Felipe97/llama-cpp-compiled

1#include "testing.h"2 3#include "mtmd-image.h"4#include "mtmd-internal.h"5 6#include <iostream>7#include <stdexcept>8#include <string>9#include <tuple>10#include <utility>11#include <vector>12 13// this test file contains:14// 1. test cases for mtmd helpers15// 2. test cases for internal mtmd components16// internal headers can be included here17 18struct test_registry {19    using fn_t = void (*)(testing &);20 21    struct entry {22        std::string name;23        fn_t fn;24    };25 26    static std::vector<entry> & all() {27        static std::vector<entry> entries;28        return entries;29    }30 31    test_registry(const char * name, fn_t fn) {32        all().push_back({ name, fn });33    }34};35 36#define MAKE_TEST(name)                                               \37    static void name(testing & t);                                    \38    static const test_registry test_registry_ ## name(#name, &name);  \39    static void name(testing & t)40 41 42//43// mtmd_image44//45 46MAKE_TEST(test_image_preprocessor_lfm2) {47    clip_hparams hparams;48    hparams.patch_size = 16;49    hparams.n_merge = 2;50    hparams.set_limit_image_tokens(64, 256);51 52    // { image size, expected tiling }53    const std::vector<std::pair<clip_image_size, bool>> cases = {54        { {  704, 704 }, false },55        // 720 / (patch_size * n_merge) is exactly 22.5, so this only matches HF56        // if round_by_factor rounds half to even (22) instead of away from zero (23)57        { {  720, 720 }, false },58        { {  736, 736 }, true  },59        { { 1024, 977 }, true  },60        { { 1056, 384 }, false },61    };62 63    for (const auto & [size, expected] : cases) {64        const bool actual = mtmd_image_preprocessor_lfm2::should_tile(hparams, size);65 66        t.assert_equal(67            "tiling for " + std::to_string(size.width) + "x" + std::to_string(size.height),68            std::string(expected ? "tiled" : "single"),69            std::string(actual   ? "tiled" : "single"));70    }71}72 73//74// mtmd temporal merge75//76 77MAKE_TEST(test_temporal_merge_grouping) {78    std::vector<mtmd::bitmap_ptr> pool; // keeps the bitmaps alive until the end of the test79 80    // spec chars:81    //   v = video frame, w = video frame of another size, a = audio, i = plain image, t = text82    auto make_parts = [&pool](const std::string & spec) {83        std::vector<mtmd_internal_part> parts;84        for (char c : spec) {85            if (c == 't') {86                parts.push_back({ "hello", nullptr });87                continue;88            }89            mtmd_bitmap * bm = nullptr;90            switch (c) {91                case 'v': bm = mtmd_bitmap_init(100, 100, nullptr);   break;92                case 'w': bm = mtmd_bitmap_init(200, 200, nullptr);   break;93                case 'a': bm = mtmd_bitmap_init_from_audio(100, nullptr); break;94                case 'i': bm = mtmd_bitmap_init(100, 100, nullptr);   break;95                default: throw std::runtime_error(std::string("unknown spec char: ") + c);96            }97            mtmd_bitmap_set_mergeable(bm, c != 'i');98            pool.emplace_back(bm);99            parts.push_back({ "", bm });100        }101        return parts;102    };103 104    // { parts, n_merge, expected size of each group }105    const std::vector<std::tuple<std::string, int, std::string>> cases = {106        { "vv",   2, "2"    },107        { "vvv",  2, "21"   },108        { "vvvv", 2, "22"   },109        { "vvi",  2, "21"   },110        { "tvvt", 2, "2"    },111        { "vtv",  2, "11"   }, // text in between breaks the merge112        { "vw",   2, "11"   }, // different sizes cannot be merged113        { "aa",   2, "11"   }, // audio is never merged114        { "ii",   2, "11"   }, // two unrelated images must stay separated115        { "iv",   2, "11"   },116        { "vi",   2, "11"   },117        { "vv",   1, "11"   }, // model without temporal merge118    };119 120    for (const auto & [spec, n_merge, expected] : cases) {121        auto parts  = make_parts(spec);122        auto groups = mtmd_group_mergeable_bitmaps(parts, n_merge);123 124        std::string actual;125        for (const auto & group : groups) {126            actual += std::to_string(group.size());127        }128 129        const std::string name = "\"" + spec + "\" with n_merge=" + std::to_string(n_merge);130        t.assert_equal("groups for " + name, expected, actual);131 132        size_t n_bitmap_parts = 0;133        for (const auto & p : parts) {134            n_bitmap_parts += p.bitmap != nullptr ? 1 : 0;135        }136        t.assert_equal("remaining bitmap parts for " + name, groups.size(), n_bitmap_parts);137    }138}139 140//141// main142//143 144int main(int argc, char ** argv) {145    testing t(std::cout);146    t.verbose = true;147 148    // usage: test-mtmd-impl [filter_regex]149    for (int i = 1; i < argc; i++) {150        t.set_filter(argv[i]);151    }152 153    for (const auto & e : test_registry::all()) {154        t.test(e.name, e.fn);155    }156 157    return t.summary();158}159