(a);
__m256i x;
_mm256_extract_epi64(x, 0); // we rely on this in our AVX2 code
return 0;
}
")
set(FMA_CODE "
#include <immintrin.h>
int main()
{
__m256 acc = _mm256_setzero_ps();
const __m256 d = _mm256_setzero_ps();
const __m256 p = _mm256_setzero_ps();
acc = _mm256_fmadd_ps( d, p, acc );
return 0;
}
")
macro(check_sse type flags)
set(__FLAG_I 1)
set(CMAKE_REQUIRED_FLAGS_SAVE ${CMAKE_REQUIRED_FLAGS})
foreach (__FLAG ${flags})
if (NOT ${type}_FOUND)
set(CMAKE_REQUIRED_FLAGS ${__FLAG})
check_c_source_runs("${${type}_CODE}" HAS_${type}_${__FLAG_I})
if (HAS_${type}_${__FLAG_I})
set(${type}_FOUND TRUE CACHE BOOL "${type} support")
set(${type}_FLAGS "${__FLAG}" CACHE STRING "${type} flags")
endif()
math(EXPR __FLAG_I "${__FLAG_I}+1")
endif()
endforeach()
set(CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS_SAVE})
if (NOT ${type}_FOUND)
set(${type}_FOUND FALSE CACHE BOOL "${type} support")
set(${type}_FLAGS "" CACHE STRING "${type} flags")
endif()
mark_as_advanced(${type}_FOUND ${type}_FLAGS)
endmacro()
# flags are for MSVC only!
check_sse("AVX" " ;/arch:AVX")
if (NOT ${AVX_FOUND})
set(LLAMA_AVX OFF)
else()
set(LLAMA_AVX ON)
endif()
check_sse("AVX2" " ;/arch:AVX2")
check_sse("FMA" " ;/arch:AVX2")
if ((NOT ${AVX2_FOUND}) OR (NOT ${FMA_FOUND}))
set(LLAMA_AVX2 OFF)
else()
set(LLAMA_AVX2 ON)
endif()
check_sse("AVX512" " ;/arch:AVX512")
if (NOT ${AVX512_FOUND})
set(LLAMA_AVX512 OFF)
else()
set(LLAMA_AVX512 ON)
endif()
requirements.txt
+1
View File
@@ -4,3 +4,4 @@ transformers>=4.34.0
gguf>=0.1.0
customtkinter>=5.1.0
protobuf>=4.21.0
psutil>=5.9.4
scripts/gen-authors.sh
-9
View File
@@ -1,9 +0,0 @@
#!/bin/bash
printf "# date: $(date)\n" > AUTHORS
printf "# this file is auto-generated by scripts/gen-authors.sh\n\n" >> AUTHORS
git log --format='%an <%ae>' --reverse --date=short master | awk '!seen[$0]++' | sort >> AUTHORS
# if necessary, update your name here. for example: jdoe -> John Doe
sed -i '' 's/^jdoe/John Doe/g' AUTHORS
tests/test-grammar-integration.cpp
-243
View File
@@ -1,243 +0,0 @@
#ifdef NDEBUG
#undef NDEBUG
#endif
#define LLAMA_API_INTERNAL
#include "ggml.h"
#include "llama.h"
#include "grammar-parser.h"
#include "unicode.h"
#include <cassert>
#include <string>
static void test_simple_grammar() {
// Test case for a simple grammar
const std::string grammar_str = R"""(root ::= expr
expr ::= term ("+" term)*
term ::= number
number ::= [0-9]+)""";
grammar_parser::parse_state parsed_grammar = grammar_parser::parse(grammar_str.c_str());
// Ensure we parsed correctly
assert(!parsed_grammar.rules.empty());
// Ensure we have a root node
assert(!(parsed_grammar.symbol_ids.find("root") == parsed_grammar.symbol_ids.end()));
std::vector<const llama_grammar_element*> grammar_rules(parsed_grammar.c_rules());
llama_grammar* grammar = llama_grammar_init(
grammar_rules.data(), grammar_rules.size(), parsed_grammar.symbol_ids.at("root"));
std::string input = "123+456";
auto decoded = decode_utf8(input, {});
const auto & code_points = decoded.first;
for (auto it = code_points.begin(), end = code_points.end() - 1; it != end; ++it) {
auto prev_stacks = grammar->stacks;
grammar->stacks = llama_grammar_accept(grammar->rules, grammar->stacks, *it);
assert(!grammar->stacks.empty());
}
bool completed_grammar = false;
for (const auto & stack : grammar->stacks) {
if (stack.empty()) {
completed_grammar = true;
break;
}
}
assert(completed_grammar);
// Clean up allocated memory
llama_grammar_free(grammar);
}
static void test_complex_grammar() {
// Test case for a more complex grammar, with both failure strings and success strings
const std::string grammar_str = R"""(root ::= expression
expression ::= term ws (("+"|"-") ws term)*
term ::= factor ws (("*"|"/") ws factor)*
factor ::= number | variable | "(" expression ")" | function-call
number ::= [0-9]+
variable ::= [a-zA-Z_][a-zA-Z0-9_]*
function-call ::= variable ws "(" (expression ("," ws expression)*)? ")"
ws ::= [ \t\n\r]?)""";
grammar_parser::parse_state parsed_grammar = grammar_parser::parse(grammar_str.c_str());
// Ensure we parsed correctly
assert(!parsed_grammar.rules.empty());
// Ensure we have a root node
assert(!(parsed_grammar.symbol_ids.find("root") == parsed_grammar.symbol_ids.end()));
std::vector<const llama_grammar_element*> grammar_rules(parsed_grammar.c_rules());
llama_grammar* grammar = llama_grammar_init(
grammar_rules.data(), grammar_rules.size(), parsed_grammar.symbol_ids.at("root"));
// Save the original grammar stacks so that we can reset after every new string we want to test
auto original_stacks = grammar->stacks;
// Test a few strings
std::vector<std::string> test_strings_pass = {
"42",
"1*2*3*4*5",
"x",
"x+10",
"x1+y2",
"(a+b)*(c-d)",
"func()",
"func(x,y+2)",
"a*(b+c)-d/e",
"f(g(x),h(y,z))",
"x + 10",
"x1 + y2",
"(a + b) * (c - d)",
"func()",
"func(x, y + 2)",
"a * (b + c) - d / e",
"f(g(x), h(y, z))",
"123+456",
"123*456*789-123/456+789*123",
"123+456*789-123/456+789*123-456/789+123*456-789/123+456*789-123/456+789*123-456"
};
std::vector<std::string> test_strings_fail = {
"+",
"/ 3x",
"x + + y",
"a * / b",
"func(,)",
"func(x y)",
"(a + b",
"x + y)",
"a + b * (c - d",
"42 +",
"x +",
"x + 10 +",
"(a + b) * (c - d",
"func(",
"func(x, y + 2",
"a * (b + c) - d /",
"f(g(x), h(y, z)",
"123+456*789-123/456+789*123-456/789+123*456-789/123+456*789-123/456+789*123-456/",
};
// Passing strings
for (const auto & test_string : test_strings_pass) {
auto decoded = decode_utf8(test_string, {});
const auto & code_points = decoded.first;
int pos = 0;
for (auto it = code_points.begin(), end = code_points.end() - 1; it != end; ++it) {
++pos;
auto prev_stacks = grammar->stacks;
grammar->stacks = llama_grammar_accept(grammar->rules, grammar->stacks, *it);
// Expect that each code point will not cause the grammar to fail
if (grammar->stacks.empty()) {
fprintf(stdout, "Error at position %d\n", pos);
fprintf(stderr, "Unexpected character '%s'\n", unicode_cpt_to_utf8(*it).c_str());
fprintf(stderr, "Input string is %s:\n", test_string.c_str());
}
assert(!grammar->stacks.empty());
}
bool completed_grammar = false;
for (const auto & stack : grammar->stacks) {
if (stack.empty()) {
completed_grammar = true;
break;
}
}
assert(completed_grammar);
// Reset the grammar stacks
grammar->stacks = original_stacks;
}
// Failing strings
for (const auto & test_string : test_strings_fail) {
auto decoded = decode_utf8(test_string, {});
const auto & code_points = decoded.first;
bool parse_failed = false;
for (auto it = code_points.begin(), end = code_points.end() - 1; it != end; ++it) {
auto prev_stacks = grammar->stacks;
grammar->stacks = llama_grammar_accept(grammar->rules, grammar->stacks, *it);
if (grammar->stacks.empty()) {
parse_failed = true;
break;
}
assert(!grammar->stacks.empty());
}
bool completed_grammar = false;
for (const auto & stack : grammar->stacks) {
if (stack.empty()) {
completed_grammar = true;
break;
}
}
// Ensure that the grammar is not completed, or that each string failed to match as-expected
assert((!completed_grammar) || parse_failed);
// Reset the grammar stacks
grammar->stacks = original_stacks;
}
// Clean up allocated memory
llama_grammar_free(grammar);
}
static void test_failure_missing_root() {
// Test case for a grammar that is missing a root rule
const std::string grammar_str = R"""(rot ::= expr
expr ::= term ("+" term)*
term ::= number
number ::= [0-9]+)""";
grammar_parser::parse_state parsed_grammar = grammar_parser::parse(grammar_str.c_str());
// Ensure we parsed correctly
assert(!parsed_grammar.rules.empty());
// Ensure we do NOT have a root node
assert(parsed_grammar.symbol_ids.find("root") == parsed_grammar.symbol_ids.end());
}
static void test_failure_missing_reference() {
// Test case for a grammar that is missing a referenced rule
const std::string grammar_str = R"""(root ::= expr
expr ::= term ("+" term)*
term ::= numero
number ::= [0-9]+)""";
fprintf(stderr, "Expected error: ");
grammar_parser::parse_state parsed_grammar = grammar_parser::parse(grammar_str.c_str());
// Ensure we did NOT parsed correctly
assert(parsed_grammar.rules.empty());
fprintf(stderr, "End of expected error. Test successful.\n");
}
int main() {
test_simple_grammar();
test_complex_grammar();
test_failure_missing_root();
test_failure_missing_reference();
return 0;
}