Results for “learning”

Explore training workflows that reduce harmful model responses using a written set of principles. The skill covers self-critique, revised training examples, and RLAIF, reinforcement learning from AI feedback, with implementation examples and hardware guidance.

• Designing principles for model safety training
• Creating training examples through self-critique and revision
• Planning reinforcement learning with AI-rated responses

Guide language model training with TRL, from learning to follow instructions to matching human preferences. Covers method selection, training data, reward models that score responses, evaluation, and adjustments for memory limits or unstable training.

• Train a model on instruction and response examples
• Align responses using preferred and rejected answer pairs
• Build a training pipeline using human feedback

Helps adapt machine learning training to multiple GPUs and machines with Ray Train. Covers resource allocation, recovery from saved training states, and Ray Tune searches that compare settings such as learning rate and batch size.

• Adapt existing PyTorch training to multiple GPUs
• Run HuggingFace training across multiple machines
• Compare learning rates and batch sizes with Ray Tune

Create learning reminders about medical guideline, consensus, and standards updates relevant to a doctor's specialty and local care setting. Check authoritative originals and summarize learning topics with source links, without providing patient treatment instructions.

• Request guideline update reminders every three days
• Check consensus updates relevant to a specialty
• Collect original sources for primary care learning

Study verified medical guidelines through section summaries linked to their original locations. The skill explains confusing points, summarizes selected topics, and builds learning sequences with estimated study time and signs of understanding.

• Study a guideline chapter with references to the original text
• Summarize screening, follow-up, or quality-control learning points
• Build a learning sequence that fits available study time

Adjust an existing law student study profile through guided questions. Update classes, learning preferences, bar exam subjects, materials, or study routines one change at a time, with explanations of how related study features will respond.

• Update enrolled classes for a new semester
• Adjust learning style and outline preferences
• Add bar exam subjects and flag weak areas

Translates Splunk SPL queries into Axiom APL, the query language used to search and analyze data in Axiom. Covers command and function equivalents, time filters, type conversions, and common patterns for migration or learning.

• Migrating Splunk queries to Axiom
• Converting log aggregation and error-rate queries
• Learning APL through familiar SPL patterns

Learn from top Kaggle solutions across text, images, time series, tabular data, and multimodal tasks that combine data types. Explore technical analyses, model settings, reusable code templates, and practical lessons to inform machine learning projects.

• Study top solutions for a Kaggle competition
• Find proven methods for a machine learning task
• Explore reusable machine learning code templates

Learn from top Kaggle solutions across text, images, time series, tabular data, and multimodal tasks that combine data types. Explore technical analyses, model settings, reusable code templates, and practical lessons to inform machine learning projects.

• Study top solutions for a Kaggle competition
• Find proven methods for a machine learning task
• Explore reusable machine learning code templates

Build reproducible empirical analyses in R, from data cleaning to statistical modeling and publication tables and figures. Covers economics, epidemiology, and machine learning approaches to causal inference, which estimates effects of interventions.

• Build a reproducible R or Quarto analysis workflow
• Compare regression models with progressively added controls
• Test robustness across samples and statistical methods

Search stored learnings from past sessions by meaning or keywords. Retrieve relevant entries in full, along with their type, confidence, and session context, to revisit earlier findings about development patterns, installation, or errors.

• Revisiting hook development patterns
• Finding earlier wizard installation learnings
• Retrieving past findings about TypeScript errors

Build reinforcement learning experiments, where agents learn actions from rewards, using Stable Baselines3. Get guidance on choosing algorithms, creating Gymnasium environments, training and evaluating agents, monitoring progress, and saving models for later use.

• Choosing an algorithm for a reinforcement learning experiment
• Creating and checking a custom Gymnasium environment
• Monitoring training and saving checkpoints

Analyze tabular datasets too large to fit in memory using Vaex. Filter rows, calculate statistics, visualize distributions, convert file formats, and prepare machine learning workflows with operations that defer computation until results are needed.

• Aggregate CSV files too large to fit in memory
• Explore large datasets with histograms and heatmaps
• Convert CSV data to HDF5 or Arrow

Organize code and data projects into clear folders with documented workflows, data fields, dependencies, and changes. The skill guides new machine learning and data processing scripts, while preserving existing project structure during modifications.

• Creating a documented machine learning training project
• Organizing data cleaning code and field definitions
• Updating scripts while preserving project structure

Explain machine learning predictions with SHAP values, which measure how each input contributes to an output. Create charts for individual predictions and overall feature importance, investigate errors, compare models, and examine differences between groups.

• Explaining an individual prediction
• Visualizing overall feature importance
• Investigating misclassifications and data leakage

Build and train reinforcement learning agents, which learn actions through rewards. The skill guides custom environment creation, parallel simulation tuning, integration with existing environments, and policy development for image or sequential observations.

• Train agents in existing reinforcement learning environments
• Create custom simulation environments
• Tune parallel simulations for higher throughput

Evaluate trained machine learning models using metrics suited to the task. Compare candidates consistently, review decision thresholds and probability estimates, and identify weaknesses in validation before deployment or publication.

• Compare candidate models using consistent metrics
• Review tradeoffs between decision thresholds
• Check whether predicted probabilities match observed outcomes

Analyze genomic regions with gtars using Python or command-line tools. Find overlaps, generate coverage tracks showing read density, prepare regions as tokens for machine learning, and split single-cell fragment data for further analysis.

• Finding overlaps between peaks and promoter regions
• Generating BigWig coverage tracks from sequencing fragments
• Converting genomic regions into tokens for machine learning

Develop workflows for satellite imagery and geographic data, using guidance and code examples to calculate vegetation indicators, classify land cover, analyze terrain, and find road routes. Covers GIS, software for working with location-based information, and spatial machine learning.

• Tracking vegetation changes in satellite imagery
• Classifying land cover with machine learning
• Aggregating location data by region

Explain an existing machine learning model by summarizing influential input features and examining individual predictions. It helps investigate unexpected feature interactions and communicate model behavior to nontechnical stakeholders, while noting limitations and potentially misleading explanations.

• Explaining why a model made a particular prediction
• Identifying influential input features
• Investigating unexpected feature interactions

Convert existing Hugging Face Trainer or TRL SFTTrainer code into an NVFLARE job for federated learning, where multiple sites train a shared model. Preserve training settings, validate locally, and export the resulting job.

• Converting Trainer-based training into a federated job
• Adapting PEFT or LoRA fine-tuning for federated learning
• Preserving evaluation metrics during conversion

Convert existing PyTorch training code into an NVFLARE federated learning job, where sites train with their own data. Adapt training and evaluation, exchange model updates through the Client API, validate locally, and export the job.

• Convert a manual PyTorch training loop to federated learning
• Adapt existing evaluation code for federated rounds
• Validate and export a FedAvg training job

Study a topic, document, or codebase through questions that adapt to your progress. Practice recalling answers, receive hints when stuck, and save a learning record that guides spaced reviews and the next session.

• Learning a new topic through questions
• Reviewing the contents of a document or URL
• Studying how a codebase works

[omh] Paper or paper PDF to understand: explain a supplied paper or paper/PDF at a selected level while preserving full section coverage and source evidence boundaries. Use when the user says: paper-learning, paper learning, paper-explainer, paper explainer, paper explanation, explain this paper, explain this arxiv paper, paper walkthrough.

The English summary is being prepared. Showing the original description.