DisciplineAI / README.md
omk4r's picture
Update README.md
76e634b verified
metadata
base_model: mistralai/Mistral-7B-v0.1
library_name: peft
pipeline_tag: text-generation
tags:
  - discipline
  - self-help
  - lora
  - qlora
  - transformers
  - Mistral
  - AI
license: apache-2.0
language:
  - en

Open In Colab

DisceplineAI

A LoRA‑tuned Mistral‑7B adapter that thinks and advises like four self‑help classics. Fine‑tuned on The 48 Laws of Power, The Way of the Superior Man, Psycho‑Cybernetics, and How to Win Friends and Influence People, DisceplineAI delivers stoic, actionable guidance in its own words.


Model Details

  • Developed by: (omk4rr)
  • Shared on: Hugging Face as omk4r/DisceplineAI
  • Model type: LoRA adapter on Mistral‑7B (7 B parameters, decoder‑only causal LM)
  • Language: English
  • License: Apache 2.0
  • Fine‑tuned from: mistralai/Mistral-7B-v0.1 via 4‑bit QLoRA and PEFT

Uses

Direct Use

  • Ask anything related to discipline, habit‑formation, influence, and personal development.
  • Ideal for chatbots, productivity assistants, or self‑help tools.

Out‑of‑Scope

  • Medical, legal, or financial advice.
  • Highly specialized technical or domain‑specific questions beyond personal growth.

Bias, Risks, and Limitations

  • Bias Sources: Draws on content from four self‑help books that reflect their authors’ worldviews—may overemphasize power dynamics, gender roles, or confidence.
  • Limitations:
    • Not a substitute for professional guidance.
    • May produce over‑generalized or overly confident advice.
  • Mitigations:
    • Users should apply critical judgment.
    • Combine AI suggestions with domain expertise.

How to Get Started

Install dependencies and run the CLI:

pip install -r requirements.txt
python inference.py "How can I stop procrastinating?"