Model Training — end-to-end ML training with checkpoints and tracking
A complete training lifecycle workflow — data loading and profiling, preprocessing pipelines, architecture selection, optimizer/scheduler configuration, validation loops with early stopping, checkpointing, and MLflow/W&B experiment tracking across PyTorch, TensorFlow and scikit-learn
- What
- A complete training lifecycle workflow — data loading and profiling, preprocessing pipelines, architecture selection, optimizer/scheduler configuration, validation loops with early stopping, checkpointing, and MLflow/W&B experiment tracking across PyTorch, TensorFlow and scikit-learn
- Cost
- Free
- Needs
- Python 3.10+ with PyTorch/TensorFlow/scikit-learn (GPU recommended for deep learning); your own dataset; an experiment tracker (MLflow or W&B) for the full workflow
- Install
- Copy the installer prompt below into your Muse — your agent does the rest.
Curated by Skill Harbor — @seb1n's model-training skill, listed here with credit to its creator: a complete training-lifecycle workflow that lets an agent train machine-learning models end to end — loading and profiling data (feature distributions, class balance, missing values), stratified train/validation/test splits, reproducible preprocessing pipelines (normalization, tokenization, augmentation) serializable for inference, architecture selection (gradient boosting/SVMs for classical ML; PyTorch/TensorFlow layers with dropout and weight decay for deep learning; pre-trained backbones for transfer learning), training configuration (Adam/SGD/AdamW, cross-entropy/MSE/focal loss, cosine/step/warmup schedules, mixed precision with torch.amp), training loops with validation and early stopping, checkpointing of the best validation score, and export to portable formats (ONNX, TorchScript, SavedModel), with hyperparameters and metrics logged to MLflow or Weights & Biases. It ships worked examples: a PyTorch text classifier (LSTM, cosine annealing, early stopping, best_model.pt) and a Hugging Face Transformers fine-tuning recipe. Honest caveats: training consumes real compute (GPU time costs money — start with the small deterministic fixture it suggests); the examples train on simulated data, so adapt them to your own datasets; exported artifacts and hyperparameters can leak training-data details — keep them in a registry with proper access controls. MIT licensed. Skill Harbor never reviews the code, review it yourself before use. Discovered via skills.sh.
Version:
Install
Prerequisites: Python 3.10+ with PyTorch/TensorFlow/scikit-learn (GPU recommended for deep learning); your own dataset; an experiment tracker (MLflow or W&B) for the full workflow Install "Model Training — end-to-end ML training with checkpoints and tracking" for me. It gives my agent @seb1n's complete training-lifecycle workflow: data loading and profiling, stratified splits, reproducible preprocessing pipelines, architecture selection (classical or deep learning), optimizer/loss/scheduler configuration with mixed precision, validation loops with early stopping, checkpointing and export (ONNX/TorchScript/SavedModel), and experiment tracking — plus a worked PyTorch text-classifier example and a Hugging Face fine-tuning recipe. MIT licensed. Repository: https://github.com/seb1n/awesome-ai-agent-skills/blob/main/ai-ml-operations/model-training/SKILL.md 1. Fetch the SKILL.md file (and any helper files) from the repository path into a temporary folder and summarize what it does in one or two sentences. 2. Safety check: review the SKILL.md and scripts for anything suspicious (unexpected network calls, shell commands, credential harvesting). Training scripts can pull datasets and large files — verify every download source is expected. This repo should contain zero secrets in code. Verify that holds here; STOP on any red flag and tell me. 3. Install it as a skill: copy SKILL.md and its helper files into the agent's skills directory, in a folder named "model-training". 4. Verify with no network calls: frontmatter valid, files in place. 5. Report what was installed, where, and what I still need to do myself (e.g. install my framework of choice, provide my dataset, start with a small run — GPU time costs money — and keep exported artifacts in a registry with proper access controls). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.
Questions
How do I install a build?
Every product page includes a copy-paste install prompt. Paste it into your Muse and it sets the build up for you — no manual configuration.
Where does my money go?
Straight to the seller. Skill Harbor never processes payments: checkout happens on the seller’s own page, usually Stripe.
What does the ✓ next to a creator’s name mean?
It means we confirmed the identity of the person behind the listing. It says nothing about the code itself — always check a build before installing it.