quantize_gptneox.cpp

#include "ggml.h"

#include "utils.h"

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

// TODO: move somewhere else
#define QK 32

// default hparams (GPT-NeoX 6B)
struct gptneox_hparams {
    int32_t n_vocab = 50400;
    int32_t n_ctx   = 2048;
    int32_t n_embd  = 4096;
    int32_t n_head  = 16;
    int32_t n_layer = 28;
    int32_t n_rot   = 32;
    int32_t use_parallel_residual = 1;
    int32_t f16     = 1;
};

// quantize a model
bool gptneox_model_quantize(const std::string & fname_inp, const std::string & fname_out, int itype) {
    ggml_type type = GGML_TYPE_Q4_1;

    switch (itype) {
        case 2: type = GGML_TYPE_Q4_0; break;
        case 3: type = GGML_TYPE_Q4_1; break;
        default: fprintf(stderr, "%s: invalid quantization type %d\n", __func__, itype); return 1;
    };

    if (type != GGML_TYPE_Q4_0 && type != GGML_TYPE_Q4_1) {
        fprintf(stderr, "%s: invalid quantization type %d\n", __func__, type);
        return false;
    }

    gpt_vocab vocab;

    printf("%s: loading model from '%s'\n", __func__, fname_inp.c_str());

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

    auto fout = std::ofstream(fname_out, std::ios::binary);
    if (!fout) {
        fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname_out.c_str());
        return false;
    }

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

        fout.write((char *) &magic, sizeof(magic));
    }

    gptneox_hparams hparams;

    // load hparams
    {
        finp.read((char *) &hparams.n_vocab, sizeof(hparams.n_vocab));
        //finp.read((char *) &hparams.n_ctx,   sizeof(hparams.n_ctx));
        finp.read((char *) &hparams.n_embd,  sizeof(hparams.n_embd));
        finp.read((char *) &hparams.n_head,  sizeof(hparams.n_head));
        finp.read((char *) &hparams.n_layer, sizeof(hparams.n_layer));
        finp.read((char *) &hparams.n_rot,  sizeof(hparams.n_rot));
        finp.read((char *) &hparams.use_parallel_residual, sizeof(hparams.use_parallel_residual));
        finp.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: n_rot   = %d\n", __func__, hparams.n_rot);
        printf("%s: use_parallel_residual = %d\n", __func__, hparams.use_parallel_residual);
        printf("%s: f16     = %d\n", __func__, hparams.f16);

        fout.write((char *) &hparams.n_vocab, sizeof(hparams.n_vocab));
        //fout.write((char *) &hparams.n_ctx,   sizeof(hparams.n_ctx));
        fout.write((char *) &hparams.n_embd,  sizeof(hparams.n_embd));
        fout.write((char *) &hparams.n_head,  sizeof(hparams.n_head));
        fout.write((char *) &hparams.n_layer, sizeof(hparams.n_layer));
        fout.write((char *) &hparams.n_rot,  sizeof(hparams.n_rot));
        fout.write((char *) &hparams.use_parallel_residual, sizeof(hparams.use_parallel_residual));
        fout.write((char *) &itype,           sizeof(hparams.f16));
    }

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

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

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

            word.resize(len);
            finp.read ((char *) word.data(), len);
            fout.write((char *) word.data(), len);

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

    // load weights
    {
        size_t total_size_org = 0;
        size_t total_size_new = 0;

        std::vector<float> work;

        std::vector<uint8_t>     data_u8;
        std::vector<ggml_fp16_t> data_f16;
        std::vector<float>       data_f32;

        std::vector<int64_t> hist_all(1 << 4, 0);

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

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

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

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

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

            {
                static const char * ftype_str[] = { "f32", "f16", "q4_0", "q4_1", };
                printf("%48s - [%5d, %5d], type = %6s ", name.data(), ne[0], ne[1], ftype_str[ftype]);
            }

            // regexes of tensor names to be quantized
            const std::vector<std::string> k_names = {
                ".*weight",
            };

            bool quantize = false;
            for (const auto & s : k_names) {
                if (std::regex_match(name, std::regex(s))) {
                    quantize = true;
                    break;
                }
            }

            // quantize only 2D tensors
            quantize &= (n_dims == 2);

            if (quantize) {
                if (ftype != 0 && ftype != 1) {
                    fprintf(stderr, "%s: unsupported ftype %d for integer quantization\n", __func__, ftype);
                    return false;
                }

                if (ftype == 1) {
                    data_f16.resize(nelements);
                    finp.read(reinterpret_cast<char *>(data_f16.data()), nelements * sizeof(ggml_fp16_t));
                    data_f32.resize(nelements);
                    for (int i = 0; i < nelements; ++i) {
                        data_f32[i] = ggml_fp16_to_fp32(data_f16[i]);
                    }
                    // if name is "gpt_neox.layers.0.attention.query.weight", then print first 10 float values.
                    if (name.find("gpt_neox.layers.0.attention") != std::string::npos) {
                        // if query.weight or key.weight or value.weight, print first 10 float values.
                        if (name.find("query.") != std::string::npos ||
                            name.find("key.") != std::string::npos ||
                            name.find("value.") != std::string::npos) {
                            printf("\n\nfirst 10 values: ");
                            for (int i = 0; i < 10; ++i) {
                                printf("%f ", data_f32[i]);
                            }
                            printf("\n");
                        }
                    }
                } else {
                    data_f32.resize(nelements);
                    finp.read(reinterpret_cast<char *>(data_f32.data()), nelements * sizeof(float));
                }

                ftype = itype;
            } else {
                const int bpe = (ftype == 0) ? sizeof(float) : sizeof(uint16_t);

                data_u8.resize(nelements*bpe);
                finp.read(reinterpret_cast<char *>(data_u8.data()), nelements * bpe);
            }

            {
                // if name is "gpt_neox.layers.0.attention.query.weight", then print first 10 float values.
                if (name.find("gpt_neox.layers.0.attention") != std::string::npos) {
                    // if query.weight or key.weight or value.weight, print first 10 float values.
                    if (name.find("query.bias") != std::string::npos ||
                        name.find("key.bias") != std::string::npos ||
                        name.find("value.bias") != std::string::npos) {
                        printf("\n\nfirst 10 values %s: ", name.data());
                        // combine four consecutive uint8_t to one float
                        for (int i = 0; i < 10; ++i) {
                            uint8_t *p = &data_u8[i*4];
                            float f = *(float *)p;
                            printf("%f ", f);
                        }
                        printf("\n");
                    }
                }
            }

            fout.write(reinterpret_cast<char *>(&n_dims), sizeof(n_dims));
            fout.write(reinterpret_cast<char *>(&length), sizeof(length));
            fout.write(reinterpret_cast<char *>(&ftype),  sizeof(ftype));
            for (int i = 0; i < n_dims; ++i) {
                fout.write(reinterpret_cast<char *>(&ne[i]), sizeof(ne[i]));
            }
            fout.write(&name[0], length);

            if (quantize) {
                printf("quantizing .. ");
                work.resize(nelements); // for quantization

                size_t cur_size = 0;
                std::vector<int64_t> hist_cur(1 << 4, 0);

                switch (type) {
                    case GGML_TYPE_Q4_0:
                        {
                            cur_size = ggml_quantize_q4_0(data_f32.data(), work.data(), nelements, ne[0], QK, hist_cur.data());
                        } break;
                    case GGML_TYPE_Q4_1:
                        {
                            cur_size = ggml_quantize_q4_1(data_f32.data(), work.data(), nelements, ne[0], QK, hist_cur.data());
                        } break;
                    default:
                        {
                            fprintf(stderr, "%s: unsupported quantization type %d\n", __func__, type);
                            return false;
                        }
                }
                // // if name is "gpt_neox.layers.0.attention.query.weight", then print first 32 quantized values from work.data()
                // if (name.find("layers.0.attention.query.weight") != std::string::npos) {
                //     printf("\n\n\n");
                //     // first value is a fp32 scale followed by 32 "int4" values.
                //     printf("scale = %f\n", work[0]);
                //     void *p = &work[1];
                //     // Since int4_t is not defined, we use int8_t to print the values two at a time, offset by 4 bits.
                //     int8_t *p8 = (int8_t *)p;
                //     for (int i = 0; i < 16; i++) {
                //         // print first 4 bits
                //         printf("%d ", (p8[i] >> 4) & 0xf);
                //         // print last 4 bits
                //         printf("%d ", p8[i] & 0xf);
                //     }
                //     printf("\n\n\n");
                // }

                fout.write(reinterpret_cast<char *>(work.data()), cur_size);
                total_size_new += cur_size;

                printf("size = %8.2f MB -> %8.2f MB | hist: ", nelements * sizeof(float)/1024.0/1024.0, cur_size/1024.0/1024.0);
                for (int i = 0; i < hist_cur.size(); ++i) {
                    hist_all[i] += hist_cur[i];
                }

                for (int i = 0; i < hist_cur.size(); ++i) {
                    printf("%5.3f ", hist_cur[i] / (float)nelements);
                }
                printf("\n");
            } else {
                printf("size = %8.3f MB\n", data_u8.size()/1024.0/1024.0);
                fout.write(reinterpret_cast<char *>(data_u8.data()), data_u8.size());
                total_size_new += data_u8.size();
            }

            total_size_org += nelements * sizeof(float);
        }

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

        {
            int64_t sum_all = 0;
            for (int i = 0; i < hist_all.size(); ++i) {
                sum_all += hist_all[i];
            }

            printf("%s: hist: ", __func__);
            for (int i = 0; i < hist_all.size(); ++i) {
                printf("%5.3f ", hist_all[i] / (float)sum_all);
            }
            printf("\n");
        }
    }

    finp.close();
    fout.close();

    return true;
}

// usage:
// ./quantize models/ggml-model-bloomz-7b1-f16.bin models/ggml-model-bloomz-7b1-f16-quant.bin 2
//
int main(int argc, char ** argv) {
    ggml_time_init();
    if (argc != 4) {
        fprintf(stderr, "usage: %s model-f32.bin model-quant.bin type\n", argv[0]);
        fprintf(stderr, "  type = 2 - q4_0\n");
        fprintf(stderr, "  type = 3 - q4_1\n");
        return 1;
    }

    // needed to initialize f16 tables
    {
        struct ggml_init_params params = { 0, NULL };
        struct ggml_context * ctx = ggml_init(params);
        ggml_free(ctx);
    }

    const std::string fname_inp = argv[1];
    const std::string fname_out = argv[2];

    const int itype = atoi(argv[3]);

    const int64_t t_main_start_us = ggml_time_us();

    int64_t t_quantize_us = 0;

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

        if (!gptneox_model_quantize(fname_inp, fname_out, itype)) {
            fprintf(stderr, "%s: failed to quantize model from '%s'\n", __func__, fname_inp.c_str());
            return 1;
        }

        t_quantize_us = ggml_time_us() - t_start_us;
    }

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

        printf("\n");
        printf("%s: quantize time = %8.2f ms\n", __func__, t_quantize_us/1000.0f);
        printf("%s:    total time = %8.2f ms\n", __func__, (t_main_end_us - t_main_start_us)/1000.0f);
    }

    return 0;
}