Llama.cpp运行流程

Llama.cpp运行流程

参考；在低配Windows上部署原版llama.cpp_git llama csdn-CSDN博客

需求：本机上运行大模型语言的算法等；

实现方式：Llama.cpp使用int4这种数值模式，显著降低内存需求；使用原始C++的项目来重写Llama代码，使得其可以在各种硬件上本地运行；

主要途径：使用优化和量化技术来量化权重的情况下，Llama.cpp使得大型语言模型可以在本地的多种硬件上运行，而无需GPU；内存带宽往往是推理的瓶颈，通过量化使用更少的精度可以减少存储模型所需的内存。

1.下载Llama.cpp

https://github.com/ggerganov/llama.cpp

2.准备编译工具

Llama.cpp是跨平台的，在Windows平台下，需要准备MinGW和Cmake

MinGW

Set-ExecutionPolicy RemoteSigned -Scope CurrentUser
iex "& {
   $(irm get.scoop.sh)} -RunAsAdmin"
scoop bucket add extras
scoop bucket add main
scoop install mingw

Cmake

https://cmake.org/download/

安装相应版本

3.编译LLaMa.cpp

进入下载好的文件夹

编译

cmake . -G "MinGW Makefiles"
cmake --build . --config Release

在bin/生成exe文件

现在llama更新很快，对于之前所提及的 quantize.exe 和 main.exe 都有相应的名称变化

main.exe : llama-cli.exe

quantize.exe : llama-quantize.exe

因为本地运行的gguf模型文件已经从hugging face下载好了，所以无需使用 llama-quantize.exe文件

部署上也参考了以下：

llama.cpp/examples/main at master · ggerganov/llama.cpp (github.com)

---

有了程序之后，我们需要下载模型

就是以上提及的GGUF文件

直接采用了

https://huggingface.co/ggml-org/gemma-1.1-7b-it-Q4_K_M-GGUF/resolve/main/gemma-1.1-7b-it.Q4_K_M.gguf?download=true

Gemma模型

下载好的gguf模型文件默认放到models/文件夹中

切换到bin目录下，运行如下代码：

llama-cli.exe -m ..\models\gemma-1.1-7b-it.Q4_K_M.gguf --prompt "Once upon a time"

上述 …\ 表示切换到上一层目录之后再寻找models文件

运行效果：

C:\……\llama.cpp\bin>llama-cli.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf --prompt "Once upon a time"
Log start
main: build = 3616 (11b84eb4)
main: built with  for x86_64-w64-mingw32
main: seed  = 1724916830
gguf_init_from_file: failed to open 'models\gemma-1.1-7b-it.Q4_K_M.gguf': 'No such file or directory'
llama_model_load: error loading model: llama_model_loader: failed to load model from models\gemma-1.1-7b-it.Q4_K_M.gguf

llama_load_model_from_file: failed to load model
llama_init_from_gpt_params: error: failed to load model 'models\gemma-1.1-7b-it.Q4_K_M.gguf'
main: error: unable to load model

C:\Users\v-yangxinyi\Llama.cpp\llama.cpp\bin>llama-cli.exe -m ..\models\gemma-1.1-7b-it.Q4_K_M.gguf --prompt "Once upon a time"
Log start
main: build = 3616 (11b84eb4)
main: built with  for x86_64-w64-mingw32
main: seed  = 1724916851
llama_model_loader: loaded meta data with 24 key-value pairs and 254 tensors from ..\models\gemma-1.1-7b-it.Q4_K_M.gguf (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = gemma
llama_model_loader: - kv   1:                               general.name str              = gemma-1.1-7b-it
llama_model_loader: - kv   2:                       gemma.context_length u32              = 8192
llama_model_loader: - kv   3:                     gemma.embedding_length u32              = 3072
llama_model_loader: - kv   4:                          gemma.block_count u32              = 28
llama_model_loader: - kv   5:                  gemma.feed_forward_length u32              = 24576
llama_model_loader: - kv   6:                 gemma.attention.head_count u32              = 16
llama_model_loader: - kv   7:              gemma.attention.head_count_kv u32              = 16
llama_model_loader: - kv   8:     gemma.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv   9:                 gemma.attention.key_length u32              = 256
llama_model_loader: - kv  10:               gemma.attention.value_length u32              = 256
llama_model_loader: - kv  11:                          general.file_type u32              = 15
llama_model_loader: - kv  12:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  13:                      tokenizer.ggml.tokens arr[str,256000]  = ["<pad>", "<eos>", "<bos>", "<unk>", ...
llama_model_loader: - kv  14:                      tokenizer.ggml.scores arr[f32,256000]  = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv  15:                  tokenizer.ggml.token_type arr[i32,256000]  = [3, 3, 3, 2, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  16:                tokenizer.ggml.bos_token_id u32              = 2
llama_model_loader: - kv  17:                tokenizer.ggml.eos_token_id u32              = 1
llama_model_loader: - kv  18:            tokenizer.ggml.unknown_token_id u32              = 3
llama_model_loader: - kv  19:            tokenizer.ggml.padding_token_id u32              = 0
llama_model_loader: - kv  20:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  21:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  22:                    tokenizer.chat_template str              = {
  { bos_token }}{% if messages[0]['rol...
llama_model_loader: - kv  23:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   57 tensors
llama_model_loader: - type q4_K:  168 tensors
llama_model_loader: - type q6_K:   29 tensors
llm_load_vocab: special tokens cache size = 4
llm_load_vocab: token to piece cache size = 1.6014 MB
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = gemma
llm_load_print_meta: vocab type       = SPM
llm_load_print_meta: n_vocab          = 256000
llm_load_print_meta: n_merges         = 0
llm_load_print_meta: vocab_only       = 0
llm_load_print_meta: n_ctx_train      = 8192
llm_load_print_meta: n_embd           = 3072
llm_load_print_meta: n_layer          = 28
llm_load_print_meta: n_head           = 16
llm_load_print_meta: n_head_kv        = 16
llm_load_print_meta: n_rot            = 256
llm_load_print_meta: n_swa            = 0
llm_load_print_meta: n_embd_head_k    = 256
llm_load_print_meta: n_embd_head_v    = 256
llm_load_print_meta: n_gqa            = 1
llm_load_print_meta: n_embd_k_gqa     = 4096
llm_load_print_meta: n_embd_v_gqa     = 4096
llm_load_print_meta: f_norm_eps       = 0.0e+00
llm_load_print_meta: f_norm_rms_eps   = 1.0e-06
llm_load_print_meta: f_clamp_kqv      = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: f_logit_scale    = 0.0e+00
llm_load_print_meta: n_ff             = 24576
llm_load_print_meta: n_expert         = 0
llm_load_print_meta: n_expert_used    = 0
llm_load_print_meta: causal attn      = 1
llm_load_print_meta: pooling type     = 0
llm_load_print_meta: rope type        = 2
llm_load_print_meta: rope scaling     = linear
llm_load_print_meta: freq_base_train  = 10000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_ctx_orig_yarn  = 8192
llm_load_print_meta: rope_finetuned   = unknown
llm_load_print_meta: ssm_d_conv       = 0
llm_load_print_meta: ssm_d_inner      = 0
llm_load_print_meta: ssm_d_state      = 0
llm_load_print_meta: ssm_dt_rank      = 0
llm_load_print_meta: ssm_dt_b_c_rms   = 0
llm_load_print_meta: model type       = 7B
llm_load_print_meta: model ftype      = Q4_K - Medium
llm_load_print_meta: model params     = 8.54 B
llm_load_print_meta: model size       = 4.96 GiB (4.99 BPW)
llm_load_print_meta: general.name     = gemma-1.1-7b-it
llm_load_print_meta: BOS token        = 2 '<bos>'
llm_load_print_meta: EOS token        = 1 '<eos>'
llm_load_print_meta: UNK token        = 3 '<unk>'
llm_load_print_meta: PAD token        &