main.cpp

#include "ggml.h"

#include "utils.h"

#include <cassert>
#include <cmath>
#include <cstdio>
#include <cstring>
#include <fstream>
#include <map>
#include <string>
#include <vector>

#if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__))
#include <signal.h>
#include <unistd.h>
#elif defined (_WIN32)
#include <signal.h>
#endif

struct bloom_hparams {
    int32_t n_vocab = 65024;
    int32_t n_ctx   = 512;   // this is provided as user input?
    int32_t n_embd  = 8192;
    int32_t n_mult  = 1;
    int32_t n_head  = 128;
    int32_t n_layer = 60;
    int32_t f16     = 1;
};

struct bloom_layer {
    // normalization
    struct ggml_tensor * attention_norm;
    struct ggml_tensor * attention_norm_b;

    // attention
    struct ggml_tensor * query_key_value;
    struct ggml_tensor * wo;

    // normalization
    struct ggml_tensor * ffn_norm;
    struct ggml_tensor * ffn_norm_b;

    // ff
    struct ggml_tensor * w1;
    struct ggml_tensor * w2;
};

struct bloom_model {
    bloom_hparams hparams;

    struct ggml_tensor * tok_embeddings;
    // struct ggml_tensor * norm;
    // struct ggml_tensor * norm_b;

    struct ggml_tensor * output_norm;
    struct ggml_tensor * output_norm_b;
    struct ggml_tensor * output;
    

    std::vector<bloom_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 bloom_model_load(const std::string & fname, bloom_model & model, gpt_vocab & vocab, int n_ctx) {
    printf("%s: loading model from '%s' - please wait ...\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;
        }
    }

    int n_ff = 0;
    int n_parts = 0;

    // 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_mult,  sizeof(hparams.n_mult));
        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));

        hparams.n_ctx = n_ctx;

        n_ff = ((4*hparams.n_embd + hparams.n_mult - 1)/hparams.n_mult)*hparams.n_mult;
        // n_parts = BLOOM_N_PARTS.at(hparams.n_embd);
        n_parts = 1;

        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_mult  = %d\n", __func__, hparams.n_mult);
        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);
        printf("%s: n_ff    = %d\n", __func__, n_ff);
        printf("%s: n_parts = %d\n", __func__, n_parts);
    }

    // load vocab
    {
        const int32_t n_vocab = model.hparams.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;

            //if (i < 30000) {
            //    printf("%s: vocab[%d] = '%s'\n", __func__, i, word.c_str());
            //}
        }
    }

    // for the big tensors, we have the option to store the data in 16-bit floats or quantized
    // in order to save memory and also to speed up the computation
    ggml_type wtype = GGML_TYPE_COUNT;
    switch (model.hparams.f16) {
        case 0: wtype = GGML_TYPE_F32;  break;
        case 1: wtype = GGML_TYPE_F16;  break;
        case 2: wtype = GGML_TYPE_Q4_0; break;
        case 3: wtype = GGML_TYPE_Q4_1; break;
        default:
                {
                    fprintf(stderr, "%s: invalid model file '%s' (bad f16 value %d)\n",
                            __func__, fname.c_str(), model.hparams.f16);
                    return false;
                }
    }

    const ggml_type wtype2 = 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_head  = hparams.n_head;
        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*n_vocab*ggml_type_sizef(wtype); // tok_embeddings

        // ctx_size += n_embd*ggml_type_sizef(GGML_TYPE_F32); // norm
        // ctx_size += n_embd*ggml_type_sizef(GGML_TYPE_F32); // norm_b

        ctx_size += n_embd*ggml_type_sizef(GGML_TYPE_F32); // output_norm
        ctx_size += n_embd*ggml_type_sizef(GGML_TYPE_F32); // output_norm_b

        ctx_size += n_embd*n_vocab*ggml_type_sizef(wtype); // output

        ctx_size += n_layer*(n_embd*ggml_type_sizef(GGML_TYPE_F32)); // attention_norm
        ctx_size += n_layer*(n_embd*ggml_type_sizef(GGML_TYPE_F32)); // attention_norm_b

        ctx_size += n_layer*(n_embd*(n_embd+2*(n_embd/n_head))*ggml_type_sizef(wtype)); // query_key_value
        ctx_size += n_layer*(n_embd*n_embd*ggml_type_sizef(wtype)); // wo

        ctx_size += n_layer*(n_embd*ggml_type_sizef(GGML_TYPE_F32)); // ffn_norm
        ctx_size += n_layer*(n_embd*ggml_type_sizef(GGML_TYPE_F32)); // ffn_norm_b

        ctx_size += n_layer*(n_ff*n_embd*ggml_type_sizef(wtype)); // w1
        ctx_size += n_layer*(n_ff*n_embd*ggml_type_sizef(wtype)); // w2

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

        ctx_size += (5 + 10*n_layer)*256; // object overhead TODO:

        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_head  = hparams.n_head;
        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.tok_embeddings = ggml_new_tensor_2d(ctx, wtype, n_embd, n_vocab);
        // model.norm   = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embd);
        // model.norm_b = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embd);

        model.output_norm = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embd);
        model.output_norm_b = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embd);
        model.output = ggml_new_tensor_2d(ctx, wtype,         n_embd, n_vocab);

        // map by name
        model.tensors["tok_embeddings.weight"] = model.tok_embeddings;
        // model.tensors["norm.weight"]   = model.norm;
        // model.tensors["norm.bias"]   = model.norm_b;
        
        model.tensors["output_norm.weight"] = model.output_norm;
        model.tensors["output_norm.bias"] = model.output_norm_b;
        model.tensors["output.weight"] = model.output;

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

            layer.attention_norm = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embd);
            layer.attention_norm_b = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embd);

            // query_key_value shape for config.multi_query == True:
            layer.query_key_value = ggml_new_tensor_2d(ctx, wtype, n_embd, n_embd + 2 * (n_embd / n_head)); 
            layer.wo = ggml_new_tensor_2d(ctx, wtype, n_embd, n_embd);

            layer.ffn_norm = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embd);
            layer.ffn_norm_b = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embd);

            layer.w1 = ggml_new_tensor_2d(ctx, wtype, n_embd,   n_ff);
            layer.w2 = ggml_new_tensor_2d(ctx, wtype,   n_ff, n_embd);

            // map by name
            model.tensors["layers." + std::to_string(i) + ".attention_norm.weight"] = layer.attention_norm;
            model.tensors["layers." + std::to_string(i) + ".attention_norm.bias"] = layer.attention_norm_b;

            model.tensors["layers." + std::to_string(i) + ".attention.query_key_value.weight"] = layer.query_key_value;
            model.tensors["layers." + std::to_string(i) + ".attention.wo.weight"] = layer.wo;

            model.tensors["layers." + std::to_string(i) + ".ffn_norm.weight"] = layer.ffn_norm;
            model.tensors["layers." + std::to_string(i) + ".ffn_norm.bias"] = layer.ffn_norm_b;

            model.tensors["layers." + std::to_string(i) + ".feed_forward.w1.weight"] = layer.w1;
            model.tensors["layers." + std::to_string(i) + ".feed_forward.w2.weight"] = layer.w2;
        }
    }

    // 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);
    }

    const size_t file_offset = fin.tellg();

    fin.close();

    std::vector<uint8_t> tmp;

    for (int i = 0; i < n_parts; ++i) {
        const int part_id = i;
        //const int part_id = n_parts - i - 1;

        std::string fname_part = fname;
        if (i > 0) {
            fname_part += "." + std::to_string(i);
        }

        printf("%s: loading model part %d/%d from '%s'\n", __func__, i+1, n_parts, fname_part.c_str());

        fin = std::ifstream(fname_part, std::ios::binary);
        fin.seekg(file_offset);

        // load weights
        {
            int n_tensors = 0;
            size_t total_size = 0;

            printf("%s: ", __func__);

            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;
                }

                // split_type = 0: split by columns
                // split_type = 1: split by rows
                int split_type = 0;

                // split_type = 0:
                // regex:
                //   - tok_embeddings.*
                //   - layers.*.attention.wo.weight
                //   - layers.*.feed_forward.w2.weight

                // split_type = 1:
                // regex:
                //   - output.*
                //   - layers.*.attention.wq.weight
                //   - layers.*.attention.wk.weight
                //   - layers.*.attention.wv.weight
                //   - layers.*.feed_forward.w1.weight
                //   - layers.*.feed_forward.w3.weight
                if (name.find("tok_embeddings") != std::string::npos) {
                    split_type = 0;
                } else if (name.find("layers") != std::string::npos) {
                    if (name.find("attention.wo.weight") != std::string::npos) {
                        split_type = 0;
                    } else if (name.find("feed_forward.w2.weight") != std::string::npos) {
                        split_type = 0;
                    } else {
                        split_type = 1;
                    }
                } else if (name.find("output") != std::string::npos) {
                    split_type = 1;
                }

                auto tensor = model.tensors[name.data()];

                if (n_dims == 1) {
                    printf("%s[%d] = %d\n", name.data(), i, ne[0]);
                    if (ggml_nelements(tensor) != nelements) {
                        fprintf(stderr, "%s: tensor '%s' has wrong size in model file\n", __func__, name.data());
                        return false;
                    }
                } else {
                    printf("%s[%d] = %d x %d\n", name.data(), i, ne[0], ne[1]);
                    if (ggml_nelements(tensor)/n_parts != nelements) {
                        fprintf(stderr, "%s: tensor '%s' has wrong size in model file\n", __func__, name.data());
                        return false;
                    }
                }

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

                if (0) {
                    static const char * ftype_str[] = { "f32", "f16", "q4_0", "q4_1", };
                    printf("%24s - [%5d, %5d], type = %6s, split = %d\n", name.data(), ne[0], ne[1], ftype_str[ftype], split_type);
                }

                size_t bpe = 0;

                switch (ftype) {
                    case 0: bpe = ggml_type_size(GGML_TYPE_F32);  break;
                    case 1: bpe = ggml_type_size(GGML_TYPE_F16);  break;
                    case 2: bpe = ggml_type_size(GGML_TYPE_Q4_0); assert(ne[0] % 64 == 0); break;
                    case 3: bpe = ggml_type_size(GGML_TYPE_Q4_1); assert(ne[0] % 64 == 0); break;
                    default:
                            {
                                fprintf(stderr, "%s: unknown ftype %d in model file\n", __func__, ftype);
                                return false;
                            }
                };

                if (n_dims == 1 || n_parts == 1) {
                    if ((nelements*bpe)/ggml_blck_size(tensor->type) != 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;
                    }

                    if (part_id == 0) {
                        fin.read(reinterpret_cast<char *>(tensor->data), ggml_nbytes(tensor));
                    } else {
                        fin.seekg(ggml_nbytes(tensor), std::ios::cur);
                    }

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

                    if (split_type == 0) {
                        const int np0 = ne[0];

                        const size_t row_size = (tensor->ne[0]/ggml_blck_size(tensor->type))*ggml_type_size(tensor->type);
                        assert(row_size == tensor->nb[1]);

                        for (int i1 = 0; i1 < ne[1]; ++i1) {
                            const size_t offset_row = i1*row_size;
                            const size_t offset = offset_row + ((part_id*np0)/ggml_blck_size(tensor->type))*ggml_type_size(tensor->type);
                            fin.read(reinterpret_cast<char *>(tensor->data) + offset, row_size/n_parts);
                        }
                    } else {
                        const int np1 = ne[1];

                        const size_t row_size = (tensor->ne[0]/ggml_blck_size(tensor->type))*ggml_type_size(tensor->type);

                        for (int i1 = 0; i1 < ne[1]; ++i1) {
                            const size_t offset_row = (i1 + part_id*np1)*row_size;
                            fin.read(reinterpret_cast<char *>(tensor->data) + offset_row, row_size);
                        }
                    }

                    total_size += ggml_nbytes(tensor)/n_parts;
                }

                //printf("%42s - [%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);
                if (++n_tensors % 8 == 0) {
                    printf(".");
                    fflush(stdout);
                }
            }

            printf(" done\n");

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

        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 logits for the next token
//
// The GPT-J model requires about 16MB of memory per input token.
//
bool bloom_eval(
        const bloom_model & model,
        const int n_threads,
        const int n_past,
        const std::vector<gpt_vocab::id> & embd_inp,
              std::vector<float>         & embd_w) {

    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;

    const int d_key = n_embd/n_head;

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

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

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

//    printf("input tokens: %d\n", N);
//    for (int i = 0; i < N; ++i) {
//        printf("%d ", embd_inp[i]);
//    }

    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 * inpL = ggml_get_rows(ctx0, model.tok_embeddings, embd);

    // word embeddings norm
    // {
    //     inpL = ggml_norm(ctx0, inpL);
    //     inpL = ggml_mul(ctx0, ggml_repeat(ctx0, model.norm, inpL), inpL);
    //     inpL = ggml_add(ctx0, ggml_repeat(ctx0, model.norm_b, inpL), inpL);
    // }

    for (int il = 0; il < 1 /*TODO: replace 1 with n_layer after porting complete! */; ++il) {
        struct ggml_tensor * inpSA = inpL; //TODO: copy?

        struct ggml_tensor * cur;

        // layernorm_output = self.input_layernorm(hidden_states)
        {
            cur = ggml_norm(ctx0, inpL);

            // cur = attention_norm*cur
            cur = ggml_mul(ctx0,
                        ggml_repeat(ctx0, model.layers[il].attention_norm, cur),
                        cur);
            cur = ggml_add(ctx0, ggml_repeat(ctx0, model.layers[il].attention_norm_b, cur), cur);
        }

        // fused_qkv = self.query_key_value(hidden_states)
        {
            cur = ggml_mul_mat(ctx0,model.layers[il].query_key_value, cur);

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

        // cur = ggml_debug(ctx0, cur);

        // self-attention
        {
            // fused_qkv = fused_qkv.view(batch_size, seq_length, self.num_heads + 2, self.head_dim)

            //struct ggml_tensor * fused_qkv_view = ggml_view_3d(ctx0, cur,
            //    n_embd/n_head, n_head+2, N,
            //    n_embd/n_head * sizeof(float), n_embd + 2 * (n_embd / n_head) * sizeof(float), 0);

            size_t head_dim = n_embd/n_head;
            size_t fused_qkv_row_nb = (n_embd + 2 * (n_embd / n_head)) * sizeof(float);


            // (query_layer, key_layer, value_layer) = self._split_heads(fused_qkv)

            struct ggml_tensor * Qcur = ggml_view_3d(ctx0, cur,
                head_dim, n_head, N,
                head_dim * sizeof(float), fused_qkv_row_nb, 0);

            struct ggml_tensor * Kcur = ggml_view_3d(ctx0, cur,
                head_dim, 1, N,
                head_dim * sizeof(float), fused_qkv_row_nb,
                n_embd * sizeof(float));

            struct ggml_tensor * Vcur = ggml_view_3d(ctx0, cur,
                head_dim, 1, N,
                head_dim * sizeof(float), fused_qkv_row_nb,
                (n_embd + head_dim) * sizeof(float));

            // Example of how we can dump a tensor (Vcur) to stdout:
            //    ggml_build_forward_expand(&gf, Vcur);
            //    ggml_graph_compute(ctx0, &gf);
            //    ggml_print_tensor_f32(Vcur);
                
            // store key and value to memory
            if (N >= 1) {
                // TODO: something strange here, in originalv version Kcur, Vcur and model.memory_(k|v)
                // has values for all heads, but in falcon/modelling_RW.py key_layer, value_layer only
                // has space for one head?... probably related to the "multi_query" algorithm

                struct ggml_tensor * k = ggml_view_1d(ctx0, model.memory_k, N*head_dim, (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*head_dim, (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));
            }

            // query_layer = query_layer.transpose(1, 2).reshape(batch_size * self.num_heads, q_length, self.head_dim)

            struct ggml_tensor * Q =
                ggml_permute(ctx0,
                            ggml_cpy(ctx0, Qcur,
                                ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, head_dim, n_head, N)),
                        0, 2, 1, 3);

            // Note: for Falcon-7B num_kv == 1
            // key_layer = key_layer.transpose(1, 2).reshape(batch_size * self.num_kv, q_length, self.head_dim)

            struct ggml_tensor * K =
                ggml_permute(ctx0,
                            ggml_cpy(ctx0, Kcur,
                                ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, head_dim, 1, N)),
                        0, 2, 1, 3);

            // value_layer = value_layer.transpose(1, 2).reshape(batch_size * self.num_kv, q_length, self.head_dim)

            struct ggml_tensor * V =
                ggml_permute(ctx0,
                            ggml_cpy(ctx0, Vcur,
                                ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, head_dim, 1, N)),
                        0, 2, 1, 3);

            // TODO: verified to match layer 0 data from falcon/modelling_RW.py up to this point
            // That is, we stand here just before self.maybe_rotary from falcon/modelling_RW.py:

            //         query_layer, key_layer = self.maybe_rotary(query_layer, key_layer)

            // TODO: UNVERIFIED from here on:
            // using mode = 2 for GPT-NeoX mode
            // TODO: the result of ggml_rope_inplace here DOES NOT match the one from modelling_RW.py!
            // Possibly buggy, there is a remark to that effect in ggml.c
            Q = ggml_rope_inplace(ctx0, Q, n_past, head_dim, 2);
            K = ggml_rope_inplace(ctx0, K, n_past, head_dim, 2);

            // attn_output = F.scaled_dot_product_attention(
            //    query_layer_, key_layer_, value_layer_, None, 0.0, is_causal=True
            // )
            // which according to torch docs is equivalent to the following with
            //  L = number of query vectors, S = number of key/value vectors:
            // attn_mask = torch.ones(L, S, dtype=torch.bool).tril(diagonal=0)
            // attn_mask = attn_mask.masked_fill(~attn_mask, -float('inf'))
            // attn_weight = torch.softmax((Q @ K.transpose(-2, -1) / math.sqrt(Q.size(-1))) + attn_mask, dim=-1)
            // return attn_weight @ V
            
            // TODO: F.scaled_dot_product_attention can somehow deal with the KQV shape as produced by the above
            // 0,2,1,3 transposition, but the K*Q matrix multiplication below can't.. maybe the transpose should not
            // be done in GGML version after all... the dimensions match without it (e.g. 0,1,2,3), but not sure
            // if the results are as intended (did not check)

            // K * Q
            struct ggml_tensor * KQ = ggml_mul_mat(ctx0, K, Q);

            // KQ_scaled = KQ / sqrt(n_embd/n_head)
            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)
            struct ggml_tensor * KQ_masked = ggml_diag_mask_inf(ctx0, KQ_scaled, n_past);

            // KQ = soft_max(KQ_masked)
            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()
            struct ggml_tensor *V_trans =
                    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, model.memory_v->type, n_past + N, n_embd / n_head, n_head));

            // KQV = transpose(V) * KQ_soft_max
            struct ggml_tensor * KQV = ggml_mul_mat(ctx0, V_trans, KQ_soft_max);

            // KQV_merged = KQV.permute(0, 2, 1, 3)
            struct ggml_tensor * KQV_merged = ggml_permute(ctx0, KQV, 0, 2, 1, 3);

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

            // projection
            cur = ggml_mul_mat(ctx0,
                    model.layers[il].wo,
                    cur);
            // cur = ggml_add(ctx0, ggml_repeat(ctx0, model.layers[il].wo_b, cur), cur);
        }

        // struct ggml_tensor * inpFF = ggml_add(ctx0, cur, inpSA);
        struct ggml_tensor * inpFF = inpSA;
        struct ggml_tensor * attn_out = ggml_cpy(ctx0, cur, ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd, N));

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

                // cur = ffn_norm*cur + ffn_norm_b
                cur = ggml_mul(ctx0,
                        ggml_repeat(ctx0, model.layers[il].ffn_norm, cur),
                        cur);
                cur = ggml_add(ctx0, ggml_repeat(ctx0, model.layers[il].ffn_norm_b, cur), cur);
            }

            cur = ggml_mul_mat(ctx0,
                    model.layers[il].w1,
                    cur);
            // cur = ggml_add(ctx0, ggml_repeat(ctx0, model.layers[il].w1_b, cur), cur);

            cur = ggml_gelu(ctx0, cur);

            cur = ggml_mul_mat(ctx0,
                    model.layers[il].w2,
                    cur);
            // cur = ggml_add(ctx0, ggml_repeat(ctx0, model.layers[il].w2_b, cur), cur);
        }

        // cur  = ggml_add(ctx0, cur, inpFF);
        cur = ggml_add(ctx0, cur, attn_out);
        cur = ggml_add(ctx0, cur, inpL);

        // input for next layer
        inpL = cur;
    }

    // norm
    {
        inpL = ggml_norm(ctx0, inpL);
        // inpL = norm*inpL
        inpL = ggml_mul(ctx0,
                    ggml_repeat(ctx0, model.output_norm, inpL),
                    inpL);
        
        inpL = ggml_add(ctx0, ggml_repeat(ctx0, model.output_norm_b, inpL), inpL);
    }

    // lm_head
    {
        inpL = ggml_mul_mat(ctx0, model.output, 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);
    //printf("used_mem = %zu\n", ggml_used_mem(ctx0));

    ggml_free(ctx0);

    return true;
}

int main(int argc, char ** argv) {
    ggml_time_init();
    const int64_t t_main_start_us = ggml_time_us();

    gpt_params params;
    params.model = "models/ggml-model-bloomz-7b1-f16-q4_0.bin";
    params.prompt = "Je vais";

    if (gpt_params_parse(argc, argv, params) == false) {
        return 1;
    }

    if (params.seed < 0) {
        params.seed = time(NULL);
    }

    printf("%s: seed = %d\n", __func__, params.seed);

    std::mt19937 rng(params.seed);
    if (params.prompt.empty()) {
        params.prompt = gpt_random_prompt(rng);
    }

//    params.prompt = R"(// this function checks if the number n is prime
//bool is_prime(int n) {)";

    int64_t t_load_us = 0;

    gpt_vocab vocab;
    bloom_model model;

    // load the model
    {
        const int64_t t_start_us = ggml_time_us();
        if (!bloom_model_load(params.model, model, vocab, params.n_ctx)) {
            fprintf(stderr, "%s: failed to load model from '%s'\n", __func__, params.model.c_str());
            return 1;
        }

        t_load_us = ggml_time_us() - t_start_us;
    }

    int n_past = 0;

    int64_t t_sample_us  = 0;
    int64_t t_predict_us = 0;

    std::vector<float> logits;

    // tokenize the prompt
    std::vector<gpt_vocab::id> embd_inp = ::bloom_tokenize(vocab, params.prompt, false); //TODO: set bos to true?

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

    printf("\n");
    printf("%s: prompt: '%s'\n", __func__, params.prompt.c_str());
    printf("%s: number of tokens in prompt = %zu\n", __func__, embd_inp.size());
    for (int i = 0; i < (int) embd_inp.size(); i++) {
        printf("%6d -> '%s'\n", embd_inp[i], vocab.id_to_token.at(embd_inp[i]).c_str());
    }
    printf("\n");
    printf("sampling parameters: temp = %f, top_k = %d, top_p = %f, repeat_last_n = %i, repeat_penalty = %f\n", params.temp, params.top_k, params.top_p, params.repeat_last_n, params.repeat_penalty);
    printf("\n\n");

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

    // determine the required inference memory per token:
    // size_t mem_per_token = 0;
    // bloom_eval(model, params.n_threads, 0, { 0, 1, 2, 3 }, logits, mem_per_token);

    int last_n_size = params.repeat_last_n;
    std::vector<gpt_vocab::id> last_n_tokens(last_n_size);
    std::fill(last_n_tokens.begin(), last_n_tokens.end(), 0);

    for (int i = embd.size(); i < embd_inp.size() + params.n_predict; i++) {
        // predict
        if (embd.size() > 0) {
            const int64_t t_start_us = ggml_time_us();

            if (!bloom_eval(model, params.n_threads, n_past, embd, logits)) { // update logits
                printf("Failed to predict\n");
                return 1;
            }

            t_predict_us += ggml_time_us() - t_start_us;
        }

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

        if (i >= embd_inp.size()) {
            // sample next token
            const float top_p = params.top_p;
            const float temp  = params.temp;
            const float repeat_penalty = params.repeat_penalty;

            const int n_vocab = model.hparams.n_vocab;

            gpt_vocab::id id = 0;

            {
                const int64_t t_start_sample_us = ggml_time_us();

                id = bloom_sample_top_p(vocab, logits.data() + (logits.size() - n_vocab), last_n_tokens, repeat_penalty, top_p, temp, rng);

                // // print
                // printf("\ngenerated token: '%s' (%d)\n", vocab.id_to_token[id].c_str(), id);

                last_n_tokens.erase(last_n_tokens.begin());
                last_n_tokens.push_back(id);

                t_sample_us += ggml_time_us() - t_start_sample_us;
            }

            // add it to the context
            embd.push_back(id);
        } else {
            // if here, it means we are still processing the input prompt
            for (int k = i; k < embd_inp.size(); k++) {
                embd.push_back(embd_inp[k]);
                last_n_tokens.erase(last_n_tokens.begin());
                last_n_tokens.push_back(embd_inp[k]);
                if (embd.size() > params.n_batch) {
                    break;
                }
            }
            i += embd.size() - 1;
        }

        // display text
        for (auto id : embd) {
            printf("%s", vocab.id_to_token[id].c_str());
        }
        fflush(stdout);

        // end of text token
        if (embd.back() == 2) {
            printf(" [end of text]\n");
            break;
        }
    }

    // report timing
    {
        const int64_t t_main_end_us = ggml_time_us();

        printf("\n\n");
        printf("%s:     load time = %8.2f ms\n", __func__, t_load_us/1000.0f);
        printf("%s:   sample time = %8.2f ms\n", __func__, t_sample_us/1000.0f);
        printf("%s:  predict time = %8.2f ms / %.2f ms per token\n", __func__, t_predict_us/1000.0f, t_predict_us/1000.0f/n_past);
        printf("%s:    total time = %8.2f ms\n", __func__, (t_main_end_us - t_main_start_us)/1000.0f);
    }

    ggml_free(model.ctx);

    return 0;
}