Is this skill compatible with Hugging Face?

While it is a management framework, it includes best practices for model naming and tagging compatible with Hugging Face Hub and Ollama deployment patterns.

Can I compare different training runs?

Yes, the skill provides features to compare configuration changes (like LoRA rank or epochs) and performance outcomes (like Macro-F1 and Accuracy) between versions.

Does it handle model evaluation data?

Yes, it tracks evaluation results in lineage.yaml, including class-specific precision, recall, and F1 scores to identify specific areas for improvement.

What versioning strategies are supported?

It supports Semantic (v1.0), Date (2025-01-01), and Hybrid (v1-20250101) versioning to fit various deployment and development styles.

How does this skill organize my projects?

It implements a standardized directory structure where each task (e.g., sentiment analysis) is a self-contained unit containing its own task.yaml, data source, and version history.

LLM Task Manager

Name: LLM Task Manager
Author: p988744

byp988744

0•

数据科学与机器学习

Orchestrates multi-task LLM training workflows with comprehensive version tracking and performance comparison tools.

The Task Manager skill provides a standardized framework for managing complex Natural Language Processing (NLP) and LLM fine-tuning projects. It organizes individual training tasks into self-contained directories, tracking the full lineage of model versions—including data snapshots, training configurations like LoRA parameters, and detailed evaluation metrics. This skill is essential for developers who need to manage multiple experiments, compare performance across iterations (using metrics like Macro-F1), and maintain a clear history of model evolution from configuration to deployment.

主要功能

01Side-by-side performance comparison of different training versions

020 GitHub stars

03Detailed version lineage tracking for data, config, and results

04Standardized multi-task directory structure for isolated experiments

05Automated task lifecycle management from creation to deployment

06Support for Semantic, Date-based, and Hybrid versioning strategies

使用场景

01Managing multiple fine-tuning experiments for sentiment analysis and NER

02Maintaining a historical record of model checkpoints and training data snapshots

03Comparing the impact of different LoRA hyperparameters on model accuracy

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add p988744/nlp-skills task-manager

For use in Claude.ai and ChatGPT

Download Skill