人工智能太疯狂,传统劳动力和内容创作平台被AI枪毙,弃尸尘埃。并非空穴来风,也不是危言耸听,人工智能AI图像增强框架ControlNet正在疯狂地改写绘画艺术的倒退过程,你问我绘画行业将来的样子?我只好指着ControlNet的方向。本次咱们在M1/M2芯片的Mac零碎下,体验人工智能登峰造极的绘画艺术。
本地装置和配置ControlNet
ControlNet在HuggingFace训练平台上也有体验版,请参见: https://huggingface.co/spaces...
但因为公共平台算力无限,同时输出参数也受到平台的限度,一次只能训练一张图片,不能让人开怀畅饮。
为了能和史上最平凡的图像增强框架ControlNet一亲芳泽,咱们抉择本地搭建ControlNet环境,首先运行Git命令拉取官网的线上代码:
git clone https://github.com/lllyasviel/ControlNet.git
拉取胜利后,进入我的项目目录:
cd ControlNet
因为Github对文件大小有限度,所以ControlNet的训练模型只能独自下载,模型都放在HuggingFace平台上:https://huggingface.co/lllyas...,须要留神的是,每个模型的体积都十分微小,达到了5.71G,令人乍舌。
下载好模型后,须要将其放到ControlNet的models目录中:
├── models │ ├── cldm_v15.yaml │ ├── cldm_v21.yaml │ └── control_sd15_canny.pth
这里笔者下载了control\_sd15\_canny.pth模型,即放入models目录中,其余模型也是一样。
随后装置运行环境,官网举荐应用conda虚拟环境,装置好conda后,运行命令激活虚拟环境即可:
conda env create -f environment.yaml conda activate control
但笔者查看了官网的environment.yaml配置文件:
name: control channels: - pytorch - defaults dependencies: - python=3.8.5 - pip=20.3 - cudatoolkit=11.3 - pytorch=1.12.1 - torchvision=0.13.1 - numpy=1.23.1 - pip: - gradio==3.16.2 - albumentations==1.3.0 - opencv-contrib-python==4.3.0.36 - imageio==2.9.0 - imageio-ffmpeg==0.4.2 - pytorch-lightning==1.5.0 - omegaconf==2.1.1 - test-tube>=0.7.5 - streamlit==1.12.1 - einops==0.3.0 - transformers==4.19.2 - webdataset==0.2.5 - kornia==0.6 - open_clip_torch==2.0.2 - invisible-watermark>=0.1.5 - streamlit-drawable-canvas==0.8.0 - torchmetrics==0.6.0 - timm==0.6.12 - addict==2.4.0 - yapf==0.32.0 - prettytable==3.6.0 - safetensors==0.2.7 - basicsr==1.4.2
一望而知,Python版本是老旧的3.8,Torch版本1.12并不反对Mac独有的Mps训练模式。
同时,Conda环境也有一些毛病:
环境隔离可能会导致一些问题。尽管虚拟环境容许您治理软件包的版本和依赖关系,但有时也可能导致环境抵触和奇怪的谬误。
Conda环境能够占用大量磁盘空间。每个环境都须要独立的软件包正本和依赖项。如果须要创立多个环境,这可能会导致磁盘空间有余的问题。
软件包可用性和兼容性也可能是一个问题。Conda环境可能不蕴含某些软件包或库,或者可能不反对特定操作系统或硬件架构。
在某些状况下,Conda环境的创立和治理可能会变得复杂和耗时。如果须要治理多个环境,并且须要在这些环境之间频繁切换,这可能会变得艰难。
所以咱们也能够用最新版的Python3.10来构建ControlNet训练环境,编写requirements.txt文件:
pytorch==1.13.0 gradio==3.16.2 albumentations==1.3.0 opencv-contrib-python==4.3.0.36 imageio==2.9.0 imageio-ffmpeg==0.4.2 pytorch-lightning==1.5.0 omegaconf==2.1.1 test-tube>=0.7.5 streamlit==1.12.1 einops==0.3.0 transformers==4.19.2 webdataset==0.2.5 kornia==0.6 open_clip_torch==2.0.2 invisible-watermark>=0.1.5 streamlit-drawable-canvas==0.8.0 torchmetrics==0.6.0 timm==0.6.12 addict==2.4.0 yapf==0.32.0 prettytable==3.6.0 safetensors==0.2.7 basicsr==1.4.2
随后,运行命令:
pip3 install -r requirements.txt
至此,基于Python3.10来构建ControlNet训练环境就实现了,对于Python3.10的装置,请移玉步至:一网成擒全端涵盖,在不同架构(Intel x86/Apple m1 silicon)不同开发平台(Win10/Win11/Mac/Ubuntu)上装置配置Python3.10开发环境,这里不再赘述。
批改训练模式(Cuda/Cpu/Mps)
ControlNet的代码中将训练模式写死为Cuda,CUDA是NVIDIA开发的一个并行计算平台和编程模型,因而不反对NVIDIA GPU的零碎将无奈运行CUDA训练模式。
除此之外,其余不反对CUDA训练模式的零碎可能包含:
没有装置NVIDIA GPU驱动程序的零碎
没有装置CUDA工具包的零碎
应用的NVIDIA GPU不反对CUDA(较旧的GPU型号可能不反对CUDA)
没有足够的GPU显存来运行CUDA训练模式(尤其是在训练大型深度神经网络时须要大量显存)
须要留神的是,即便零碎反对CUDA,也须要确保所应用的机器学习框架反对CUDA,否则无奈应用CUDA进行训练。
咱们能够批改代码将训练模式改为Mac反对的Mps,请参见:闻其声而知雅意,M1 Mac基于PyTorch(mps/cpu/cuda)的人工智能AI本地语音辨认库Whisper(Python3.10),这里不再赘述。
如果代码运行过程中,报上面的谬误:
RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.
阐明以后零碎不反对cuda模型,须要批改几个中央,以我的项目中的gradio\_canny2image.py为例子,须要将gradio\_canny2image.py文件中的cuda替换为cpu,同时批改/ControlNet/ldm/modules/encoders/modules.py文件,将cuda替换为cpu,批改/ControlNet/cldm/ddim\_hacked.py文件,将cuda替换为cpu。至此,训练模式就改成cpu了。
开始训练
批改完代码后,间接在终端运行gradio\_canny2image.py文件:
python3 gradio_canny2image.py
程序返回:
➜ ControlNet git:(main) ✗ /opt/homebrew/bin/python3.10 "/Users/liuyue/wodfan/work/ControlNet/gradio_cann y2image.py" logging improved. No module 'xformers'. Proceeding without it. /opt/homebrew/lib/python3.10/site-packages/pytorch_lightning/utilities/distributed.py:258: LightningDeprecationWarning: `pytorch_lightning.utilities.distributed.rank_zero_only` has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from `pytorch_lightning.utilities` instead. rank_zero_deprecation( ControlLDM: Running in eps-prediction mode DiffusionWrapper has 859.52 M params. making attention of type 'vanilla' with 512 in_channels Working with z of shape (1, 4, 32, 32) = 4096 dimensions. making attention of type 'vanilla' with 512 in_channels Loaded model config from [./models/cldm_v15.yaml] Loaded state_dict from [./models/control_sd15_canny.pth] Running on local URL: http://0.0.0.0:7860 To create a public link, set `share=True` in `launch()`.
此时,在本地零碎的7860端口上会运行ControlNet的Web客户端服务。
拜访 http://localhost:7860,就能够间接上传图片进行训练了。
这里以本站的Logo图片为例子:
通过输出疏导词和其余训练参数,就能够对现有图片进行扩散模型的加强解决,这里的疏导词的意思是:红宝石、黄金、油画。训练后果堪称是言有尽而意无穷了。
除了主疏导词,零碎默认会增加一些辅助疏导词,比方要求图像品质的best quality, extremely detailed等等,残缺代码:
from share import * import config import cv2 import einops import gradio as gr import numpy as np import torch import random from pytorch_lightning import seed_everything from annotator.util import resize_image, HWC3 from annotator.canny import CannyDetector from cldm.model import create_model, load_state_dict from cldm.ddim_hacked import DDIMSampler apply_canny = CannyDetector() model = create_model('./models/cldm_v15.yaml').cpu() model.load_state_dict(load_state_dict('./models/control_sd15_canny.pth', location='cpu')) model = model.cpu() ddim_sampler = DDIMSampler(model) def process(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, guess_mode, strength, scale, seed, eta, low_threshold, high_threshold): with torch.no_grad(): img = resize_image(HWC3(input_image), image_resolution) H, W, C = img.shape detected_map = apply_canny(img, low_threshold, high_threshold) detected_map = HWC3(detected_map) control = torch.from_numpy(detected_map.copy()).float().cpu() / 255.0 control = torch.stack([control for _ in range(num_samples)], dim=0) control = einops.rearrange(control, 'b h w c -> b c h w').clone() if seed == -1: seed = random.randint(0, 65535) seed_everything(seed) if config.save_memory: model.low_vram_shift(is_diffusing=False) cond = {"c_concat": [control], "c_crossattn": [model.get_learned_conditioning([prompt + ', ' + a_prompt] * num_samples)]} un_cond = {"c_concat": None if guess_mode else [control], "c_crossattn": [model.get_learned_conditioning([n_prompt] * num_samples)]} shape = (4, H // 8, W // 8) if config.save_memory: model.low_vram_shift(is_diffusing=True) model.control_scales = [strength * (0.825 ** float(12 - i)) for i in range(13)] if guess_mode else ([strength] * 13) # Magic number. IDK why. Perhaps because 0.825**12<0.01 but 0.826**12>0.01 samples, intermediates = ddim_sampler.sample(ddim_steps, num_samples, shape, cond, verbose=False, eta=eta, unconditional_guidance_scale=scale, unconditional_conditioning=un_cond) if config.save_memory: model.low_vram_shift(is_diffusing=False) x_samples = model.decode_first_stage(samples) x_samples = (einops.rearrange(x_samples, 'b c h w -> b h w c') * 127.5 + 127.5).cpu().numpy().clip(0, 255).astype(np.uint8) results = [x_samples[i] for i in range(num_samples)] return [255 - detected_map] + results block = gr.Blocks().queue() with block: with gr.Row(): gr.Markdown("## Control Stable Diffusion with Canny Edge Maps") with gr.Row(): with gr.Column(): input_image = gr.Image(source='upload', type="numpy") prompt = gr.Textbox(label="Prompt") run_button = gr.Button(label="Run") with gr.Accordion("Advanced options", open=False): num_samples = gr.Slider(label="Images", minimum=1, maximum=12, value=1, step=1) image_resolution = gr.Slider(label="Image Resolution", minimum=256, maximum=768, value=512, step=64) strength = gr.Slider(label="Control Strength", minimum=0.0, maximum=2.0, value=1.0, step=0.01) guess_mode = gr.Checkbox(label='Guess Mode', value=False) low_threshold = gr.Slider(label="Canny low threshold", minimum=1, maximum=255, value=100, step=1) high_threshold = gr.Slider(label="Canny high threshold", minimum=1, maximum=255, value=200, step=1) ddim_steps = gr.Slider(label="Steps", minimum=1, maximum=100, value=20, step=1) scale = gr.Slider(label="Guidance Scale", minimum=0.1, maximum=30.0, value=9.0, step=0.1) seed = gr.Slider(label="Seed", minimum=-1, maximum=2147483647, step=1, randomize=True) eta = gr.Number(label="eta (DDIM)", value=0.0) a_prompt = gr.Textbox(label="Added Prompt", value='best quality, extremely detailed') n_prompt = gr.Textbox(label="Negative Prompt", value='longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality') with gr.Column(): result_gallery = gr.Gallery(label='Output', show_label=False, elem_id="gallery").style(grid=2, height='auto') ips = [input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, guess_mode, strength, scale, seed, eta, low_threshold, high_threshold] run_button.click(fn=process, inputs=ips, outputs=[result_gallery]) block.launch(server_name='0.0.0.0')
其余的模型,比方gradio\_hed2image.py,它能够保留输出图像中的许多细节,适宜图像的从新着色和款式化的场景:
还记得AnimeGANv2模型吗:神工鬼斧惟肖惟妙,M1 mac零碎深度学习框架Pytorch的二次元动漫动画格调迁徙滤镜AnimeGANv2+Ffmpeg(图片+视频)疾速实际,之前还只能通过对立模型滤镜进行转化,当初只有批改疏导词,咱们就能够肆意地变动出不同的滤镜,人工智能技术的倒退,就像发情的海,波澜壮阔。
结语
“人类嘛时候会被人工智能代替呀?”
“就是当初!就在明天!”
就算是达芬奇还魂,齐白石再生,他们也会被现今的人工智能AI技术所震撼,纵横恣肆的笔墨,抑扬变动的状态,左右跌宕的心气,焕然飞动的神采!历史长河中这一刻,大千世界里这一处,让咱们变得疯狂!
最初奉上批改后的基于Python3.10的Cpu训练版本的ControlNet,与众亲同飨:https://github.com/zcxey2911/...\_py3.10\_cpu\_NoConda