套件安裝 - HackMD

# 待開源 * https://mdnice.com/writing/68ce8d0aa0574184a4fc9ad96309f328 * Outfit Anyone * [tryondiffusion](https://github.com/tryonlabs/tryondiffusion) * [diffuse2choosehttps](https://diffuse2choose.github.io/) # [OminiControlGP](https://github.com/deepbeepmeep/OminiControlGP) # [Wan2.1](https://github.com/Wan-Video/Wan2.1)(2/25開源) * 照github安裝即可 * conda activate wen * python generate.py --task t2v-14B --size 1280*720 --ckpt_dir ./Wan2.1-T2V-14B --prompt "Two anthropomorphic cats in comfy boxing gear and bright gloves fight intensely on a spotlighted stage." ## comfyui * prompt * 擁抱:best quality video of a handsome boy and a cute girl embrace with each other * 親吻:best quality video of a handsome boy and a cute girl embrace and kiss, * 握手:best quality video of a handsome boy and a girl Shake hands with each other ## [加速版Wan2GP](https://github.com/deepbeepmeep/Wan2GP) # MMAudio * python demo.py --duration=8 --video=<path to video> * python gradio_demo.py # [PuLID](https://github.com/ToTheBeginning/PuLID) * [ERROR: Exception in ASGI application](https://github.com/ToTheBeginning/PuLID/issues/61) * python app_v1_1.py --base F:\GenAI\PuLID\models\Juggernaut-XL-v9 # [Seed-VC](https://github.com/Plachtaa/seed-vc/blob/main/README-ZH.md) # [ IMAGDressing](https://github.com/v3ucn/IMAGdressing_WebUi_For_Windows) * 改運行.bat跟app.py中的python路徑 * 安裝insightface * 安裝onnxruntime-gpus * 安裝cuda --- * 改程式碼 * IMAGdressing_WebUi_For_Windows\dressing_sd\pipelines\IMAGDressing_v1_pipeline_ipa_controlnet.py加上這行 --- * from diffusers.loaders import LoraLoaderMixin --- * 改 F:\GenAI\IMAGdressing_WebUi_For_Windows\inference_IMAGdressing_controlnetinpainting.py裡的183行 * openpose_model = OpenPose(1) -> openpose_model = OpenPose(0) --- * pip install tb-nightly -i https://mirrors.aliyun.com/pypi/simple * pip install basicsr --- * 改F:\anaconda3\envs\IMAGDressing_window\Lib\site-packages\basicsr\data\degradations.py裡的第8行 * from torchvision.transforms.functional_tensor import rgb_to_grayscale -> from torchvision.transforms._functional_tensor import rgb_to_grayscale * https://mirrors.aliyun.com/pypi/simple ### 沒有這個檔案 python inference_IMAGdressing_counterfeit-v30.py --cloth_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\images\garment\c1.png" --model_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\model.jpg" ### 指定臉型和姿勢穿著指定的衣服可用 python inference_IMAGdressing_ipa_controlnetpose.py --face_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\model.jpg" --pose_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\model.jpg" --cloth_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\images\garment\c1.png" ### 指定模特兒穿指定衣服（實驗功能） python inference_IMAGdressing_controlnetinpainting.py --model_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\model.jpg" --cloth_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\images\garment\c1.png" * 提詞無效 ### 隨機臉孔使用給定的姿勢來穿著給定的服裝 python inference_IMAGdressing_controlnetpose.py --cloth_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\images\garment\c1.png" --pose_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\model.jpg" ### 隨機的臉孔和姿勢來穿上指定的衣服可用 python inference_IMAGdressing.py --cloth_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\images\garment\c1.png" ### 卡通 python inference_IMAGdressing_cartoon_style.py --cloth_path "F:\GenAI\IMAGdressing_WebUi_For_Windows\assets\images\garment\c1.png" # AnyV2V * https://github.com/TIGER-AI-Lab/AnyV2V * 開啟: * cd D:\III_GAI_DIY\GenAI\AnyV2V * conda activate anyv2v-i2vgen-xl * python gradio_demo.py * 使用: * 預處理影片階段:動作影片 * 影像編輯階段:根據第一禎圖片下提詞重繪 * 影片編輯階段:指定動作 * 建議:影像編輯用同類型 ex.男性生成男性 --- # Fatezero * https://github.com/ChenyangQiQi/FateZero * conda activate fatezero38 --- # MimicMotion * https://github.com/Tencent/MimicMotion * 開啟: * cd D:\III_GAI_DIY\GenAI\MimicMotion * conda activate mimicmotion * 遇到huggingface登入問題: * `huggingface-cli login` * shift+insert huggingface token * python inference.py --inference_config configs/test.yaml --- # StoryDiffusion * https://github.com/HVision-NKU/StoryDiffusion --- # T2X * https://github.com/Alpha-VLLM/Lumina-T2X * pip install flash_attn-2.4.1+cu121torch2.1cxx11abiFALSE-cp311-cp311-win_amd64.whl * https://github.com/bdashore3/flash-attention/releases * cuda python win 都要正確 ### 開啟4090 【t2i】(未完成) * conda activate Lumina_T2X * python -u demo.py --ckpt "D:\III_GAI_DIY\GenAI\Lumina-T2X\Lumina-T2I" 【t2i_mini】 * conda activate t2x * python -u demo.py --ckpt "D:\III_GAI_DIY\GenAI\Lumina-T2X\Lumina-Next-SFT 【t2music】(fail) * conda activate Lumina_T2X_music * cd Lumina-T2X/lumina_music ### 開啟3080 【t2i_mini】 * conda activate Lumina_T2X_mini * cd Lumina-T2X/lumina_T2X_mini * python -u demo.py --ckpt "D:\Lumina-T2X\lumina_next_t2i_mini\Lumina-Next-SFT" --hf_token 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 【t2music】(linux) * conda activate Lumina_T2X_music * cd Lumina-T2X/lumina_music * bash run_music.sh ### 安裝錯誤 * TCPStore() RuntimeError：格式字串中不符合的“}” https://github.com/pytorch/pytorch/issues/118378 127.0.0.1->localhost * RuntimeError: Distributed package doesn't have NCCL built in * nccl->gloo * Cannot access gated repo for url https://huggingface.co/google/gemma-2b/resolve/main/config.json. Access to model google/gemma-2b is restricted. You must be authenticated to access it. * 開啟加上 --hf_token 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 * lumina_next convert D:/III_GAI_DIY/GenAI/Lumina-T2X/Lumina-T2I/model_args.pth D:/III_GAI_DIY/GenAI/Lumina-T2X/Lumina-T2I --- # VGen 開啟 * conda activate vgen-py310 * cd D:\III_GAI_DIY\GenAI\VGen * python gradio_app.py 安裝 * 參考:https://github.com/ali-vilab/VGen/issues/142 * 更改requirements.txt * Triton install:https://github.com/PrashantSaikia/Triton-for-Windows * 下載triton-2.0.0-cp310-cp310-win_amd64.whl * pip install triton-2.0.0-cp310-cp310-win_amd64.whl * 把所有.yaml中的 NCCL 改成gloo * 在 Python 3.10 和 miniconda3 運行 * 跟随日志報錯来解决其他錯误 --- # [MagicClothing](https://github.com/shinechen1024/magicclothing) * conda activate magicloth * 衣服+提詞:python gradio_generate.py --model_path "D:\III_GAI_DIY\GenAI\MagicClothing\checkpoints\magic_clothing_768_vitonhd_joint.safetensors" * 衣服+提詞+controlnet:python gradio_controlnet_openpose.py --model_path "D:\III_GAI_DIY\GenAI\MagicClothing\checkpoints\magic_clothing_768_vitonhd_joint.safetensors" * 衣服+提詞controlnet+ipadapter:python gradio_ipadapter_openpose.py --model_path "D:\III_GAI_DIY\GenAI\MagicClothing\checkpoints\magic_clothing_768_vitonhd_joint.safetensors" --enable_cloth_guidance * 模型:https://github.com/ShineChen1024/MagicClothing/issues/65 --- # [oms-Diffusion](https://github.com/ShineChen1024/MagicClothing/tree/earlyAccess?tab=readme-ov-file) python gradio_generate.py --model_path "D:\III_GAI_DIY\GenAI\oms-Diffusion\checkpoints\magic_clothing_768_vitonhd_joint.safetensors" python inference.py --cloth_path "D:\III_GAI_DIY\GenAI\oms-Diffusion\valid_cloth\t1.png" --model_path "D:\III_GAI_DIY\GenAI\oms-Diffusion\checkpoints\magic_clothing_768_vitonhd_joint.safetensors" python gradio_ipadapter_faceid.py --model_path "D:\III_GAI_DIY\GenAI\oms-Diffusion\checkpoints\magic_clothing_768_vitonhd_joint.safetensors" python gradio_animatediff.py --cloth_path "D:\III_GAI_DIY\GenAI\oms-Diffusion\valid_cloth\t1.png" --ckpt_dir "D:\III_GAI_DIY\GenAI\ComfyUI_windows_portable\ComfyUI\models\animatediff_models\AnimateLCM_sd15_t2v.ckpt" * 錯誤:huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': 'D:/III_GAI_DIY/GenAI/oms-Diffusion/checkpoints/ipadapter_faceid/ip-adapter-faceid-plus_sd15_lora.safetensors'. Use `repo_type` argument if needed. * 改為絕對路徑 * 與magicloth共用模型 --- # [Live_Portrait_Monitor](https://github.com/Mrkomiljon/Live_Portrait_Monitor) * conda activate LivePortrait2 * to use monitor `python inference_monitor.py -s assets/examples/source/MY_photo.jpg` * disable pasting back `python inference_monitor.py -s assets/examples/source/s9.jpg -d assets/examples/driving/t.mp4 --no_flag_pasteback` * to use original code for inference ``` python inference_org.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d0.mp4 --no_flag_pasteback ``` * 一鍵開啟(無法使用) * "D:\III_GAI_DIY\Live_Portrait_Monitor_High_FPS_window\Monitor.bat" --- # [Webcam_Live_Portrait](https://github.com/Mrkomiljon/Webcam_Live_Portrait) ## 開這個 * conda activate LivePortrait2 * cd D:\III_GAI_DIY\Webcam_Live_Portrait **即時監控使用webcam** `python inference.py -s assets/examples/source/MY_photo.jpg` **螢幕錄影** `python screen.py -s assets/examples/source/MY_photo.jpg` **螢幕錄影+開webcam** `python test.py -s assets/examples/source/MY_photo.jpg` (3080) * conda activate LivePortrait * cd D:\Webcam_Live_Portrait ## 測試 * 眼睛過大扎眼效果不好 * 可用卡通:白雪公主、長髮公主、娜美、魔法公主、霍爾的移動城堡-蘇菲 * 不可用卡通(動物版可以用、眼睛不會動):蠟筆小新、皮卡丘、小丸子、神奇寶貝小智、真珠美人魚、熊貓 * 結論:要有鼻子、基本五官要有 # [LivePortrait](https://github.com/KwaiVGI/LivePortrait) * conda activate LivePortrait2 * 有gradio `python app.py `人物 `python app_animals.py `動物 * 影片-圖片-影片 `python inference.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d0.mp4` * 影片-影片-影片 `python inference.py -s assets/examples/source/s13.mp4 -d assets/examples/driving/d0.mp4` * gpu error:https://github.com/KwaiVGI/LivePortrait/issues/290 * 一鍵安裝包(開啟gradio): * animal:run_windows_animal.bat * human:run_windows_human.bat ## 開啟 * conda activate lpanimal * cd D:\III_GAI_DIY\LivePortrait-animal --- # [COTTON-size-does-matter](https://github.com/cotton6/COTTON-size-does-matter) ## 常見安裝 * Triton-for-Windows * https://github.com/PrashantSaikia/Triton-for-Windows * whl檔在D:\III_GAI_DIY\GenAI * pip install D:\III_GAI_DIY\GenAI\triton-2.0.0-cp310-cp310-win_amd64.whl python create_seal_api.py -t 小吉籤詩預喜訊，福德漸滿事漸行，進步雖緩現成效，就待春風報喜來 -p D:\III_GAI_DIY\GenAI\Lumina-T2X\lumina_next_t2i_mini\output\e5431b8f9a4d4f0dbb17ba55905371f0.jpg --- # [FasterLivePortrait](https://github.com/warmshao/FasterLivePortrait) * [window整合包](https://drive.google.com/file/d/1fRJ5AJeXLLrI2q5lgnxE42CK04FQ29hw/view) 1. 下載後解壓縮 2. 進入FasterLivePortrait-windows後點擊all_onnx2trt.bat對onnx文件進行轉換 3. 更換目標圖片進入命令行運行 ``` bash camera.bat assets/examples/source/s9.jpg ```