开源C++版AI画图大模型框架stable-diffusion.cpp开发使用初体验

本文主要是介绍开源C++版AI画图大模型框架stable-diffusion.cpp开发使用初体验,希望对大家解决编程问题提供一定的参考价值,需要的开发者们随着小编来一起学习吧!

stable-diffusion.cpp是一个C++编写的轻量级开源类AIGC大模型框架,可以支持在消费级普通设备上本地部署运行大模型进行AI画图,以及作为依赖库集成的到应用程序中提供类似于网页版stable-diffusion的功能。

以下基于stable-diffusion.cpp的源码利用C++ api来开发实例demo演示加载本地模型文件输入提示词生成画图,这里采用显卡CUDA加速计算,如果没有显卡也可以直接使用CPU。

项目结构

stable_diffusion_cpp_starter- stable-diffusion.cpp- src|- main.cpp- CMakeLists.txt

有两个前置操作:

  • 在系统安装好CUDA Toolkit
  • 将stable-diffusion.cpp源码根目录的CMakeLists.txt里面SD_CUBLAS选项打开设为ON

不过,如果没有支持CUDA的显卡,默认采用CPU计算,则可以忽略以上两项

CMakeLists.txt

cmake_minimum_required(VERSION 3.15)project(stable_diffusion_cpp_starter)set(CMAKE_CXX_STANDARD 14)
set(CMAKE_CXX_STANDARD_REQUIRED ON)add_subdirectory(stable-diffusion.cpp)include_directories(${CMAKE_CURRENT_SOURCE_DIR}/stable-diffusion.cpp${CMAKE_CURRENT_SOURCE_DIR}/stable-diffusion.cpp/thirdparty
)file(GLOB SRCsrc/*.hsrc/*.cpp
)add_executable(${PROJECT_NAME} ${SRC})target_link_libraries(${PROJECT_NAME} stable-diffusion ${CMAKE_THREAD_LIBS_INIT} # means pthread on unix
)

main.cpp

#include <stdio.h>
#include <string.h>
#include <time.h>
#include <iostream>
#include <random>
#include <string>
#include <vector>#include "stable-diffusion.h"#define STB_IMAGE_IMPLEMENTATION
#define STB_IMAGE_STATIC
#include "stb_image.h"#define STB_IMAGE_WRITE_IMPLEMENTATION
#define STB_IMAGE_WRITE_STATIC
#include "stb_image_write.h"#define STB_IMAGE_RESIZE_IMPLEMENTATION
#define STB_IMAGE_RESIZE_STATIC
#include "stb_image_resize.h"const char* rng_type_to_str[] = {"std_default","cuda",
};// Names of the sampler method, same order as enum sample_method in stable-diffusion.h
const char* sample_method_str[] = {"euler_a","euler","heun","dpm2","dpm++2s_a","dpm++2m","dpm++2mv2","lcm",
};// Names of the sigma schedule overrides, same order as sample_schedule in stable-diffusion.h
const char* schedule_str[] = {"default","discrete","karras","ays",
};const char* modes_str[] = {"txt2img","img2img","img2vid","convert",
};enum SDMode 
{TXT2IMG,IMG2IMG,IMG2VID,CONVERT,MODE_COUNT
};struct SDParams 
{int n_threads = -1;SDMode mode   = TXT2IMG;std::string model_path;std::string vae_path;std::string taesd_path;std::string esrgan_path;std::string controlnet_path;std::string embeddings_path;std::string stacked_id_embeddings_path;std::string input_id_images_path;sd_type_t wtype = SD_TYPE_COUNT;std::string lora_model_dir;std::string output_path = "output.png";std::string input_path;std::string control_image_path;std::string prompt;std::string negative_prompt;float min_cfg     = 1.0f;float cfg_scale   = 7.0f;float style_ratio = 20.f;int clip_skip     = -1;  // <= 0 represents unspecifiedint width         = 512;int height        = 512;int batch_count   = 1;int video_frames         = 6;int motion_bucket_id     = 127;int fps                  = 6;float augmentation_level = 0.f;sample_method_t sample_method = EULER_A;schedule_t schedule           = DEFAULT;int sample_steps              = 20;float strength                = 0.75f;float control_strength        = 0.9f;rng_type_t rng_type           = CUDA_RNG;int64_t seed                  = 42;bool verbose                  = false;bool vae_tiling               = false;bool control_net_cpu          = false;bool normalize_input          = false;bool clip_on_cpu              = false;bool vae_on_cpu               = false;bool canny_preprocess         = false;bool color                    = false;int upscale_repeats           = 1;
};static std::string sd_basename(const std::string& path) 
{size_t pos = path.find_last_of('/');if (pos != std::string::npos) {return path.substr(pos + 1);}pos = path.find_last_of('\\');if (pos != std::string::npos) {return path.substr(pos + 1);}return path;
}std::string get_image_params(SDParams params, int64_t seed) 
{std::string parameter_string = params.prompt + "\n";if (params.negative_prompt.size() != 0) {parameter_string += "Negative prompt: " + params.negative_prompt + "\n";}parameter_string += "Steps: " + std::to_string(params.sample_steps) + ", ";parameter_string += "CFG scale: " + std::to_string(params.cfg_scale) + ", ";parameter_string += "Seed: " + std::to_string(seed) + ", ";parameter_string += "Size: " + std::to_string(params.width) + "x" + std::to_string(params.height) + ", ";parameter_string += "Model: " + sd_basename(params.model_path) + ", ";parameter_string += "RNG: " + std::string(rng_type_to_str[params.rng_type]) + ", ";parameter_string += "Sampler: " + std::string(sample_method_str[params.sample_method]);if (params.schedule == KARRAS) {parameter_string += " karras";}parameter_string += ", ";parameter_string += "Version: stable-diffusion.cpp";return parameter_string;
}/* Enables Printing the log level tag in color using ANSI escape codes */
void sd_log_cb(enum sd_log_level_t level, const char* log, void* data) 
{SDParams* params = (SDParams*)data;int tag_color;const char* level_str;FILE* out_stream = (level == SD_LOG_ERROR) ? stderr : stdout;if (!log || (!params->verbose && level <= SD_LOG_DEBUG)) return;switch (level) {case SD_LOG_DEBUG:tag_color = 37;level_str = "DEBUG";break;case SD_LOG_INFO:tag_color = 34;level_str = "INFO";break;case SD_LOG_WARN:tag_color = 35;level_str = "WARN";break;case SD_LOG_ERROR:tag_color = 31;level_str = "ERROR";break;default: /* Potential future-proofing */tag_color = 33;level_str = "?????";break;}if (params->color == true) fprintf(out_stream, "\033[%d;1m[%-5s]\033[0m ", tag_color, level_str);else fprintf(out_stream, "[%-5s] ", level_str);fputs(log, out_stream);fflush(out_stream);
}int main(int argc, const char* argv[]) 
{// set sd paramsconst std::string model_path = "./v1-5-pruned-emaonly.ckpt";const std::string img_output_path = "./gen_img.png";const std::string prompt = "a cute little dog with flowers";SDParams params;params.model_path = model_path;params.output_path = img_output_path;params.prompt = prompt;sd_set_log_callback(sd_log_cb, (void*)&params);if (params.mode == CONVERT) {bool success = convert(params.model_path.c_str(), params.vae_path.c_str(), params.output_path.c_str(), params.wtype);if (!success) {fprintf(stderr,"convert '%s'/'%s' to '%s' failed\n",params.model_path.c_str(),params.vae_path.c_str(),params.output_path.c_str());return 1;} else {printf("convert '%s'/'%s' to '%s' success\n",params.model_path.c_str(),params.vae_path.c_str(),params.output_path.c_str());return 0;}}if (params.mode == IMG2VID) {fprintf(stderr, "SVD support is broken, do not use it!!!\n");return 1;}// prepare image bufferbool vae_decode_only          = true;uint8_t* input_image_buffer   = NULL;uint8_t* control_image_buffer = NULL;if (params.mode == IMG2IMG || params.mode == IMG2VID) {vae_decode_only = false;int c              = 0;int width          = 0;int height         = 0;input_image_buffer = stbi_load(params.input_path.c_str(), &width, &height, &c, 3);if (input_image_buffer == NULL) {fprintf(stderr, "load image from '%s' failed\n", params.input_path.c_str());return 1;}if (c < 3) {fprintf(stderr, "the number of channels for the input image must be >= 3, but got %d channels\n", c);free(input_image_buffer);return 1;}if (width <= 0) {fprintf(stderr, "error: the width of image must be greater than 0\n");free(input_image_buffer);return 1;}if (height <= 0) {fprintf(stderr, "error: the height of image must be greater than 0\n");free(input_image_buffer);return 1;}// Resize input image ...if (params.height != height || params.width != width) {printf("resize input image from %dx%d to %dx%d\n", width, height, params.width, params.height);int resized_height = params.height;int resized_width  = params.width;uint8_t* resized_image_buffer = (uint8_t*)malloc(resized_height * resized_width * 3);if (resized_image_buffer == NULL) {fprintf(stderr, "error: allocate memory for resize input image\n");free(input_image_buffer);return 1;}stbir_resize(input_image_buffer, width, height, 0,resized_image_buffer, resized_width, resized_height, 0, STBIR_TYPE_UINT8,3 /*RGB channel*/, STBIR_ALPHA_CHANNEL_NONE, 0,STBIR_EDGE_CLAMP, STBIR_EDGE_CLAMP,STBIR_FILTER_BOX, STBIR_FILTER_BOX,STBIR_COLORSPACE_SRGB, nullptr);// Save resized resultfree(input_image_buffer);input_image_buffer = resized_image_buffer;}}// init sd contextsd_ctx_t* sd_ctx = new_sd_ctx(params.model_path.c_str(),params.vae_path.c_str(),params.taesd_path.c_str(),params.controlnet_path.c_str(),params.lora_model_dir.c_str(),params.embeddings_path.c_str(),params.stacked_id_embeddings_path.c_str(),vae_decode_only,params.vae_tiling,true,params.n_threads,params.wtype,params.rng_type,params.schedule,params.clip_on_cpu,params.control_net_cpu,params.vae_on_cpu);if (sd_ctx == NULL) {printf("new_sd_ctx_t failed\n");return 1;}sd_image_t* control_image = NULL;if (params.controlnet_path.size() > 0 && params.control_image_path.size() > 0) {int c                = 0;control_image_buffer = stbi_load(params.control_image_path.c_str(), &params.width, &params.height, &c, 3);if (control_image_buffer == NULL) {fprintf(stderr, "load image from '%s' failed\n", params.control_image_path.c_str());return 1;}control_image = new sd_image_t{(uint32_t)params.width,(uint32_t)params.height,3,control_image_buffer};if (params.canny_preprocess) {  // apply preprocessorcontrol_image->data = preprocess_canny(control_image->data,control_image->width,control_image->height,0.08f,0.08f,0.8f,1.0f,false);}}// generate imagesd_image_t* results;if (params.mode == TXT2IMG) {results = txt2img(sd_ctx,params.prompt.c_str(),params.negative_prompt.c_str(),params.clip_skip,params.cfg_scale,params.width,params.height,params.sample_method,params.sample_steps,params.seed,params.batch_count,control_image,params.control_strength,params.style_ratio,params.normalize_input,params.input_id_images_path.c_str());} else {sd_image_t input_image = {(uint32_t)params.width,(uint32_t)params.height,3,input_image_buffer};if (params.mode == IMG2VID) {results = img2vid(sd_ctx,input_image,params.width,params.height,params.video_frames,params.motion_bucket_id,params.fps,params.augmentation_level,params.min_cfg,params.cfg_scale,params.sample_method,params.sample_steps,params.strength,params.seed);if (results == NULL) {printf("generate failed\n");free_sd_ctx(sd_ctx);return 1;}size_t last            = params.output_path.find_last_of(".");std::string dummy_name = last != std::string::npos ? params.output_path.substr(0, last) : params.output_path;for (int i = 0; i < params.video_frames; i++) {if (results[i].data == NULL) continue;std::string final_image_path = i > 0 ? dummy_name + "_" + std::to_string(i + 1) + ".png" : dummy_name + ".png";stbi_write_png(final_image_path.c_str(), results[i].width, results[i].height, results[i].channel,results[i].data, 0, get_image_params(params, params.seed + i).c_str());printf("save result image to '%s'\n", final_image_path.c_str());free(results[i].data);results[i].data = NULL;}free(results);free_sd_ctx(sd_ctx);return 0;} else {results = img2img(sd_ctx,input_image,params.prompt.c_str(),params.negative_prompt.c_str(),params.clip_skip,params.cfg_scale,params.width,params.height,params.sample_method,params.sample_steps,params.strength,params.seed,params.batch_count,control_image,params.control_strength,params.style_ratio,params.normalize_input,params.input_id_images_path.c_str());}}if (results == NULL) {printf("generate failed\n");free_sd_ctx(sd_ctx);return 1;}int upscale_factor = 4;  // unused for RealESRGAN_x4plus_anime_6B.pthif (params.esrgan_path.size() > 0 && params.upscale_repeats > 0) {upscaler_ctx_t* upscaler_ctx = new_upscaler_ctx(params.esrgan_path.c_str(),params.n_threads,params.wtype);if (upscaler_ctx == NULL) printf("new_upscaler_ctx failed\n");else {for (int i = 0; i < params.batch_count; i++) {if (results[i].data == NULL) {continue;}sd_image_t current_image = results[i];for (int u = 0; u < params.upscale_repeats; ++u) {sd_image_t upscaled_image = upscale(upscaler_ctx, current_image, upscale_factor);if (upscaled_image.data == NULL) {printf("upscale failed\n");break;}free(current_image.data);current_image = upscaled_image;}results[i] = current_image;  // Set the final upscaled image as the result}}}size_t last            = params.output_path.find_last_of(".");std::string dummy_name = last != std::string::npos ? params.output_path.substr(0, last) : params.output_path;for (int i = 0; i < params.batch_count; i++) {if (results[i].data == NULL) continue;std::string final_image_path = i > 0 ? dummy_name + "_" + std::to_string(i + 1) + ".png" : dummy_name + ".png";stbi_write_png(final_image_path.c_str(), results[i].width, results[i].height, results[i].channel,results[i].data, 0, get_image_params(params, params.seed + i).c_str());printf("save result image to '%s'\n", final_image_path.c_str());free(results[i].data);results[i].data = NULL;}free(results);free_sd_ctx(sd_ctx);free(control_image_buffer);free(input_image_buffer);return 0;
}

运行结果

ggml_cuda_init: GGML_CUDA_FORCE_MMQ:   no
ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes
ggml_cuda_init: found 1 CUDA devices:Device 0: NVIDIA GeForce GTX 1060 with Max-Q Design, compute capability 6.1, VMM: yes
[INFO ] stable-diffusion.cpp:169  - loading model from './v1-5-pruned-emaonly.ckpt'
[INFO ] model.cpp:736  - load ./v1-5-pruned-emaonly.ckpt using checkpoint format
[INFO ] stable-diffusion.cpp:192  - Stable Diffusion 1.x
[INFO ] stable-diffusion.cpp:198  - Stable Diffusion weight type: f32
[INFO ] stable-diffusion.cpp:419  - total params memory size = 2719.24MB (VRAM 2719.24MB, RAM 0.00MB): clip 469.44MB(VRAM), unet 2155.33MB(VRAM), vae 94.47MB(VRAM), controlnet 0.00MB(VRAM), pmid 0.00MB(VRAM)
[INFO ] stable-diffusion.cpp:423  - loading model from './v1-5-pruned-emaonly.ckpt' completed, taking 18.72s
[INFO ] stable-diffusion.cpp:440  - running in eps-prediction mode
[INFO ] stable-diffusion.cpp:556  - Attempting to apply 0 LoRAs
[INFO ] stable-diffusion.cpp:1203 - apply_loras completed, taking 0.00s
ggml_gallocr_reserve_n: reallocating CUDA0 buffer from size 0.00 MiB to 1.40 MiB
ggml_gallocr_reserve_n: reallocating CUDA0 buffer from size 0.00 MiB to 1.40 MiB
[INFO ] stable-diffusion.cpp:1316 - get_learned_condition completed, taking 514 ms
[INFO ] stable-diffusion.cpp:1334 - sampling using Euler A method
[INFO ] stable-diffusion.cpp:1338 - generating image: 1/1 - seed 42
ggml_gallocr_reserve_n: reallocating CUDA0 buffer from size 0.00 MiB to 559.90 MiB|==================================================| 20/20 - 1.40s/it
[INFO ] stable-diffusion.cpp:1381 - sampling completed, taking 35.05s
[INFO ] stable-diffusion.cpp:1389 - generating 1 latent images completed, taking 35.07s
[INFO ] stable-diffusion.cpp:1392 - decoding 1 latents
ggml_gallocr_reserve_n: reallocating CUDA0 buffer from size 0.00 MiB to 1664.00 MiB
[INFO ] stable-diffusion.cpp:1402 - latent 1 decoded, taking 3.03s
[INFO ] stable-diffusion.cpp:1406 - decode_first_stage completed, taking 3.03s
[INFO ] stable-diffusion.cpp:1490 - txt2img completed in 38.64s
save result image to './gen_img.png'

注:

  • stable_diffusion支持的模型文件需要自己去下载,推荐到huggingface官网下载ckpt格式文件
  • 提示词要使用英文
  • 支持文字生成图和以图辅助生成图,参数很多,可以多尝试

源码

stable_diffusion_cpp_starter

这篇关于开源C++版AI画图大模型框架stable-diffusion.cpp开发使用初体验的文章就介绍到这儿,希望我们推荐的文章对编程师们有所帮助!



http://www.chinasem.cn/article/1088143

相关文章

使用Docker构建Python Flask程序的详细教程

《使用Docker构建PythonFlask程序的详细教程》在当今的软件开发领域,容器化技术正变得越来越流行,而Docker无疑是其中的佼佼者,本文我们就来聊聊如何使用Docker构建一个简单的Py... 目录引言一、准备工作二、创建 Flask 应用程序三、创建 dockerfile四、构建 Docker

Python使用vllm处理多模态数据的预处理技巧

《Python使用vllm处理多模态数据的预处理技巧》本文深入探讨了在Python环境下使用vLLM处理多模态数据的预处理技巧,我们将从基础概念出发,详细讲解文本、图像、音频等多模态数据的预处理方法,... 目录1. 背景介绍1.1 目的和范围1.2 预期读者1.3 文档结构概述1.4 术语表1.4.1 核

Python使用pip工具实现包自动更新的多种方法

《Python使用pip工具实现包自动更新的多种方法》本文深入探讨了使用Python的pip工具实现包自动更新的各种方法和技术,我们将从基础概念开始,逐步介绍手动更新方法、自动化脚本编写、结合CI/C... 目录1. 背景介绍1.1 目的和范围1.2 预期读者1.3 文档结构概述1.4 术语表1.4.1 核

Conda与Python venv虚拟环境的区别与使用方法详解

《Conda与Pythonvenv虚拟环境的区别与使用方法详解》随着Python社区的成长,虚拟环境的概念和技术也在不断发展,:本文主要介绍Conda与Pythonvenv虚拟环境的区别与使用... 目录前言一、Conda 与 python venv 的核心区别1. Conda 的特点2. Python v

Spring Boot中WebSocket常用使用方法详解

《SpringBoot中WebSocket常用使用方法详解》本文从WebSocket的基础概念出发,详细介绍了SpringBoot集成WebSocket的步骤,并重点讲解了常用的使用方法,包括简单消... 目录一、WebSocket基础概念1.1 什么是WebSocket1.2 WebSocket与HTTP

C#中Guid类使用小结

《C#中Guid类使用小结》本文主要介绍了C#中Guid类用于生成和操作128位的唯一标识符,用于数据库主键及分布式系统,支持通过NewGuid、Parse等方法生成,感兴趣的可以了解一下... 目录前言一、什么是 Guid二、生成 Guid1. 使用 Guid.NewGuid() 方法2. 从字符串创建

Python使用python-can实现合并BLF文件

《Python使用python-can实现合并BLF文件》python-can库是Python生态中专注于CAN总线通信与数据处理的强大工具,本文将使用python-can为BLF文件合并提供高效灵活... 目录一、python-can 库:CAN 数据处理的利器二、BLF 文件合并核心代码解析1. 基础合

Python使用OpenCV实现获取视频时长的小工具

《Python使用OpenCV实现获取视频时长的小工具》在处理视频数据时,获取视频的时长是一项常见且基础的需求,本文将详细介绍如何使用Python和OpenCV获取视频时长,并对每一行代码进行深入解析... 目录一、代码实现二、代码解析1. 导入 OpenCV 库2. 定义获取视频时长的函数3. 打开视频文

Spring IoC 容器的使用详解(最新整理)

《SpringIoC容器的使用详解(最新整理)》文章介绍了Spring框架中的应用分层思想与IoC容器原理,通过分层解耦业务逻辑、数据访问等模块,IoC容器利用@Component注解管理Bean... 目录1. 应用分层2. IoC 的介绍3. IoC 容器的使用3.1. bean 的存储3.2. 方法注

Python内置函数之classmethod函数使用详解

《Python内置函数之classmethod函数使用详解》:本文主要介绍Python内置函数之classmethod函数使用方式,具有很好的参考价值,希望对大家有所帮助,如有错误或未考虑完全的地... 目录1. 类方法定义与基本语法2. 类方法 vs 实例方法 vs 静态方法3. 核心特性与用法(1编程客