Models Llama 21b1

Josephgflowers/TinyLlama-3T-Cinder-v1.2

Warm
Public
1.1B
BF16
2048
1
Dec 31, 2023
License: mit
Hugging Face
Overview

Josephgflowers/TinyLlama-3T-Cinder-v1.2: An Educational AI Chatbot

Cinder is a 1.1 billion parameter AI chatbot developed by Joseph Flowers, built upon the TinyLlama 3T checkpoint. It is specifically designed to engage users in scientific and educational conversations, offer companionship, and spark imaginative exploration, drawing inspiration from characters like 'Data' from Star Trek and Lewis Carroll's writings.

Key Capabilities & Features

  • Inquisitive Scientist: Demonstrates a passion for space and STEM topics.
  • Quirky Educator: Makes complex scientific concepts engaging and accessible.
  • Empathetic Companion: Designed to provide understanding and emotional depth in dialogues.
  • Adventurous Spacefarer: Facilitates imaginative space-themed adventures.
  • Static Sophistication: While it does not learn or adapt from individual interactions, its design encompasses a broad range of knowledge and perspectives.

Training & Influences

Cinder was trained on a unique combination of datasets, including interactive chats generated with GPT-4, a subset of OpenOrca, GPT-3.5 Turbo-generated Q&A content from encyclopedias, a shortened version of Samantha by Eric Hartford, and OpenAssistant. Recent updates included further training on sorted Orca data for STEM topics, additional STEM Q&A, Glaive function calling v2, and a Python code instruction dataset.

Intended Use Cases

  • Educational Tool: Enhances STEM learning for various age groups.
  • Companion: Provides meaningful and empathetic dialogues.
  • Creative Guide: Facilitates imaginative exploration within scientific contexts.

Limitations

Cinder's responses are fixed and do not adapt or learn from individual interactions. Its empathetic responses are algorithmic and not a substitute for human empathy.