skills.rest
sovr610
sovr610COMMUNITY
Explore all skills by this author →
WebsiteGitHub

Task2Vec Task Embeddings + Curriculum/Clustering

Compute Fisher-based Task2Vec embeddings for episodic few-shot tasks and cluster them for curriculum workflows.

Updated Feb 28, 2026
Search Tags:#embeddings#registry#meta-learning#curriculum#clustering#task2vec#fisher
One-click install
npx skills add https://github.com/sovr610/refffiy --skill task2vec-task-embeddings-curriculum-clustering
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: Task2Vec Task Embeddings + Curriculum/Clustering
Source: https://github.com/sovr610/refffiy/tree/main/brain-ai-dev/skills/task2vec-curriculum
Command: npx skills add https://github.com/sovr610/refffiy --skill task2vec-task-embeddings-curriculum-clustering

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires numpy, torch, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

Task space is large and complex; practitioners need a deterministic, end-to-end workflow to convert episodic tasks into compact embeddings that support curriculum design and diverse meta-batches.

Core Features & Use Cases

  • Deterministic Task2Vec embeddings using diagonal Fisher on a fixed probe.
  • Clustering, stability diagnostics, and cohort analysis to reveal task-space structure.
  • Curriculum ordering and diversity-aware meta-batch construction using pre-computed embeddings and an optional registry for persistence.

Quick Start

Instantiate with production config and run extraction on a small synthetic episode to produce the first embedding.

Frequently Asked Questions about Task2Vec Task Embeddings + Curriculum/Clustering

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I compute Task2Vec embeddings for few-shot tasks?▼

Task2Vec embeddings are computed for episodic few-shot tasks using diagonal Fisher information on a fixed probe, producing compact vectors that represent task-space structure for curriculum design.

How do I build a curriculum for meta-learning from task embeddings?▼

Curriculum ordering for meta-learning is built by applying clustering and stability diagnostics to pre-computed Task2Vec embeddings, organizing episodic tasks into curriculum-friendly structures.

Can I diversify meta-batches using Task2Vec clustering?▼

Yes, diversity-aware meta-batch construction uses pre-computed Task2Vec embeddings and clustering to organize episodic tasks, ensuring varied task sampling across clusters during meta-training.

What is needed to extract deterministic Task2Vec embeddings for offline analysis?▼

Deterministic Task2Vec extraction requires a fixed probe and a registry to persist embeddings and diagnostics, running on PyTorch and NumPy to support offline task-space analysis.

Does Task2Vec curriculum generation work with PyTorch meta-learning workflows?▼

Yes, Task2Vec embeddings integrate into Phase 7 meta-learning workflows, using PyTorch and NumPy to provide task-space clustering and curriculum ordering for episodic training.

Why use diagonal Fisher information for task embeddings instead of full matrices?▼

Diagonal Fisher information provides a compact, deterministic approximation for Task2Vec embeddings, reducing computational overhead while still capturing task-space structure for curriculum ordering.

Related Skills

Explore More Data & Analytics Skills→
Pixel-Process-UGPixel-Process-UG
1

senior-prompt-engineer

Engineer prompts for large language models with structured outputs and evaluation.

Official
Advanced
NetRxnNetRxn
2

meta-learning-orchestrator

Route user requests to appropriate meta-learning component skills by pipeline phase.

Community
Advanced
stephendorstephendor
1

tda-representation-diagnostics

Diagnose embedding stability, PCA loadings, UMAP/t-SNE projections, and representation drift in TDL datasets.

Community
Advanced
snoodleboot-iosnoodleboot-io
2

dimensionality-reduction

Reduces dataset feature dimensions using PCA or t-SNE for ML preprocessing.

Official
Intermediate