skills.rest
Metta-AI
Metta-AIOFFICIAL
Explore all skills by this author →
Website𝕏@SoftmaxResearchGitHub

fabric-training-loop

Plan and execute Fabric training experiments with structured logs under ai_docs/.

Updated Aug 27, 2026
Search Tags:#logging#comparison#evidence#training#hypothesis#fabric#experiments
One-click install
npx skills add https://github.com/Metta-AI/cortical --skill fabric-training-loop
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: fabric-training-loop
Source: https://github.com/Metta-AI/cortical/tree/main/skills/fabric-training-loop
Command: npx skills add https://github.com/Metta-AI/cortical --skill fabric-training-loop

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill enables disciplined planning, execution, and evaluation of Fabric training experiments to generate reliable, evidence-based insights for model improvements.

Core Features & Use Cases

  • High-level Fabric declarations: Start experiments with clear architectural intents and minimal backend changes.
  • Controlled comparisons: Perform side-by-side ablations with consistent seeds, tasks, and logging to isolate effects.
  • Evidence-driven decisions: Maintain an active progress document and structured logs under ai_docs/ to justify decisions.

Quick Start

Define a training hypothesis, configure the Fabric stack, run a short controlled experiment, and log the results under ai_docs/.

Frequently Asked Questions about fabric-training-loop

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

FAQPage Schema
How do I log Fabric training experiments with empirical evidence?▼

To log Fabric training with empirical evidence, you define a training hypothesis, configure the Fabric stack, run a controlled experiment, and output a structured experiment log under ai_docs/ for traceability.

What is the best way to run controlled comparisons in Fabric training?▼

The best way to run controlled comparisons in Fabric training is performing side-by-side ablations using consistent seeds, tasks, and logging to isolate specific effects.

Can I use high-level Fabric APIs for supervised and continual-learning tasks?▼

Yes, you can use high-level Fabric APIs for supervised, sequence, continual-learning, and task-specific Fabric tasks to start experiments with clear architectural intents and minimal backend changes.

How do I plan a Fabric training experiment to test a specific hypothesis?▼

You plan a Fabric training experiment by defining a clear training hypothesis, configuring the Fabric stack, and maintaining an active progress document to justify evidence-driven decisions before backend changes.

Do I need evidence gates before making backend changes in Fabric training?▼

Yes, evidence gates are required before backend changes to ensure disciplined execution and evaluation of Fabric training experiments, generating reliable insights for model improvements.

What limitations exist when running side-by-side ablations in Fabric?▼

When running side-by-side ablations in Fabric, you must maintain consistent seeds, tasks, and logging to isolate effects; without this consistency, your evidence-driven decisions and comparisons may be unreliable.

Related Skills

Explore More Software Engineering Skills→
CUexterCUexter
2

simpo-training

Automate reference-free preference optimization for LLM alignment with PyTorch.

Community
Advanced
tbroadleytbroadley

read-inspect-eval

Extract samples, messages, events, and metrics from Inspect AI .eval logs.

Community
Intermediate
SimHackerSimHacker
49

debugging

Systematically debug issues by hypothesizing, testing, and tracking evidence.

Community
Advanced
xqinagxqinag

ablation-planner

Generate targeted ablation plans for machine learning method components.

Community
Advanced