Skill UI

Pyvene is a declarative framework designed for performing causal interventions on PyTorch neural networks. It allows researchers to conduct advanced experiments such as activation patching, causal tracing (ROME-style), and interchange intervention training (IIT). Use this library when you need to test causal hypotheses about model behavior, deeply interpret model components, or ensure reproducibility in advanced AI research.

Automates the setup of ML experiment tracking with MLflow or W&B, configures the environment, initializes servers, and supplies snippets for logging parameters, metrics, and artifacts so runs stay reproducible and easy to compare.

TensorBoard delivers a unified interface to track training metrics, debug network behavior, compare experiments, inspect model graphs, and profile performance for PyTorch or TensorFlow workflows.

Assists with tensorboard visualizer tasks for ML training by guiding setup, best practices, and experiment tracking whenever tensorboard visualization is mentioned, producing production-ready configs and validated outputs.

torchforge is Meta's advanced PyTorch-native library designed for agentic Reinforcement Learning (RL). It fundamentally separates core RL algorithms from complex distributed infrastructure concerns. This allows researchers to rapidly prototype and experiment with novel RL methods (like GRPO) using clean, PyTorch-native abstractions, while automatically handling large-scale tasks like distributed training across thousands of GPUs using Monarch and TorchTitan.

Automates wandb experiment logger tasks within ML training, providing guidance, best practices, and code or configuration examples plus validation to keep data prep, training, hyperparameter tuning, and experiment tracking consistent.

Weights & Biases (W&B) automates experiment logging, real-time dashboards, hyperparameter sweeps, artifact tracking, and model registry/versioning for ML teams using PyTorch, TensorFlow, or HuggingFace workflows.

Denario orchestrates a multi-agent AI to automate research workflows, covering idea generation, methodology development, computational experiments, and LaTeX publication output.

Assess research claims and experimental rigor by evaluating methodology, statistical validity, evidence quality, and biases using GRADE and Cochrane ROB frameworks, plus supporting schematics for complex concepts.

Stable Baselines3 is a robust, PyTorch-based library providing production-ready implementations of core reinforcement learning algorithms (PPO, SAC, DQN, etc.). It facilitates the training of single-agent RL models, supports the creation of custom Gymnasium environments, and handles advanced parallel training via vectorized environments. It is ideal for rapid prototyping and rigorous RL experiments.

This skill orchestrates end-to-end autonomous AI research projects using a sophisticated two-loop architecture. It manages the entire research lifecycle, from initial literature review and hypothesis formation to running constrained experiments and synthesizing findings into publishable papers. It acts as a project manager, autonomously driving the process while structuring the workspace and presenting progress to human users.

This skill transforms a raw research proposal or method idea into a rigorous, paper-ready experimental roadmap. It guides the user through defining core claims, identifying necessary ablations (e.g., novelty, simplicity, frontier necessity), and structuring experiments into actionable, phased stages (Sanity, Baseline, Main, Decision). Ideal for advanced academic research in AI/LLM/VLM fields that requires strong validation and systematic reporting.