emscripten.cpp

#include "ggml.h"
#include "whisper.h"

#include <emscripten.h>
#include <emscripten/bind.h>

#include <atomic>
#include <cassert>
#include <cmath>
#include <cstdio>
#include <cstring>
#include <fstream>
#include <map>
#include <mutex>
#include <string>
#include <thread>
#include <vector>
#include <regex>
#include <random>

std::string to_timestamp(int64_t t) {
    int64_t sec = t/100;
    int64_t msec = t - sec*100;
    int64_t min = sec/60;
    sec = sec - min*60;

    char buf[32];
    snprintf(buf, sizeof(buf), "%02d:%02d.%03d", (int) min, (int) sec, (int) msec);

    return std::string(buf);
}

/////////////////////// GPT-2 BEGIN /////////////////////////
// TODO: move to a separate file

//
// Vocab utils
//

struct gpt_vocab {
    using id    = int32_t;
    using token = std::string;

    std::map<token, id> token_to_id;
    std::map<id, token> id_to_token;
};

void replace(std::string & str, const std::string & needle, const std::string & replacement) {
    size_t pos = 0;
    while ((pos = str.find(needle, pos)) != std::string::npos) {
        str.replace(pos, needle.length(), replacement);
        pos += replacement.length();
    }
}

std::map<std::string, int32_t> json_parse(const std::string & fname) {
    std::map<std::string, int32_t> result;

    // read file into string
    std::string json;
    {
        std::ifstream ifs(fname);
        if (!ifs) {
            fprintf(stderr, "Failed to open %s\n", fname.c_str());
            exit(1);
        }

        json = std::string((std::istreambuf_iterator<char>(ifs)),
                (std::istreambuf_iterator<char>()));
    }

    if (json[0] != '{') {
        return result;
    }

    // parse json
    {
        bool has_key  = false;
        bool in_token = false;

        std::string str_key = "";
        std::string str_val = "";

        int n = json.size();
        for (int i = 1; i < n; ++i) {
            if (!in_token) {
                if (json[i] == ' ') continue;
                if (json[i] == '"') {
                    in_token = true;
                    continue;
                }
            } else {
                if (json[i] == '\\' && i+1 < n) {
                    if (has_key == false) {
                        str_key += json[i];
                    } else {
                        str_val += json[i];
                    }
                    ++i;
                } else if (json[i] == '"') {
                    if (has_key == false) {
                        has_key = true;
                        ++i;
                        while (json[i] == ' ') ++i;
                        ++i; // :
                        while (json[i] == ' ') ++i;
                        if (json[i] != '\"') {
                            while (json[i] != ',' && json[i] != '}') {
                                str_val += json[i++];
                            }
                            has_key = false;
                        } else {
                            in_token = true;
                            continue;
                        }
                    } else {
                        has_key = false;
                    }

                    ::replace(str_key, "\\u0120", " " ); // \u0120 -> space
                    ::replace(str_key, "\\u010a", "\n"); // \u010a -> new line
                    ::replace(str_key, "\\\"",    "\""); // \\\"   -> "

                    try {
                        result[str_key] = std::stoi(str_val);
                    } catch (...) {
                        //fprintf(stderr, "%s: ignoring key '%s' with value '%s'\n", fname.c_str(), str_key.c_str(), str_val.c_str());

                    }
                    str_key = "";
                    str_val = "";
                    in_token = false;
                    continue;
                }
                if (has_key == false) {
                    str_key += json[i];
                } else {
                    str_val += json[i];
                }
            }
        }
    }

    return result;
}

std::vector<gpt_vocab::id> gpt_tokenize(const gpt_vocab & vocab, const std::string & text) {
    std::vector<std::string> words;

    // first split the text into words
    {
        std::string str = text;
        std::string pat = R"('s|'t|'re|'ve|'m|'ll|'d| ?[[:alpha:]]+| ?[[:digit:]]+| ?[^\s[:alpha:][:digit:]]+|\s+(?!\S)|\s+)";

        std::regex re(pat);
        std::smatch m;

        while (std::regex_search(str, m, re)) {
            for (auto x : m) {
                words.push_back(x);
            }
            str = m.suffix();
        }
    }

    // find the longest tokens that form the words:
    std::vector<gpt_vocab::id> tokens;
    for (const auto & word : words) {
        if (word.size() == 0) continue;

        int i = 0;
        int n = word.size();
        while (i < n) {
            int j = n;
            while (j > i) {
                auto it = vocab.token_to_id.find(word.substr(i, j-i));
                if (it != vocab.token_to_id.end()) {
                    tokens.push_back(it->second);
                    i = j;
                    break;
                }
                --j;
            }
            if (i == n) {
                break;
            }
            if (j == i) {
                auto sub = word.substr(i, 1);
                if (vocab.token_to_id.find(sub) != vocab.token_to_id.end()) {
                    tokens.push_back(vocab.token_to_id.at(sub));
                } else {
                    fprintf(stderr, "%s: unknown token '%s'\n", __func__, sub.data());
                }
                ++i;
            }
        }
    }

    return tokens;
}

bool gpt_vocab_init(const std::string & fname, gpt_vocab & vocab) {
    printf("%s: loading vocab from '%s'\n", __func__, fname.c_str());

    vocab.token_to_id = ::json_parse(fname);

    for (const auto & kv : vocab.token_to_id) {
        vocab.id_to_token[kv.second] = kv.first;
    }

    printf("%s: vocab size = %d\n", __func__, (int) vocab.token_to_id.size());

    // print the vocabulary
    //for (auto kv : vocab.token_to_id) {
    //    printf("'%s' -> %d\n", kv.first.data(), kv.second);
    //}

    return true;
}

gpt_vocab::id gpt_sample_top_k_top_p(
        const gpt_vocab & vocab,
        const float * logits,
        int    top_k,
        double top_p,
        double temp,
        std::mt19937 & rng) {
    int n_logits = vocab.id_to_token.size();

    std::vector<std::pair<double, gpt_vocab::id>> logits_id;
    logits_id.reserve(n_logits);

    for (int i = 0; i < n_logits; i++) {
        logits_id.push_back(std::make_pair(logits[i], i));
    }

    // find the top K tokens
    std::partial_sort(
            logits_id.begin(),
            logits_id.begin() + top_k, logits_id.end(),
            [](const std::pair<double, gpt_vocab::id> & a, const std::pair<double, gpt_vocab::id> & b) {
        return a.first > b.first;
    });

    logits_id.resize(top_k);

    // normalize
    {
        double sum = 0.0f;
        for (int i = 0; i < (int)logits_id.size(); i++) {
            sum += logits_id[i].first;
        }

        sum = 1.0/sum;
        for (int i = 0; i < (int)logits_id.size(); i++) {
            logits_id[i].first *= sum;
        }
    }

    if (top_p < 1.0f) {
        {
            double cumsum = 0.0f;
            for (int i = 0; i < top_k; i++) {
                cumsum += logits_id[i].first;
                if (cumsum >= top_p) {
                    logits_id.resize(i+1);
                    break;
                }
            }
        }

        // normalize again
        {
            double sum = 0.0f;
            for (int i = 0; i < (int)logits_id.size(); i++) {
                sum += logits_id[i].first;
            }

            sum = 1.0/sum;
            for (int i = 0; i < (int)logits_id.size(); i++) {
                logits_id[i].first *= sum;
            }
        }
    }

    //printf("\n");
    //for (int i = 0; i < (int)logits_id.size(); i++) {
    //    printf("%d: '%s' %f\n", i, vocab.id_to_token.at(logits_id[i].second).c_str(), logits_id[i].first);
    //}
    //exit(0);

    // sample from the obtained distribution
    std::vector<double> probs;
    probs.reserve(logits_id.size());

    for (int i = 0; i < (int) logits_id.size(); i++) {
        probs.push_back(logits_id[i].first);
    }

    std::discrete_distribution<> dist(probs.begin(), probs.end());
    int idx = dist(rng);

    return logits_id[idx].second;
}

// default hparams (GPT-2 117M)
struct gpt2_hparams {
    int32_t n_vocab = 50257;
    int32_t n_ctx   = 1024;
    int32_t n_embd  = 768;
    int32_t n_head  = 12;
    int32_t n_layer = 12;
    int32_t f16     = 1;
};

struct gpt2_layer {
    // normalization
    struct ggml_tensor * ln_1_g;
    struct ggml_tensor * ln_1_b;

    struct ggml_tensor * ln_2_g;
    struct ggml_tensor * ln_2_b;

    // attention
    struct ggml_tensor * c_attn_attn_w;
    struct ggml_tensor * c_attn_attn_b;

    struct ggml_tensor * c_attn_proj_w;
    struct ggml_tensor * c_attn_proj_b;

    // mlp
    struct ggml_tensor * c_mlp_fc_w;
    struct ggml_tensor * c_mlp_fc_b;

    struct ggml_tensor * c_mlp_proj_w_trans; // transposed for efficiency
    struct ggml_tensor * c_mlp_proj_b;
};

struct gpt2_model {
    gpt2_hparams hparams;

    // normalization
    struct ggml_tensor * ln_f_g;
    struct ggml_tensor * ln_f_b;

    struct ggml_tensor * wte; // position embedding
    struct ggml_tensor * wpe; //    token embedding

    std::vector<gpt2_layer> layers;

    // key + value memory
    struct ggml_tensor * memory_k;
    struct ggml_tensor * memory_v;

    //
    struct ggml_context * ctx;
    std::map<std::string, struct ggml_tensor *> tensors;
};

// load the model's weights from a file
bool gpt2_model_load(const std::string & fname, gpt2_model & model, gpt_vocab & vocab) {
    printf("%s: loading model from '%s'\n", __func__, fname.c_str());

    auto fin = std::ifstream(fname, std::ios::binary);
    if (!fin) {
        fprintf(stderr, "%s: failed to open '%s'\n", __func__, fname.c_str());
        return false;
    }

    // verify magic
    {
        uint32_t magic;
        fin.read((char *) &magic, sizeof(magic));
        if (magic != 0x67676d6c) {
            fprintf(stderr, "%s: invalid model file '%s' (bad magic)\n", __func__, fname.c_str());
            return false;
        }
    }

    // load hparams
    {
        auto & hparams = model.hparams;

        fin.read((char *) &hparams.n_vocab, sizeof(hparams.n_vocab));
        fin.read((char *) &hparams.n_ctx,   sizeof(hparams.n_ctx));
        fin.read((char *) &hparams.n_embd,  sizeof(hparams.n_embd));
        fin.read((char *) &hparams.n_head,  sizeof(hparams.n_head));
        fin.read((char *) &hparams.n_layer, sizeof(hparams.n_layer));
        fin.read((char *) &hparams.f16,     sizeof(hparams.f16));

        printf("%s: n_vocab = %d\n", __func__, hparams.n_vocab);
        printf("%s: n_ctx   = %d\n", __func__, hparams.n_ctx);
        printf("%s: n_embd  = %d\n", __func__, hparams.n_embd);
        printf("%s: n_head  = %d\n", __func__, hparams.n_head);
        printf("%s: n_layer = %d\n", __func__, hparams.n_layer);
        printf("%s: f16     = %d\n", __func__, hparams.f16);
    }

    // load vocab
    {
        int32_t n_vocab = 0;
        fin.read((char *) &n_vocab, sizeof(n_vocab));

        if (n_vocab != model.hparams.n_vocab) {
            fprintf(stderr, "%s: invalid model file '%s' (bad vocab size %d != %d)\n",
                    __func__, fname.c_str(), n_vocab, model.hparams.n_vocab);
            return false;
        }

        std::string word;
        for (int i = 0; i < n_vocab; i++) {
            uint32_t len;
            fin.read((char *) &len, sizeof(len));

            word.resize(len);
            fin.read((char *) word.data(), len);

            vocab.token_to_id[word] = i;
            vocab.id_to_token[i] = word;
        }
    }

    // for the big tensors, we have the option to store the data in 16-bit floats
    // in order to save memory and also to speed up the computation
    const ggml_type wtype = model.hparams.f16 ? GGML_TYPE_F16 : GGML_TYPE_F32;

    auto & ctx = model.ctx;

    size_t ctx_size = 0;

    {
        const auto & hparams = model.hparams;

        const int n_embd  = hparams.n_embd;
        const int n_layer = hparams.n_layer;
        const int n_ctx   = hparams.n_ctx;
        const int n_vocab = hparams.n_vocab;

        ctx_size += n_embd*ggml_type_size(GGML_TYPE_F32); // ln_f_g
        ctx_size += n_embd*ggml_type_size(GGML_TYPE_F32); // ln_f_b

        ctx_size += n_vocab*n_embd*ggml_type_size(wtype);         // wte
        ctx_size +=   n_ctx*n_embd*ggml_type_size(GGML_TYPE_F32); // wpe

        ctx_size += n_layer*(n_embd*ggml_type_size(GGML_TYPE_F32)); // ln_1_g
        ctx_size += n_layer*(n_embd*ggml_type_size(GGML_TYPE_F32)); // ln_1_b

        ctx_size += n_layer*(n_embd*ggml_type_size(GGML_TYPE_F32)); // ln_2_g
        ctx_size += n_layer*(n_embd*ggml_type_size(GGML_TYPE_F32)); // ln_2_b

        ctx_size += n_layer*(3*n_embd*n_embd*ggml_type_size(wtype));         // c_attn_attn_w
        ctx_size += n_layer*(       3*n_embd*ggml_type_size(GGML_TYPE_F32)); // c_attn_attn_b

        ctx_size += n_layer*(n_embd*n_embd*ggml_type_size(wtype));           // c_attn_proj_w
        ctx_size += n_layer*(       n_embd*ggml_type_size(GGML_TYPE_F32));   // c_attn_proj_b

        ctx_size += n_layer*(4*n_embd*n_embd*ggml_type_size(wtype));         // c_mlp_fc_w
        ctx_size += n_layer*(       4*n_embd*ggml_type_size(GGML_TYPE_F32)); // c_mlp_fc_b

        ctx_size += n_layer*(4*n_embd*n_embd*ggml_type_size(wtype));         // c_mlp_proj_w
        ctx_size += n_layer*(         n_embd*ggml_type_size(GGML_TYPE_F32)); // c_mlp_proj_b

        ctx_size += n_ctx*n_layer*n_embd*ggml_type_size(GGML_TYPE_F32); // memory_k
        ctx_size += n_ctx*n_layer*n_embd*ggml_type_size(GGML_TYPE_F32); // memory_v

        ctx_size += (6 + 12*n_layer)*256; // object overhead

        printf("%s: ggml ctx size = %6.2f MB\n", __func__, ctx_size/(1024.0*1024.0));
    }

    // create the ggml context
    {
        struct ggml_init_params params = {
            .mem_size   = ctx_size,
            .mem_buffer = NULL,
        };

        model.ctx = ggml_init(params);
        if (!model.ctx) {
            fprintf(stderr, "%s: ggml_init() failed\n", __func__);
            return false;
        }
    }

    // prepare memory for the weights
    {
        const auto & hparams = model.hparams;

        const int n_embd  = hparams.n_embd;
        const int n_layer = hparams.n_layer;
        const int n_ctx   = hparams.n_ctx;
        const int n_vocab = hparams.n_vocab;

        model.layers.resize(n_layer);

        model.ln_f_g = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embd);
        model.ln_f_b = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embd);

        model.wte = ggml_new_tensor_2d(ctx, wtype,         n_embd, n_vocab);
        model.wpe = ggml_new_tensor_2d(ctx, GGML_TYPE_F32, n_embd, n_ctx);

        // map by name
        model.tensors["model/ln_f/g"] = model.ln_f_g;
        model.tensors["model/ln_f/b"] = model.ln_f_b;

        model.tensors["model/wte"] = model.wte;
        model.tensors["model/wpe"] = model.wpe;

        for (int i = 0; i < n_layer; ++i) {
            auto & layer = model.layers[i];

            layer.ln_1_g             = ggml_new_tensor_1d(ctx, GGML_TYPE_F32,   n_embd);
            layer.ln_1_b             = ggml_new_tensor_1d(ctx, GGML_TYPE_F32,   n_embd);

            layer.ln_2_g             = ggml_new_tensor_1d(ctx, GGML_TYPE_F32,   n_embd);
            layer.ln_2_b             = ggml_new_tensor_1d(ctx, GGML_TYPE_F32,   n_embd);

            layer.c_attn_attn_w      = ggml_new_tensor_2d(ctx, wtype,         3*n_embd, n_embd);
            layer.c_attn_attn_b      = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 3*n_embd);

            layer.c_attn_proj_w      = ggml_new_tensor_2d(ctx, wtype,           n_embd, n_embd);
            layer.c_attn_proj_b      = ggml_new_tensor_1d(ctx, GGML_TYPE_F32,   n_embd);

            layer.c_mlp_fc_w         = ggml_new_tensor_2d(ctx, wtype,         4*n_embd, n_embd);
            layer.c_mlp_fc_b         = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 4*n_embd);

            layer.c_mlp_proj_w_trans = ggml_new_tensor_2d(ctx, wtype,         4*n_embd, n_embd);
            layer.c_mlp_proj_b       = ggml_new_tensor_1d(ctx, GGML_TYPE_F32,   n_embd);

            // map by name
            model.tensors["model/h" + std::to_string(i) + "/ln_1/g"]        = layer.ln_1_g;
            model.tensors["model/h" + std::to_string(i) + "/ln_1/b"]        = layer.ln_1_b;

            model.tensors["model/h" + std::to_string(i) + "/ln_2/g"]        = layer.ln_2_g;
            model.tensors["model/h" + std::to_string(i) + "/ln_2/b"]        = layer.ln_2_b;

            model.tensors["model/h" + std::to_string(i) + "/attn/c_attn/w"] = layer.c_attn_attn_w;
            model.tensors["model/h" + std::to_string(i) + "/attn/c_attn/b"] = layer.c_attn_attn_b;

            model.tensors["model/h" + std::to_string(i) + "/attn/c_proj/w"] = layer.c_attn_proj_w;
            model.tensors["model/h" + std::to_string(i) + "/attn/c_proj/b"] = layer.c_attn_proj_b;

            model.tensors["model/h" + std::to_string(i) + "/mlp/c_fc/w"]    = layer.c_mlp_fc_w;
            model.tensors["model/h" + std::to_string(i) + "/mlp/c_fc/b"]    = layer.c_mlp_fc_b;

            model.tensors["model/h" + std::to_string(i) + "/mlp/c_proj/w"]  = layer.c_mlp_proj_w_trans;
            model.tensors["model/h" + std::to_string(i) + "/mlp/c_proj/b"]  = layer.c_mlp_proj_b;
        }
    }

    // key + value memory
    {
        const auto & hparams = model.hparams;

        const int n_embd  = hparams.n_embd;
        const int n_layer = hparams.n_layer;
        const int n_ctx   = hparams.n_ctx;

        const int n_mem      = n_layer*n_ctx;
        const int n_elements = n_embd*n_mem;

        model.memory_k = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_elements);
        model.memory_v = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_elements);

        const size_t memory_size = ggml_nbytes(model.memory_k) + ggml_nbytes(model.memory_v);

        printf("%s: memory size = %8.2f MB, n_mem = %d\n", __func__, memory_size/1024.0/1024.0, n_mem);
    }

    // load weights
    {
        size_t total_size = 0;

        while (true) {
            int32_t n_dims;
            int32_t length;
            int32_t ftype;

            fin.read(reinterpret_cast<char *>(&n_dims), sizeof(n_dims));
            fin.read(reinterpret_cast<char *>(&length), sizeof(length));
            fin.read(reinterpret_cast<char *>(&ftype),  sizeof(ftype));

            if (fin.eof()) {
                break;
            }

            int32_t nelements = 1;
            int32_t ne[2] = { 1, 1 };
            for (int i = 0; i < n_dims; ++i) {
                fin.read(reinterpret_cast<char *>(&ne[i]), sizeof(ne[i]));
                nelements *= ne[i];
            }

            std::string name(length, 0);
            fin.read(&name[0], length);

            if (model.tensors.find(name.data()) == model.tensors.end()) {
                fprintf(stderr, "%s: unknown tensor '%s' in model file\n", __func__, name.data());
                return false;
            }

            auto tensor = model.tensors[name.data()];
            if (ggml_nelements(tensor) != nelements) {
                fprintf(stderr, "%s: tensor '%s' has wrong size in model file\n", __func__, name.data());
                return false;
            }

            if (tensor->ne[0] != ne[0] || tensor->ne[1] != ne[1]) {
                fprintf(stderr, "%s: tensor '%s' has wrong shape in model file: got [%d, %d], expected [%d, %d]\n",
                        __func__, name.data(), tensor->ne[0], tensor->ne[1], ne[0], ne[1]);
                return false;
            }

            const size_t bpe = (ftype == 0) ? sizeof(float) : sizeof(ggml_fp16_t);

            if (nelements*bpe != ggml_nbytes(tensor)) {
                fprintf(stderr, "%s: tensor '%s' has wrong size in model file: got %zu, expected %zu\n",
                        __func__, name.data(), ggml_nbytes(tensor), nelements*bpe);
                return false;
            }

            fin.read(reinterpret_cast<char *>(tensor->data), ggml_nbytes(tensor));

            //printf("%24s - [%5d, %5d], type = %6s, %6.2f MB\n", name.data(), ne[0], ne[1], ftype == 0 ? "float" : "f16", ggml_nbytes(tensor)/1024.0/1024.0);
            total_size += ggml_nbytes(tensor);
        }

        printf("%s: model size  = %8.2f MB\n", __func__, total_size/1024.0/1024.0);
    }

    fin.close();

    return true;
}

// evaluate the transformer
//
//   - model:     the model
//   - n_threads: number of threads to use
//   - n_past:    the context size so far
//   - embd_inp:  the embeddings of the tokens in the context
//   - embd_w:    the predicted probabilities of the next token
//
bool gpt2_eval(
        const gpt2_model & model,
        const int n_threads,
        const int n_past,
        const std::vector<gpt_vocab::id> & embd_inp,
              std::vector<float>         & embd_w,
              size_t                     & mem_per_token) {
    const int N = embd_inp.size();

    const auto & hparams = model.hparams;

    const int n_embd  = hparams.n_embd;
    const int n_layer = hparams.n_layer;
    const int n_ctx   = hparams.n_ctx;
    const int n_head  = hparams.n_head;
    const int n_vocab = hparams.n_vocab;

    static size_t buf_size = 512u*1024*1024;
    static void * buf = malloc(buf_size);

    if (mem_per_token > 0 && mem_per_token*N > buf_size) {
        const size_t buf_size_new = 1.1*(mem_per_token*N); // add 10% to account for ggml object overhead
        printf("\n%s: reallocating buffer from %zu to %zu bytes\n", __func__, buf_size, buf_size_new);

        // reallocate
        buf_size = buf_size_new;
        buf = realloc(buf, buf_size);
        if (buf == nullptr) {
            fprintf(stderr, "%s: failed to allocate %zu bytes\n", __func__, buf_size);
            return false;
        }
    }

    struct ggml_init_params params = {
        .mem_size   = buf_size,
        .mem_buffer = buf,
    };

    struct ggml_context * ctx0 = ggml_init(params);
    struct ggml_cgraph gf = { .n_threads = n_threads };

    struct ggml_tensor * embd = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, N);
    memcpy(embd->data, embd_inp.data(), N*ggml_element_size(embd));

    struct ggml_tensor * position = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, N);
    for (int i = 0; i < N; ++i) {
        ((int32_t *) position->data)[i] = n_past + i;
    }

    // wte + wpe
    struct ggml_tensor * inpL =
        ggml_add(ctx0,
                ggml_get_rows(ctx0, model.wte, embd),
                ggml_get_rows(ctx0, model.wpe, position));

    for (int il = 0; il < n_layer; ++il) {
        struct ggml_tensor * cur;

        // norm
        {
            // [ 768, N]
            cur = ggml_norm(ctx0, inpL);

            // cur = ln_1_g*cur + ln_1_b
            // [ 768, N]
            cur = ggml_add(ctx0,
                    ggml_mul(ctx0,
                        ggml_repeat(ctx0, model.layers[il].ln_1_g, cur),
                        cur),
                    ggml_repeat(ctx0, model.layers[il].ln_1_b, cur));
        }

        // attn
        // [2304, 768] - model.layers[il].c_attn_attn_w
        // [2304,   1] - model.layers[il].c_attn_attn_b
        // [ 768,   N] - cur (in)
        // [2304,   N] - cur (out)
        //
        // cur = attn_w*cur + attn_b
        // [2304, N]
        {
            cur = ggml_mul_mat(ctx0,
                    ggml_transpose(ctx0, model.layers[il].c_attn_attn_w),
                    cur);

            cur = ggml_add(ctx0,
                    ggml_repeat(ctx0, model.layers[il].c_attn_attn_b, cur),
                    cur);
        }

        // self-attention
        {
            struct ggml_tensor * Qcur = ggml_view_2d(ctx0, cur, n_embd, N, cur->nb[1], 0*sizeof(float)*n_embd);
            struct ggml_tensor * Kcur = ggml_view_2d(ctx0, cur, n_embd, N, cur->nb[1], 1*sizeof(float)*n_embd);
            struct ggml_tensor * Vcur = ggml_view_2d(ctx0, cur, n_embd, N, cur->nb[1], 2*sizeof(float)*n_embd);

            // store key and value to memory
            if (N >= 1) {
                struct ggml_tensor * k = ggml_view_1d(ctx0, model.memory_k, N*n_embd, (ggml_element_size(model.memory_k)*n_embd)*(il*n_ctx + n_past));
                struct ggml_tensor * v = ggml_view_1d(ctx0, model.memory_v, N*n_embd, (ggml_element_size(model.memory_v)*n_embd)*(il*n_ctx + n_past));

                ggml_build_forward_expand(&gf, ggml_cpy(ctx0, Kcur, k));
                ggml_build_forward_expand(&gf, ggml_cpy(ctx0, Vcur, v));
            }

            // Q = Qcur.contiguous().view(n_embd/n_head, n_head, N).permute(0, 2, 1, 3)
            // [64, N, 12]
            struct ggml_tensor * Q =
                ggml_permute(ctx0,
                        ggml_cpy(ctx0,
                            Qcur,
                            ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, n_embd/n_head, n_head, N)),
                        0, 2, 1, 3);

            // K = Kmem.view(n_embd/n_head, n_head, n_past + N).permute(0, 2, 1, 3)
            // [64, n_past + N, 12]
            struct ggml_tensor * K =
                ggml_permute(ctx0,
                        ggml_reshape_3d(ctx0,
                            ggml_view_1d(ctx0, model.memory_k, (n_past + N)*n_embd, il*n_ctx*ggml_element_size(model.memory_k)*n_embd),
                            n_embd/n_head, n_head, n_past + N),
                        0, 2, 1, 3);

            // GG: flash attention
            //struct ggml_tensor * V =
            //    ggml_cpy(ctx0,
            //            ggml_permute(ctx0,
            //                ggml_reshape_3d(ctx0,
            //                    ggml_view_1d(ctx0, model.memory_v, (n_past + N)*n_embd, il*n_ctx*ggml_element_size(model.memory_v)*n_embd),
            //                    n_embd/n_head, n_head, n_past + N),
            //                1, 2, 0, 3),
            //            ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, n_past + N, n_embd/n_head, n_head));

            //struct ggml_tensor * KQV = ggml_flash_attn(ctx0, Q, K, V, true);

            // K * Q
            // [n_past + N, N, 12]
            struct ggml_tensor * KQ = ggml_mul_mat(ctx0, K, Q);

            // KQ_scaled = KQ / sqrt(n_embd/n_head)
            // [n_past + N, N, 12]
            struct ggml_tensor * KQ_scaled =
                ggml_scale(ctx0,
                        KQ,
                        ggml_new_f32(ctx0, 1.0f/sqrt(float(n_embd)/n_head))
                        );

            // KQ_masked = mask_past(KQ_scaled)
            // [n_past + N, N, 12]
            struct ggml_tensor * KQ_masked = ggml_diag_mask_inf(ctx0, KQ_scaled, n_past);

            // KQ = soft_max(KQ_masked)
            // [n_past + N, N, 12]
            struct ggml_tensor * KQ_soft_max = ggml_soft_max(ctx0, KQ_masked);

            // V_trans = Vmem.view(n_embd/n_head, n_head, n_past + N).permute(1, 2, 0, 3).contiguous()
            // [n_past + N, 64, 12]
            struct ggml_tensor * V_trans =
                ggml_permute(ctx0,
                        ggml_reshape_3d(ctx0,
                            ggml_view_1d(ctx0, model.memory_v, (n_past + N)*n_embd, il*n_ctx*ggml_element_size(model.memory_v)*n_embd),
                            n_embd/n_head, n_head, n_past + N),
                        1, 2, 0, 3);

            // KQV = transpose(V) * KQ_soft_max
            // [64, N, 12]
            struct ggml_tensor * KQV = ggml_mul_mat(ctx0, V_trans, KQ_soft_max);

            // KQV_merged = KQV.permute(0, 2, 1, 3)
            // [64, 12, N]
            struct ggml_tensor * KQV_merged = ggml_permute(ctx0, KQV, 0, 2, 1, 3);

            // cur = KQV_merged.contiguous().view(n_embd, N)
            // [768, N]
            cur = ggml_cpy(ctx0,
                    KQV_merged,
                    ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd, N));
        }

        // projection
        // [ 768, 768] - model.layers[il].c_attn_proj_w
        // [ 768,   1] - model.layers[il].c_attn_proj_b
        // [ 768,   N] - cur (in)
        // [ 768,   N] - cur (out)
        //
        // cur = proj_w*cur + proj_b
        // [768, N]
        {
            cur = ggml_mul_mat(ctx0,
                    ggml_transpose(ctx0, model.layers[il].c_attn_proj_w),
                    cur);

            cur = ggml_add(ctx0,
                    ggml_repeat(ctx0, model.layers[il].c_attn_proj_b, cur),
                    cur);
        }

        // add the input
        cur = ggml_add(ctx0, cur, inpL);

        struct ggml_tensor * inpFF = cur;

        // feed-forward network
        {
            // norm
            {
                cur = ggml_norm(ctx0, inpFF);

                // cur = ln_2_g*cur + ln_2_b
                // [ 768, N]
                cur = ggml_add(ctx0,
                        ggml_mul(ctx0,
                            ggml_repeat(ctx0, model.layers[il].ln_2_g, cur),
                            cur),
                        ggml_repeat(ctx0, model.layers[il].ln_2_b, cur));
            }

            // fully connected
            // [3072, 768] - model.layers[il].c_mlp_fc_w
            // [3072,   1] - model.layers[il].c_mlp_fc_b
            // [ 768,   N] - cur (in)
            // [3072,   N] - cur (out)
            //
            // cur = fc_w*cur + fc_b
            // [3072, N]
            cur = ggml_mul_mat(ctx0,
                    ggml_transpose(ctx0, model.layers[il].c_mlp_fc_w),
                    cur);

            cur = ggml_add(ctx0,
                    ggml_repeat(ctx0, model.layers[il].c_mlp_fc_b, cur),
                    cur);

            // GELU activation
            // [3072, N]
            cur = ggml_gelu(ctx0, cur);

            // projection
            // [ 768, 3072] - model.layers[il].c_mlp_proj_w
            // [ 768,    1] - model.layers[il].c_mlp_proj_b
            // [3072,    N] - cur (in)
            // [ 768,    N] - cur (out)
            //
            // cur = proj_w*cur + proj_b
            // [768, N]
            cur = ggml_mul_mat(ctx0,
                    model.layers[il].c_mlp_proj_w_trans,
                    cur);

            cur = ggml_add(ctx0,
                    ggml_repeat(ctx0, model.layers[il].c_mlp_proj_b, cur),
                    cur);
        }

        // input for next layer
        inpL = ggml_add(ctx0, cur, inpFF);
    }

    // norm
    {
        // [ 768, N]
        inpL = ggml_norm(ctx0, inpL);

        // inpL = ln_f_g*inpL + ln_f_b
        // [ 768, N]
        inpL = ggml_add(ctx0,
                ggml_mul(ctx0,
                    ggml_repeat(ctx0, model.ln_f_g, inpL),
                    inpL),
                ggml_repeat(ctx0, model.ln_f_b, inpL));
    }

    // inpL = WTE * inpL
    // [ 768, 50257] - model.wte
    // [ 768, N]     - inpL
    inpL = ggml_mul_mat(ctx0, model.wte, inpL);

    // logits -> probs
    inpL = ggml_soft_max(ctx0, inpL);

    // run the computation
    ggml_build_forward_expand(&gf, inpL);
    ggml_graph_compute       (ctx0, &gf);

    //if (n_past%100 == 0) {
    //    ggml_graph_print   (&gf);
    //    ggml_graph_dump_dot(&gf, NULL, "gpt-2.dot");
    //}

    //embd_w.resize(n_vocab*N);
    //memcpy(embd_w.data(), ggml_get_data(inpL), sizeof(float)*n_vocab*N);

    // return result for just the last token
    embd_w.resize(n_vocab);
    memcpy(embd_w.data(), (float *) ggml_get_data(inpL) + (n_vocab*(N-1)), sizeof(float)*n_vocab);

    if (mem_per_token == 0) {
        mem_per_token = ggml_used_mem(ctx0)/N;
    }
    //printf("used_mem = %zu\n", ggml_used_mem(ctx0));

    ggml_free(ctx0);

    return true;
}

/////////////////////////////// GPT-2 END ////////////////////////////////

constexpr int N_THREAD = 8;

struct gpt2_state {
    std::string prompt_base = R"(Hello, how are you?
I'm fine, thanks. How are you?
Thanks, I'm fine too. What are you doing?
I'm just sitting here.
It's a lovely day, isn't it?
Yes, it is.
Did you know that I'm a robot?
I wasn't aware of that.
)";

    std::mt19937 rng;

    gpt_vocab vocab;
    gpt2_model model;

    int32_t n_threads = std::min(N_THREAD, (int) std::thread::hardware_concurrency());
    int32_t n_predict = 32; // new tokens to predict

    // sampling parameters
    int32_t top_k = 40;
    float   top_p = 0.9f;
    float   temp  = 1.0f;
};

struct gpt2_state g_gpt2;

std::vector<float> g_pcmf32;

std::vector<struct whisper_context *> g_contexts(4, nullptr);

std::mutex g_mutex;
std::thread g_worker;
std::atomic<bool> g_running(false);

bool g_force_speak = false;
std::string g_text_to_speak = "";
std::string g_status = "";
std::string g_status_forced = "";

std::string gpt2_gen_text(const std::string & prompt) {
    int n_past = 0;

    std::vector<float> embd_w;

    // tokenize the prompt
    std::vector<gpt_vocab::id> embd_inp = ::gpt_tokenize(g_gpt2.vocab, prompt);

    g_gpt2.n_predict = std::min(g_gpt2.n_predict, g_gpt2.model.hparams.n_ctx - (int) embd_inp.size());

    std::vector<gpt_vocab::id> embd = embd_inp;

    size_t mem_per_token = 3000000;

    std::string result;

    for (int i = embd.size(); i < embd_inp.size() + g_gpt2.n_predict; i++) {
        // predict
        if (embd.size() > 0) {
            if (!gpt2_eval(g_gpt2.model, g_gpt2.n_threads, n_past, embd, embd_w, mem_per_token)) {
                printf("gpt-2: failed to generate text\n");
                return "";
            }
        }

        n_past += embd.size();
        embd.clear();

        {
            // sample next token
            const int   top_k = g_gpt2.top_k;
            const float top_p = g_gpt2.top_p;
            const float temp  = g_gpt2.temp;

            const int n_vocab = g_gpt2.model.hparams.n_vocab;

            const gpt_vocab::id id = gpt_sample_top_k_top_p(g_gpt2.vocab, embd_w.data() + (embd_w.size() - n_vocab), top_k, top_p, temp, g_gpt2.rng);

            // add it to the context
            embd.push_back(id);
        }

        result += g_gpt2.vocab.id_to_token[embd[0]];

        // end of text token
        if (embd.back() == 50256 ||
            g_gpt2.vocab.id_to_token[embd.back()] == "." ||
            g_gpt2.vocab.id_to_token[embd.back()] == "!" ||
            g_gpt2.vocab.id_to_token[embd.back()] == "?") {
            break;
        }
    }

    return result;
}

void talk_set_status(const std::string & status) {
    std::lock_guard<std::mutex> lock(g_mutex);
    g_status = status;
}

void talk_main(size_t index) {
    talk_set_status("loading data ...");

    struct whisper_full_params wparams = whisper_full_default_params(whisper_sampling_strategy::WHISPER_SAMPLING_GREEDY);

    wparams.n_threads            = std::min(N_THREAD, (int) std::thread::hardware_concurrency());
    wparams.offset_ms            = 0;
    wparams.translate            = false;
    wparams.no_context           = true;
    wparams.single_segment       = true;
    wparams.print_realtime       = false;
    wparams.print_progress       = false;
    wparams.print_timestamps     = true;
    wparams.print_special_tokens = false;

    wparams.max_tokens           = 32;
    wparams.audio_ctx            = 768;

    wparams.language             = "en";

    g_gpt2.rng = std::mt19937(time(NULL));

    // load the model
    {
        const int64_t t_start_us = ggml_time_us();

        if (!gpt2_model_load("gpt-2.bin", g_gpt2.model, g_gpt2.vocab)) {
            fprintf(stderr, "%s: failed to load model from '%s'\n", __func__, "gpt-2.bin");
            return;
        }

        const int64_t t_load_us = ggml_time_us() - t_start_us;

        printf("gpt-2: model loaded in %d ms\n", (int) (t_load_us/1000));
    }

    printf("talk: using %d threads\n", N_THREAD);

    std::vector<float> pcmf32;

    auto & ctx = g_contexts[index];

    const int64_t step_samples = 2*WHISPER_SAMPLE_RATE;
    const int64_t step_ms = (step_samples*1000)/WHISPER_SAMPLE_RATE;
    const int64_t window_samples = 9*WHISPER_SAMPLE_RATE;

    auto t_last = std::chrono::high_resolution_clock::now();

    talk_set_status("listening ...");

    while (g_running) {

        const auto t_now = std::chrono::high_resolution_clock::now();
        if (std::chrono::duration_cast<std::chrono::milliseconds>(t_now - t_last).count() < step_ms) {
            {
                std::lock_guard<std::mutex> lock(g_mutex);
                g_pcmf32.clear();
            }
            std::this_thread::sleep_for(std::chrono::milliseconds(10));
            continue;
        }

        talk_set_status("listening ...");

        {
            std::unique_lock<std::mutex> lock(g_mutex);

            if (g_pcmf32.size() < step_samples) {
                lock.unlock();

                std::this_thread::sleep_for(std::chrono::milliseconds(10));

                continue;
            }

            pcmf32 = std::vector<float>(g_pcmf32.end() - std::min((int64_t) g_pcmf32.size(), window_samples), g_pcmf32.end());
        }

        // if energy in during last second is above threshold, then skip
        {
            float energy_all = 0.0f;
            float energy_1s  = 0.0f;

            for (size_t i = 0; i < pcmf32.size(); i++) {
                energy_all += fabsf(pcmf32[i]);

                if (i >= pcmf32.size() - WHISPER_SAMPLE_RATE) {
                    energy_1s += fabsf(pcmf32[i]);
                }
            }

            energy_all /= pcmf32.size();
            energy_1s  /= WHISPER_SAMPLE_RATE;

            if (energy_1s > 0.1f*energy_all && !g_force_speak) {
                std::this_thread::sleep_for(std::chrono::milliseconds(10));
                continue;
            }
        }

        talk_set_status("processing ...");

        g_force_speak = false;

        t_last = t_now;

        {
            const auto t_start = std::chrono::high_resolution_clock::now();

            int ret = whisper_full(ctx, wparams, pcmf32.data(), pcmf32.size());
            if (ret != 0) {
                printf("whisper_full() failed: %d\n", ret);
                break;
            }

            const auto t_end = std::chrono::high_resolution_clock::now();

            printf("whisper_full() returned %d in %f seconds\n", ret, std::chrono::duration<double>(t_end - t_start).count());
        }

        {
            std::string text_heard;

            const int n_segments = whisper_full_n_segments(ctx);
            for (int i = n_segments - 1; i < n_segments; ++i) {
                const char * text = whisper_full_get_segment_text(ctx, i);

                const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
                const int64_t t1 = whisper_full_get_segment_t1(ctx, i);

                printf ("[%s --> %s]  %s\n", to_timestamp(t0).c_str(), to_timestamp(t1).c_str(), text);

                text_heard += text;
            }

            // remove text between brackets using regex
            {
                std::regex re("\\[.*?\\]");
                text_heard = std::regex_replace(text_heard, re, "");
            }

            // remove text between brackets using regex
            {
                std::regex re("\\(.*?\\)");
                text_heard = std::regex_replace(text_heard, re, "");
            }

            // remove all characters, except for letters, numbers, punctuation and ':', '\'', '-', ' '
            text_heard = std::regex_replace(text_heard, std::regex("[^a-zA-Z0-9\\.,\\?!\\s\\:\\'\\-]"), "");

            // take first line
            text_heard = text_heard.substr(0, text_heard.find_first_of("\n"));

            // remove leading and trailing whitespace
            text_heard = std::regex_replace(text_heard, std::regex("^\\s+"), "");
            text_heard = std::regex_replace(text_heard, std::regex("\\s+$"), "");

            talk_set_status("'" + text_heard + "' - thinking how to respond ...");

            const std::vector<gpt_vocab::id> tokens = ::gpt_tokenize(g_gpt2.vocab, text_heard);

            printf("whisper: number of tokens: %d, '%s'\n", (int) tokens.size(), text_heard.c_str());

            std::string text_to_speak;
            std::string prompt_base;

            {
                std::lock_guard<std::mutex> lock(g_mutex);
                prompt_base = g_gpt2.prompt_base;
            }

            if (tokens.size() > 0) {
                text_to_speak = gpt2_gen_text(prompt_base + text_heard + "\n");
                text_to_speak = std::regex_replace(text_to_speak, std::regex("[^a-zA-Z0-9\\.,\\?!\\s\\:\\'\\-]"), "");
                text_to_speak = text_to_speak.substr(0, text_to_speak.find_first_of("\n"));

                std::lock_guard<std::mutex> lock(g_mutex);

                // remove first 2 lines of base prompt
                {
                    const size_t pos = prompt_base.find_first_of("\n");
                    if (pos != std::string::npos) {
                        prompt_base = prompt_base.substr(pos + 1);
                    }
                }
                {
                    const size_t pos = prompt_base.find_first_of("\n");
                    if (pos != std::string::npos) {
                        prompt_base = prompt_base.substr(pos + 1);
                    }
                }
                prompt_base += text_heard + "\n" + text_to_speak + "\n";
            } else {
                text_to_speak = gpt2_gen_text(prompt_base);
                text_to_speak = std::regex_replace(text_to_speak, std::regex("[^a-zA-Z0-9\\.,\\?!\\s\\:\\'\\-]"), "");
                text_to_speak = text_to_speak.substr(0, text_to_speak.find_first_of("\n"));

                std::lock_guard<std::mutex> lock(g_mutex);

                const size_t pos = prompt_base.find_first_of("\n");
                if (pos != std::string::npos) {
                    prompt_base = prompt_base.substr(pos + 1);
                }
                prompt_base += text_to_speak + "\n";
            }

            printf("gpt-2: %s\n", text_to_speak.c_str());

            //printf("========================\n");
            //printf("gpt-2: prompt_base:\n'%s'\n", prompt_base.c_str());
            //printf("========================\n");

            {
                std::lock_guard<std::mutex> lock(g_mutex);
                t_last = std::chrono::high_resolution_clock::now();
                g_text_to_speak = text_to_speak;
                g_pcmf32.clear();
                g_gpt2.prompt_base = prompt_base;
            }

            talk_set_status("speaking ...");
        }
    }

    if (index < g_contexts.size()) {
        whisper_free(g_contexts[index]);
        g_contexts[index] = nullptr;
    }
}

EMSCRIPTEN_BINDINGS(talk) {
    emscripten::function("init", emscripten::optional_override([](const std::string & path_model) {
        for (size_t i = 0; i < g_contexts.size(); ++i) {
            if (g_contexts[i] == nullptr) {
                g_contexts[i] = whisper_init(path_model.c_str());
                if (g_contexts[i] != nullptr) {
                    g_running = true;
                    if (g_worker.joinable()) {
                        g_worker.join();
                    }
                    g_worker = std::thread([i]() {
                        talk_main(i);
                    });

                    return i + 1;
                } else {
                    return (size_t) 0;
                }
            }
        }

        return (size_t) 0;
    }));

    emscripten::function("free", emscripten::optional_override([](size_t index) {
        if (g_running) {
            g_running = false;
        }
    }));

    emscripten::function("set_audio", emscripten::optional_override([](size_t index, const emscripten::val & audio) {
        --index;

        if (index >= g_contexts.size()) {
            return -1;
        }

        if (g_contexts[index] == nullptr) {
            return -2;
        }

        {
            std::lock_guard<std::mutex> lock(g_mutex);
            const int n = audio["length"].as<int>();

            emscripten::val heap = emscripten::val::module_property("HEAPU8");
            emscripten::val memory = heap["buffer"];

            g_pcmf32.resize(n);

            emscripten::val memoryView = audio["constructor"].new_(memory, reinterpret_cast<uintptr_t>(g_pcmf32.data()), n);
            memoryView.call<void>("set", audio);
        }

        return 0;
    }));

    emscripten::function("force_speak", emscripten::optional_override([](size_t index) {
        {
            std::lock_guard<std::mutex> lock(g_mutex);
            g_force_speak = true;
        }
    }));

    emscripten::function("get_text_context", emscripten::optional_override([]() {
        std::string text_context;

        {
            std::lock_guard<std::mutex> lock(g_mutex);
            text_context = g_gpt2.prompt_base;
        }

        return text_context;
    }));

    emscripten::function("get_text_to_speak", emscripten::optional_override([]() {
        std::string text_to_speak;

        {
            std::lock_guard<std::mutex> lock(g_mutex);
            text_to_speak = std::move(g_text_to_speak);
        }

        return text_to_speak;
    }));

    emscripten::function("get_status", emscripten::optional_override([]() {
        std::string status;

        {
            std::lock_guard<std::mutex> lock(g_mutex);
            status = g_status_forced.empty() ? g_status : g_status_forced;
        }

        return status;
    }));

    emscripten::function("set_status", emscripten::optional_override([](const std::string & status) {
        {
            std::lock_guard<std::mutex> lock(g_mutex);
            g_status_forced = status;
        }
    }));

    emscripten::function("set_prompt", emscripten::optional_override([](const std::string & prompt) {
        {
            std::lock_guard<std::mutex> lock(g_mutex);
            g_gpt2.prompt_base = prompt;
        }
    }));
}