- name
SKILL.md
O arquivo da skill permanece no repositório do publicador.
Cole o comando copiado em um terminal. Ele instala este SKILL.md nesse agente.
npx skills add artokun/comfyui-mcp --skill anima-lora-trainer -a cursor -yCole o comando copiado em um terminal. Ele instala este SKILL.md nesse agente.
npx skills add artokun/comfyui-mcp --skill anima-lora-trainer -a claude-code -yCole o comando copiado em um terminal. Ele instala este SKILL.md nesse agente.
npx skills add artokun/comfyui-mcp --skill anima-lora-trainer -a codex -yCole o comando copiado em um terminal. Ele instala este SKILL.md nesse agente.
npx skills add artokun/comfyui-mcp --skill anima-lora-trainer -a grok -yEnvie a mensagem copiada em um chat do Grok Bot. O Bot busca este SKILL.md no GitHub e o adiciona às skills que todos os seus Bots compartilham.
Install this skill: https://github.com/artokun/comfyui-mcp/tree/HEAD/plugin/skills/anima-lora-trainerDo publicador
Citron's Anima LoRA Trainer (app.py, titled "Citron's Anima LoRA Trainer" in the UI) is a local Gradio UI for training LoRA adapters on the Anima diffusion model using kohya-ss/sd-scripts. It trains on ~6GB VRAM with the default settings, the same low-VRAM profile as Anima generation.
https://github.com/citronlegacy/citron-anima-lora-trainer-ui. The Aitrepreneur adaptive installers clone the fork https://github.com/aitrepreneur/citron-anima-lora-trainer-ui.kohya-ss/sd-scripts (https://github.com/kohya-ss/sd-scripts), launched via accelerate launch..safetensors LoRA usable directly in the anima-base ComfyUI workflow.The network module is
networks.lora_animaand the training script issd-scripts/anima_train_network.py(an Anima-specific kohya script the installer expects). Confirm these exist after the installer'sgit cloneof sd-scripts.app.pyreferences them, but they are pulled from the upstream repo at install time.
Run CITRON_ANIMA_LORA_TRAINER-V2.bat. It:
app.py defaults (base_model → anima-preview3-base, mixed_precision → detected value), writes app_configs/accelerate_gpu.yaml..venv, installs PyTorch, clones and installs sd-scripts, installs app requirements.txt.models/anima/{dit,text_encoder,vae}/ from https://huggingface.co/circlestone-labs/Anima/resolve/main/split_files/...:
dit/anima-base-v1.0.safetensors (~4GB)text_encoder/qwen_3_06b_base.safetensors (~1.19GB)vae/qwen_image_vae.safetensors (~254MB)run_anima_base_windows.bat.Run CITRON_ANIMA_LORA_TRAINER-RUNPOD-V2.sh. Same flow into /workspace/citron-anima-lora-trainer-ui; it patches server_name to 0.0.0.0. Expose HTTP port 7860 and open Connect → HTTP Service 7860 (or https://${RUNPOD_POD_ID}-7860.proxy.runpod.net).
app.py runs Gradio on 0.0.0.0:7860, so open http://127.0.0.1:7860. Re-launch later with run_anima_base_windows.bat (Win) or ./run_anima_base_runpod.sh (RunPod). The DiT base model auto-downloads on the first "Start Training" if not already present (uses wget).
A flat folder of images, each with a matching .txt caption of the same basename (image-side captioning, kohya style):
my_dataset/
001.png 001.txt
002.jpg 002.txt
...
.jpg .jpeg .png .webp .bmp .gif..txt.caption_extension = .txt; shuffle_caption = false; caption_dropout_rate default 0.1 (set per dataset).The UI tab "Training" takes Image Directory (the flat folder above) and Output Directory (where the LoRA is saved). "Configure Training" validates the dataset, prints a step estimate (steps_per_epoch = ceil(images × repeats / (batch × grad_accum)), total = spe × epochs), then writes two TOMLs into configs/.
app.py)Fixed in the generated training TOML (not exposed): network_module = networks.lora_anima, network_train_unet_only = true, qwen3_max_token_length = 512, t5_max_token_length = 512, save_model_as = safetensors, save_precision = bf16 (fp16 on older GPUs).
configs/<project>_training_<timestamp>.toml references the DiT (pretrained_model_name_or_path), qwen3 text encoder, and vae paths from models/anima/, plus all params above.
configs/<project>_dataset_<timestamp>.toml:
[general]
resolution = 768
enable_bucket = true
bucket_no_upscale = false
bucket_reso_steps = 64
min_bucket_reso = 256
max_bucket_reso = 4096
[[datasets]]
resolution = 768
[[datasets.subsets]]
num_repeats = 10
image_dir = "/path/to/my_dataset"
caption_extension = ".txt"
caption_dropout_rate = 0.1
"Start Training" runs the following and streams logs live to the UI and to logs/<project>_<timestamp>.log:
accelerate launch \
--config_file app_configs/accelerate_gpu.yaml \
--num_cpu_threads_per_process 1 \
--gpu_ids 0 \
sd-scripts/anima_train_network.py \
--config_file configs/<project>_training_<timestamp>.toml \
--dataset_config configs/<project>_dataset_<timestamp>.toml
accelerate_gpu.yaml pins use_cpu: false, mixed_precision: <bf16|fp16>, single process/machine. CUDA_VISIBLE_DEVICES is set to the selected GPU index.
<project_name>.safetensors, plus per-epoch checkpoints (the last save_last_n_epochs are kept).models/loras/ and load it in the anima-base workflow via LoraLoaderModelOnly (or rgthree Power Lora Loader):
{ "class_type": "LoraLoaderModelOnly",
"inputs": { "model": ["<unet>", 0], "lora_name": "<project_name>.safetensors", "strength_model": 1.0 } }
network_dim=8 and/or resolution=512. Also keep batch 1 and use AdamW8bit.images × repeats × epochs / (batch × grad_accum). The UI prints the exact estimate before you train.logs/. Training config and last paths persist in config.json so you can re-run.sd-scripts/anima_train_network.py and networks.lora_anima come from the kohya fork pulled at install time. app.py expects them, but they are not in the local downloaded files here.<project_name>.safetensors per output_name; confirm in your Output Directory after a run.Train a custom anime LoRA on the ANIMA base model with Citron's local Gradio trainer (kohya sd-scripts), <6GB VRAM, character/style LoRAs; covers setup, dataset prep, training params, and using the re
---name: anima-lora-trainerdescription: Train a custom anime LoRA on the ANIMA base model with Citron's local Gradio trainer (kohya sd-scripts), <6GB…---# plugin/skills/anima-lora-trainer/SKILL.mdVisão geral
Train a custom anime LoRA on the ANIMA base model with Citron's local Gradio trainer (kohya sd-scripts), <6GB VRAM, character/style LoRAs; covers setup, dataset prep, training params, and using the result in the anima-base workflow
Perguntas frequentes
Baixe plugin/skills/anima-lora-trainer/SKILL.md do publicador ou copie o caminho e abra o repositório. O PlugMCP lista a skill e não hospeda o arquivo.
O publicador o mantém em https://github.com/artokun/comfyui-mcp.
Claude, Cursor, Codex e Grok Build conseguem ler um SKILL.md que você adiciona a partir do repositório do publicador. O Grok Bot o instala quando você envia em um chat o link da pasta da skill no GitHub.