[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-80076":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":23,"hasPages":23,"topics":25,"createdAt":10,"pushedAt":10,"updatedAt":32,"readmeContent":33,"aiSummary":34,"trendingCount":16,"starSnapshotCount":16,"syncStatus":15,"lastSyncTime":35,"discoverSource":36},80076,"Anima-TrainFlow","ThetaCursed\u002FAnima-TrainFlow","ThetaCursed","The most efficient one-page LoRA trainer for Anima 2B. Optimized for 6GB+ VRAM, featuring a smart dataset analyzer and real-time previews.","",null,"Python",107,11,4,2,0,1,19,41,6,58.84,"MIT License",false,"main",[26,27,28,29,30,31],"anima-2b","gradio-ui","lora-training","low-vram","prodigy-optimizer","sd-scripts","2026-06-12 04:01:26","# Anima TrainFlow\n\nAnima TrainFlow is a streamlined, one-page GUI for training LoRA on the **Anima 2B** model. Optimized to run on hardware with as little as **6GB of VRAM**, it eliminates technical overhead by focusing on the essential settings that impact training results the most.\n\n![Anima TrainFlow Interface Preview](preview.png)\n\n## Quick Start (Portable)\n1. [**Download Portable Version (3GB)**](https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002FThetaCursed\u002FAnima-TrainFlow-Portable\u002Fresolve\u002Fmain\u002FAnima-TrainFlow-v1.1.0-Portable.7z?download=true)\n2. **Extract the archive using [7-Zip](https:\u002F\u002F7-zip.org\u002F) or WinRAR**.\n3. Run `start_trainer.bat`.\n4. Open the `🔧 Paths to Models \u003C- Set Once` accordion and specify the paths to your model files.\n5. Specify your **Dataset Path** (images + .txt files) and **Trigger Word**, then click **Start**.\n\n## Manual Installation\nIf you prefer to set up the environment manually instead of using the portable version, follow these steps:\n\n1. **Clone the repository:**\n   ```bash\n   git clone https:\u002F\u002Fgithub.com\u002FThetaCursed\u002FAnima-TrainFlow\n   cd Anima-TrainFlow\n   ```\n\n2. **Install dependencies:**  \n   Run `Install_Requirements.bat`\n\n3. **Download Required Models:**\n   Run the following commands from the root folder:\n\n   *   **WD Tagger** (used for auto-captioning):\n       ```bash\n       git clone https:\u002F\u002Fhuggingface.co\u002FSmilingWolf\u002Fwd-eva02-large-tagger-v3 models\u002Fwd-eva02-large-tagger-v3\n       ```\n   *   **U2Net Model** (used for Smart Cropping):\n       ```bash\n       curl -L -o models\u002Fu2net\u002Fu2net.onnx https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fu2net.onnx\n       ```\n\n4. **Launch the Trainer:** \n   Run `start_trainer.bat`\n\n## Key Features\n* **Zero-Tab Interface:** All critical parameters (Trigger Word, Rank, LR, Steps) are accessible on a single screen.\n* **Live Training Previews:** Watch your LoRA improve in real-time. The built-in gallery automatically updates whenever a new sample is generated.\n* **AI-Powered Smart Cropping:** Integrated U2Net model automatically performs subject-aware, head-first cropping and resizes images to optimal aspect-ratio buckets via multi-threading.\n* **Built-in Auto-Captioning:** Integrated WD14 Tagger (EVA02 v3) automatically generates multi-threaded `.txt` tags for your dataset with customizable general and character thresholds.\n* **Pre-Flight Validation:** Automatically scans the dataset for missing captions, oversized images (>=2048px), and missing model paths to prevent crashes before training starts.\n* **Persistent Sessions:** All UI inputs, paths, and slider positions are instantly auto-saved and restored on the next launch.\n* **Portable Edition:** Includes an embedded Python environment to avoid installation or complex setup.\n* **Low VRAM Friendly:** Specifically tuned for 6GB+ NVIDIA GPUs with aggressive RAM\u002FVRAM clearing between tasks.\n* **Optimized Defaults:** Pre-configured for BF16 precision and latent caching to ensure maximum performance and stability.\n* **Prodigy Native:** Intelligent Learning Rate handling and optimized defaults for the Prodigy optimizer.\n\n## Dataset Preparation\nPlace all your training images (.png, .jpg, .webp) in a single folder. Every image must have a matching text file with the same name containing its tags\u002Fcaptions (e.g., `image1.png` and `image1.txt`). You can easily generate these text files using the built-in **Auto-Caption Dataset** tool.\n\n## System Requirements\n* **OS:** Windows 10\u002F11.\n* **GPU:** NVIDIA GPU (6GB+ VRAM recommended for Anima 2B training).\n* **Storage:** ~6.5GB of free space (SSD recommended).\n\n## Technical Details\n* **Core:** Based on a modified version of `sd-scripts` for Anima 2B architecture.\n* **UI:** Built with Gradio featuring a customized dark theme.\n* **Backend:** Utilizes `accelerate launch` for optimized execution.\n* **Auto-Save:** All paths and configurations are automatically saved to `settings.json`.","Anima TrainFlow 是一个专为 Anima 2B 模型设计的高效单页面 LoRA 训练工具。该项目通过优化仅需6GB VRAM即可运行，特别适合硬件资源有限的用户。它提供了一个简洁的界面，将所有关键参数集中在一页上，并具备实时预览功能，使用户能够即时看到训练效果。此外，项目还集成了智能裁剪和自动标注功能，利用AI技术自动处理数据集，提高训练效率。非常适合需要在较低配置硬件上进行模型微调的研究人员或开发者使用。","2026-06-11 03:59:09","CREATED_QUERY"]