diff --git a/Makefile b/Makefile index a44295312..5a2b3a52e 100644 --- a/Makefile +++ b/Makefile @@ -194,7 +194,7 @@ gpttype_adapter.o: $(CXX) $(CXXFLAGS) -c gpttype_adapter.cpp -o gpttype_adapter.o clean: - rm -vf *.o main quantize quantize-stats perplexity embedding main.exe quantize.exe koboldcpp.dll koboldcpp_openblas.dll koboldcpp_clblast.dll gptj.exe gpt2.exe + rm -vf *.o main quantize_llama quantize_gpt2 quantize_gptj quantize-stats perplexity embedding main.exe quantize_llama.exe quantize_gptj.exe quantize_gpt2.exe koboldcpp.dll koboldcpp_openblas.dll koboldcpp_openblas_noavx2.dll koboldcpp_clblast.dll gptj.exe gpt2.exe main: examples/main/main.cpp ggml.o llama.o common.o $(CXX) $(CXXFLAGS) examples/main/main.cpp ggml.o llama.o common.o -o main $(LDFLAGS) @@ -214,25 +214,24 @@ llamalib_openblas_noavx2: ggml_openblas_noavx2.o ggml_v1_noavx2.o expose.o commo llamalib_clblast: ggml_clblast.o ggml_v1.o expose.o common.o llama_adapter.o gpttype_adapter.o $(CLBLAST_BUILD) -quantize: examples/quantize/quantize.cpp ggml.o llama.o - $(CXX) $(CXXFLAGS) examples/quantize/quantize.cpp ggml.o llama.o -o quantize $(LDFLAGS) +quantize_llama: examples/quantize/quantize.cpp ggml.o llama.o + $(CXX) $(CXXFLAGS) examples/quantize/quantize.cpp ggml.o llama.o -o quantize_llama $(LDFLAGS) quantize-stats: examples/quantize-stats/quantize-stats.cpp ggml.o llama.o $(CXX) $(CXXFLAGS) examples/quantize-stats/quantize-stats.cpp ggml.o llama.o -o quantize-stats $(LDFLAGS) - quantize_gptj: ggml.o llama.o $(CXX) $(CXXFLAGS) otherarch/gptj_quantize.cpp ggml.o llama.o -o quantize_gptj $(LDFLAGS) +quantize_gpt2: ggml.o llama.o + $(CXX) $(CXXFLAGS) otherarch/gpt2_quantize.cpp ggml.o llama.o -o quantize_gpt2 $(LDFLAGS) + perplexity: examples/perplexity/perplexity.cpp ggml.o llama.o common.o $(CXX) $(CXXFLAGS) examples/perplexity/perplexity.cpp ggml.o llama.o common.o -o perplexity $(LDFLAGS) embedding: examples/embedding/embedding.cpp ggml.o llama.o common.o $(CXX) $(CXXFLAGS) examples/embedding/embedding.cpp ggml.o llama.o common.o -o embedding $(LDFLAGS) -gpt2: ggml.o - $(CXX) $(CXXFLAGS) otherarch/gpt2_v2.cpp otherarch/utils.cpp ggml.o -o gpt2 $(LDFLAGS) - libllama.so: llama.o ggml.o $(CXX) $(CXXFLAGS) -shared -fPIC -o libllama.so llama.o ggml.o $(LDFLAGS) diff --git a/otherarch/gpt2_quantize.cpp b/otherarch/gpt2_quantize.cpp new file mode 100644 index 000000000..9eb7b79b1 --- /dev/null +++ b/otherarch/gpt2_quantize.cpp @@ -0,0 +1,331 @@ +#include "ggml.h" + +#include "otherarch/utils.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +// 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; +}; + +// quantize a model +bool gpt2_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)); + } + + gpt2_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.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); + + 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 *) &itype, sizeof(hparams.f16)); + } + + // load vocab + { + int32_t n_vocab = 0; + finp.read ((char *) &n_vocab, sizeof(n_vocab)); + fout.write((char *) &n_vocab, sizeof(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 work; + + std::vector data_u8; + std::vector data_f16; + std::vector data_f32; + + std::vector hist_all(1 << 4, 0); + + while (true) { + int32_t n_dims; + int32_t length; + int32_t ftype; + + finp.read(reinterpret_cast(&n_dims), sizeof(n_dims)); + finp.read(reinterpret_cast(&length), sizeof(length)); + finp.read(reinterpret_cast(&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(&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("%24s - [%5d, %5d], type = %6s ", name.data(), ne[0], ne[1], ftype_str[ftype]); + } + + // regexes of tensor names to be quantized + const std::vector k_names = { + "model/wte", + "model/lm_head", + "model/h.*/attn/c_attn/w", + "model/h.*/attn/c_proj/w", + "model/h.*/mlp/c_fc/w", + "model/h.*/mlp/c_proj/w", + }; + + bool quantize = false; + for (const auto & s : k_names) { + if (std::regex_match(name, std::regex(s))) { + quantize = true; + break; + } + } + + 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(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]); + } + } else { + data_f32.resize(nelements); + finp.read(reinterpret_cast(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(data_u8.data()), nelements * bpe); + } + + fout.write(reinterpret_cast(&n_dims), sizeof(n_dims)); + fout.write(reinterpret_cast(&length), sizeof(length)); + fout.write(reinterpret_cast(&ftype), sizeof(ftype)); + for (int i = 0; i < n_dims; ++i) { + fout.write(reinterpret_cast(&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 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], hist_cur.data()); + } break; + case GGML_TYPE_Q4_1: + { + cur_size = ggml_quantize_q4_1(data_f32.data(), work.data(), nelements, ne[0], hist_cur.data()); + } break; + default: + { + fprintf(stderr, "%s: unsupported quantization type %d\n", __func__, type); + return false; + } + } + + fout.write(reinterpret_cast(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(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: +// ./gpt-2-quantize models/gpt-2-117M/ggml-model.bin models/gpt-2-117M/ggml-model-quant.bin type +// +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, false }; + 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 (!gpt2_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; +} \ No newline at end of file diff --git a/otherarch/gptj_quantize.cpp b/otherarch/gptj_quantize.cpp index 8d29f7748..46eb94792 100644 --- a/otherarch/gptj_quantize.cpp +++ b/otherarch/gptj_quantize.cpp @@ -293,7 +293,7 @@ int main(int argc, char ** argv) { // needed to initialize f16 tables { - struct ggml_init_params params = { 0, NULL }; + struct ggml_init_params params = { 0, NULL, false }; struct ggml_context * ctx = ggml_init(params); ggml_free(ctx); }