gpt4all/gpt4all-backend/llamamodel.cpp

#define LLAMAMODEL_H_I_KNOW_WHAT_I_AM_DOING_WHEN_INCLUDING_THIS_FILE
#include "llamamodel_impl.h"

#include <cassert>
#include <cmath>
#include <cstdio>
#include <cstring>
#include <fstream>
#include <map>
#include <string>
#include <vector>
#include <iostream>
#if defined(_WIN32) && defined(_MSC_VER)
    #define WIN32_LEAN_AND_MEAN
    #ifndef NOMINMAX
        #define NOMINMAX
    #endif
    #include <windows.h>
    #include <io.h>
    #include <stdio.h>
#else
    #include <unistd.h>
#endif
#include <random>
#include <thread>
#include <unordered_set>

#include <llama.h>
#include <ggml.h>

#ifdef GGML_USE_KOMPUTE
#include "ggml-kompute.h"
#endif

// Maximum supported GGUF version
static constexpr int GGUF_VER_MAX = 3;

namespace {
const char *modelType_ = "LLaMA";
}

static bool llama_verbose() {
    const char* var = getenv("GPT4ALL_VERBOSE_LLAMACPP");
    return var && *var;
}

static void llama_log_callback(enum ggml_log_level level, const char *text, void *userdata) {
    (void)userdata;
    if (llama_verbose() || level <= GGML_LOG_LEVEL_ERROR) {
        fputs(text, stderr);
    }
}

struct gpt_params {
    int32_t seed          = -1;   // RNG seed
    int32_t n_keep        = 0;    // number of tokens to keep from initial prompt

    // sampling parameters
    float   tfs_z         = 1.0f; // 1.0 = disabled
    float   typical_p     = 1.0f; // 1.0 = disabled

    std::string prompt = "";

    enum ggml_type kv_type = GGML_TYPE_F16; // use f16 instead of f32 for memory kv

    bool use_mmap          = true;  // use mmap for faster loads
    bool use_mlock         = false; // use mlock to keep model in memory
};

static int llama_sample_top_p_top_k(
        llama_context *ctx,
        const llama_token *last_n_tokens_data,
        int last_n_tokens_size,
        int top_k,
        float top_p,
        float temp,
        float repeat_penalty,
        int32_t pos) {
    auto logits = llama_get_logits_ith(ctx, pos);
    auto n_vocab = llama_n_vocab(llama_get_model(ctx));
    // Populate initial list of all candidates
    std::vector<llama_token_data> candidates;
    candidates.reserve(n_vocab);
    for (int token_id = 0; token_id < n_vocab; token_id++) {
        candidates.emplace_back(llama_token_data{token_id, logits[token_id], 0.0f});
    }
    llama_token_data_array candidates_p = {candidates.data(), candidates.size(), false};
    // Sample repeat penalty
    llama_sample_repetition_penalties(nullptr, &candidates_p, last_n_tokens_data, last_n_tokens_size, repeat_penalty, 0.0f, 0.0f);
    // Temperature sampling
    llama_sample_top_k(ctx, &candidates_p, top_k, 1);
    llama_sample_tail_free(ctx, &candidates_p, 1.0f, 1);
    llama_sample_typical(ctx, &candidates_p, 1.0f, 1);
    llama_sample_top_p(ctx, &candidates_p, top_p, 1);
    llama_sample_temp(ctx, &candidates_p, temp);
    return llama_sample_token(ctx, &candidates_p);
}

struct LLamaPrivate {
    const std::string modelPath;
    bool modelLoaded;
    int device = -1;
    llama_model *model = nullptr;
    llama_context *ctx = nullptr;
    llama_model_params model_params;
    llama_context_params ctx_params;
    int64_t n_threads = 0;
    std::vector<LLModel::Token> end_tokens;
};

LLamaModel::LLamaModel()
    : d_ptr(new LLamaPrivate) {
    d_ptr->modelLoaded = false;
}

// default hparams (LLaMA 7B)
struct llama_file_hparams {
    uint32_t n_vocab = 32000;
    uint32_t n_embd  = 4096;
    uint32_t n_mult  = 256;
    uint32_t n_head  = 32;
    uint32_t n_layer = 32;
    uint32_t n_rot   = 64;
    enum llama_ftype ftype = LLAMA_FTYPE_MOSTLY_F16;
};

size_t LLamaModel::requiredMem(const std::string &modelPath, int n_ctx, int ngl) {
    // TODO(cebtenzzre): update to GGUF
    (void)ngl; // FIXME(cetenzzre): use this value
    auto fin = std::ifstream(modelPath, std::ios::binary);
    fin.seekg(0, std::ios_base::end);
    size_t filesize = fin.tellg();
    fin.seekg(0, std::ios_base::beg);
    uint32_t magic = 0;
    fin.read(reinterpret_cast<char*>(&magic), sizeof(magic));
    if (magic != 0x67676a74) return 0;
    uint32_t version = 0;
    fin.read(reinterpret_cast<char*>(&version), sizeof(version));
    llama_file_hparams hparams;
    fin.read(reinterpret_cast<char*>(&hparams.n_vocab), sizeof(hparams.n_vocab));
    fin.read(reinterpret_cast<char*>(&hparams.n_embd), sizeof(hparams.n_embd));
    fin.read(reinterpret_cast<char*>(&hparams.n_head), sizeof(hparams.n_head));
    fin.read(reinterpret_cast<char*>(&hparams.n_layer), sizeof(hparams.n_layer));
    fin.read(reinterpret_cast<char*>(&hparams.n_rot), sizeof(hparams.n_rot));
    fin.read(reinterpret_cast<char*>(&hparams.ftype), sizeof(hparams.ftype));
    const size_t kvcache_element_size = 2; // fp16
    const size_t est_kvcache_size = hparams.n_embd * hparams.n_layer * 2u * n_ctx * kvcache_element_size;
    return filesize + est_kvcache_size;
}

bool LLamaModel::loadModel(const std::string &modelPath, int n_ctx, int ngl)
{
    d_ptr->modelLoaded = false;

    // clean up after previous loadModel()
    if (d_ptr->model) {
        llama_free_model(d_ptr->model);
        d_ptr->model = nullptr;
    }
    if (d_ptr->ctx) {
        llama_free(d_ptr->ctx);
        d_ptr->ctx = nullptr;
    }

    if (n_ctx < 8) {
        std::cerr << "warning: minimum context size is 8, using minimum size.\n";
        n_ctx = 8;
    }

    // -- load the model --

    gpt_params params;

    d_ptr->model_params = llama_model_default_params();

    d_ptr->model_params.use_mmap  = params.use_mmap;
#if defined (__APPLE__)
    d_ptr->model_params.use_mlock = true;
#else
    d_ptr->model_params.use_mlock = params.use_mlock;
#endif

    d_ptr->model_params.progress_callback = &LLModel::staticProgressCallback;
    d_ptr->model_params.progress_callback_user_data = this;

#ifdef GGML_USE_METAL
    if (llama_verbose()) {
        std::cerr << "llama.cpp: using Metal" << std::endl;
    }

    // always fully offload on Metal
    // TODO(cebtenzzre): use this parameter to allow using more than 53% of system RAM to load a model
    d_ptr->model_params.n_gpu_layers = 100;
#elif defined(GGML_USE_KOMPUTE)
    if (d_ptr->device != -1) {
        d_ptr->model_params.main_gpu = d_ptr->device;
        d_ptr->model_params.n_gpu_layers = ngl;
    }
#endif

    d_ptr->model = llama_load_model_from_file_gpt4all(modelPath.c_str(), &d_ptr->model_params);
    if (!d_ptr->model) {
        fflush(stdout);
        d_ptr->device = -1;
        std::cerr << "LLAMA ERROR: failed to load model from " <<  modelPath << std::endl;
        return false;
    }

    const int n_ctx_train = llama_n_ctx_train(d_ptr->model);
    if (n_ctx > n_ctx_train) {
        std::cerr << "warning: model was trained on only " << n_ctx_train << " context tokens ("
                  << n_ctx << " specified)\n";
    }

    // -- initialize the context --

    d_ptr->ctx_params = llama_context_default_params();

    d_ptr->ctx_params.n_ctx   = n_ctx;
    d_ptr->ctx_params.seed    = params.seed;
    d_ptr->ctx_params.type_k  = params.kv_type;
    d_ptr->ctx_params.type_v  = params.kv_type;

    // The new batch API provides space for n_vocab*n_tokens logits. Tell llama.cpp early
    // that we want this many logits so the state serializes consistently.
    d_ptr->ctx_params.logits_all = true;

    d_ptr->n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency());
    d_ptr->ctx_params.n_threads       = d_ptr->n_threads;
    d_ptr->ctx_params.n_threads_batch = d_ptr->n_threads;

    d_ptr->ctx = llama_new_context_with_model(d_ptr->model, d_ptr->ctx_params);
    if (!d_ptr->ctx) {
        fflush(stdout);
        std::cerr << "LLAMA ERROR: failed to init context for model " <<  modelPath << std::endl;
        llama_free_model(d_ptr->model);
        d_ptr->model = nullptr;
        d_ptr->device = -1;
        return false;
    }

    d_ptr->end_tokens = {llama_token_eos(d_ptr->model)};

#ifdef GGML_USE_KOMPUTE
    if (usingGPUDevice() && ggml_vk_has_device()) {
        std::cerr << "llama.cpp: using Vulkan on " << ggml_vk_current_device().name << std::endl;
    }
#endif

    fflush(stdout);
    d_ptr->modelLoaded = true;
    return true;
}

void LLamaModel::setThreadCount(int32_t n_threads) {
    d_ptr->n_threads = n_threads;
    llama_set_n_threads(d_ptr->ctx, n_threads, n_threads);
}

int32_t LLamaModel::threadCount() const {
    return d_ptr->n_threads;
}

LLamaModel::~LLamaModel()
{
    if (d_ptr->ctx) {
        llama_free(d_ptr->ctx);
    }
    llama_free_model(d_ptr->model);
}

bool LLamaModel::isModelLoaded() const
{
    return d_ptr->modelLoaded;
}

size_t LLamaModel::stateSize() const
{
    return llama_get_state_size(d_ptr->ctx);
}

size_t LLamaModel::saveState(uint8_t *dest) const
{
    return llama_copy_state_data(d_ptr->ctx, dest);
}

size_t LLamaModel::restoreState(const uint8_t *src)
{
    // const_cast is required, see: https://github.com/ggerganov/llama.cpp/pull/1540
    return llama_set_state_data(d_ptr->ctx, const_cast<uint8_t*>(src));
}

std::vector<LLModel::Token> LLamaModel::tokenize(PromptContext &ctx, const std::string &str) const
{
    const bool useBOS = ctx.n_past == 0 && (ctx.tokens.empty() || ctx.tokens.front() != llama_token_bos(d_ptr->model));
    std::vector<LLModel::Token> fres(str.size()+4);
    // TODO(cebtenzzre): we may want to use special=true here to process special tokens
    auto fres_len = llama_tokenize(d_ptr->model, str.c_str(), str.length(), fres.data(), fres.size(), useBOS, false);
    fres.resize(fres_len);
    return fres;
}

std::string LLamaModel::tokenToString(Token id) const
{
    return llama_token_to_piece(d_ptr->ctx, id);
}

LLModel::Token LLamaModel::sampleToken(PromptContext &promptCtx) const
{
    const size_t n_prev_toks = std::min((size_t) promptCtx.repeat_last_n, promptCtx.tokens.size());
    return llama_sample_top_p_top_k(d_ptr->ctx,
        promptCtx.tokens.data() + promptCtx.tokens.size() - n_prev_toks,
        n_prev_toks, promptCtx.top_k, promptCtx.top_p, promptCtx.temp,
        promptCtx.repeat_penalty, promptCtx.n_last_batch_tokens - 1);
}

bool LLamaModel::evalTokens(PromptContext &ctx, const std::vector<int32_t> &tokens) const
{
    llama_kv_cache_seq_rm(d_ptr->ctx, 0, ctx.n_past, -1);

    llama_batch batch = llama_batch_init(tokens.size(), 0, 1);

    batch.n_tokens = tokens.size();
    ctx.n_last_batch_tokens = tokens.size();

    for (int32_t i = 0; i < batch.n_tokens; i++) {
        batch.token   [i] = tokens[i];
        batch.pos     [i] = ctx.n_past + i;
        batch.n_seq_id[i] = 1;
        batch.seq_id  [i][0] = 0;
        batch.logits  [i] = false;
    }

    // llama_decode will output logits only for the last token of the prompt
    batch.logits[batch.n_tokens - 1] = true;

    int res = llama_decode(d_ptr->ctx, batch);
    llama_batch_free(batch);
    return res == 0;
}

int32_t LLamaModel::contextLength() const
{
    return llama_n_ctx(d_ptr->ctx);
}

const std::vector<LLModel::Token> &LLamaModel::endTokens() const
{
    return d_ptr->end_tokens;
}

std::string get_arch_name(gguf_context *ctx_gguf) {
    std::string arch_name;
    const int kid = gguf_find_key(ctx_gguf, "general.architecture");
    enum gguf_type ktype = gguf_get_kv_type(ctx_gguf, kid);
    if (ktype != (GGUF_TYPE_STRING)) {
        throw std::runtime_error("ERROR: Can't get general architecture from gguf file.");
    }
    return gguf_get_val_str(ctx_gguf, kid);
}

static gguf_context *load_gguf(const char *fname, std::string &arch) {
    struct gguf_init_params params = {
        /*.no_alloc = */ true,
        /*.ctx      = */ nullptr,
    };
    gguf_context *ctx = gguf_init_from_file(fname, params);
    if (!ctx) {
        std::cerr << __func__ << ": gguf_init_from_file failed\n";
        return nullptr;
    }

    int gguf_ver = gguf_get_version(ctx);
    if (gguf_ver > GGUF_VER_MAX) {
        std::cerr << __func__ << ": unsupported gguf version: " << gguf_ver << "\n";
        gguf_free(ctx);
        return nullptr;
    }

    arch = get_arch_name(ctx);
    return ctx;
}

static int32_t get_arch_key_u32(std::string const &modelPath, std::string const &archKey) {
    std::string arch;
    auto * ctx = load_gguf(modelPath.c_str(), arch);

    int32_t value = -1;
    if (ctx) {
        auto key = arch + "." + archKey;
        int keyidx = gguf_find_key(ctx, key.c_str());
        if (keyidx != -1) {
            value = gguf_get_val_u32(ctx, keyidx);
        } else {
            std::cerr << __func__ << ": " << key << "not found in " << modelPath << "\n";
        }
    }

    gguf_free(ctx);
    return value;
}

int32_t LLamaModel::maxContextLength(std::string const &modelPath) const
{
    return get_arch_key_u32(modelPath, "context_length");
}

int32_t LLamaModel::layerCount(std::string const &modelPath) const
{
    return get_arch_key_u32(modelPath, "block_count");
}

std::vector<LLModel::GPUDevice> LLamaModel::availableGPUDevices(size_t memoryRequired) const
{
#ifdef GGML_USE_KOMPUTE
    size_t count = 0;
    auto * vkDevices = ggml_vk_available_devices(memoryRequired, &count);

    if (vkDevices) {
        std::vector<LLModel::GPUDevice> devices;
        devices.reserve(count);

        for (size_t i = 0; i < count; ++i) {
            auto & dev = vkDevices[i];
            devices.emplace_back(
                /* index    = */ dev.index,
                /* type     = */ dev.type,
                /* heapSize = */ dev.heapSize,
                /* name     = */ dev.name,
                /* vendor   = */ dev.vendor
            );
            ggml_vk_device_destroy(&dev);
        }

        free(vkDevices);
        return devices;
    }
#else
    std::cerr << __func__ << ": built without Kompute\n";
#endif

    return {};
}

bool LLamaModel::initializeGPUDevice(size_t memoryRequired, const std::string &name) const
{
#if defined(GGML_USE_KOMPUTE)
    ggml_vk_device device;
    bool ok = ggml_vk_get_device(&device, memoryRequired, name.c_str());
    if (ok) {
        d_ptr->device = device.index;
        return true;
    }
#else
    (void)memoryRequired;
    (void)name;
#endif
    return false;
}

bool LLamaModel::initializeGPUDevice(int device, std::string *unavail_reason) const
{
#if defined(GGML_USE_KOMPUTE)
    (void)unavail_reason;
    d_ptr->device = device;
    return true;
#else
    (void)device;
    if (unavail_reason) {
        *unavail_reason = "built without Kompute";
    }
    return false;
#endif
}

bool LLamaModel::hasGPUDevice()
{
#if defined(GGML_USE_KOMPUTE)
    return d_ptr->device != -1;
#else
    return false;
#endif
}

bool LLamaModel::usingGPUDevice()
{
#if defined(GGML_USE_KOMPUTE)
    return hasGPUDevice() && d_ptr->model_params.n_gpu_layers > 0;
#elif defined(GGML_USE_METAL)
    return true;
#else
    return false;
#endif
}

#if defined(_WIN32)
#define DLL_EXPORT __declspec(dllexport)
#else
#define DLL_EXPORT __attribute__ ((visibility ("default")))
#endif

extern "C" {
DLL_EXPORT bool is_g4a_backend_model_implementation() {
    return true;
}

DLL_EXPORT const char *get_model_type() {
    return modelType_;
}

DLL_EXPORT const char *get_build_variant() {
    return GGML_BUILD_VARIANT;
}

DLL_EXPORT bool magic_match(const char *fname) {
    std::string arch;
    auto * ctx = load_gguf(fname, arch);

    bool valid = true;

    static const std::vector<const char *> known_arches {
        "baichuan", "bloom", "codeshell", "falcon", "gpt2", "llama", "mpt", "orion", "persimmon", "phi2", "plamo",
        "qwen", "qwen2", "refact", "stablelm", "starcoder"
    };

    if (std::find(known_arches.begin(), known_arches.end(), arch) == known_arches.end()) {
        // not supported by this version of llama.cpp
        if (!(arch == "gptj" || arch == "bert")) { // we support these via other modules
            std::cerr << __func__ << ": unsupported model architecture: " << arch << "\n";
        }
        valid = false;
    }

    gguf_free(ctx);
    return valid;
}

DLL_EXPORT LLModel *construct() {
    llama_log_set(llama_log_callback, nullptr);
    return new LLamaModel;
}
}