SmolLM-135M-Tarot-Zodiac-Reasoning

By 11-47

A fine-tuned 135M parameter model specialized in Tarot interpretation, Zodiac insights, and mystical reasoning. Built on top of HuggingFaceTB/SmolLM-135M-Instruct using Supervised Fine-Tuning (SFT) with TRL.

This tiny but mighty model blends modern LLM capabilities with esoteric knowledge—perfect for lightweight tarot readings, astrological guidance, symbolic reasoning, and creative spiritual conversations on edge devices or low-resource setups.

Model Details

• Base Model: HuggingFaceTB/SmolLM-135M-Instruct
• Parameters: 135M (super efficient!)
• Tensor Type: F32 (Safetensors)
• Training Method: SFT (Supervised Fine-Tuning) via TRL
• Datasets:
  • 11-47/The_Prophet_Tarot_Zodiac_50K (main dataset)
  • 11-47/The_Prophet_Tarot_Zodiac_5K
  • 11-47/The_Prophet_Tarot_Zodiac_5K_seed (seed / validation)

The training focuses on conversational tarot spreads, zodiac interpretations, symbolic reasoning, and blending mystical traditions with logical explanations.

Quick Start

from transformers import pipeline

generator = pipeline(
    "text-generation",
    model="11-47/SmolLM-135M-Tarot-Zodiac-Reasoning",
    device="cuda"  # or "cpu"
)

question = "What does The Tower card mean for someone with a Scorpio rising going through a major life transition?"
output = generator(
    [{"role": "user", "content": question}],
    max_new_tokens=256,
    return_full_text=False
)[0]

print(output["generated_text"])
Intended Uses
	•	Tarot Readings: Generate insightful interpretations of cards and spreads (single card, 3-card, Celtic Cross, etc.).
	•	Zodiac & Astrology: Personalized horoscopes, compatibility analysis, and planetary transits.
	•	Mystical Reasoning: Symbolic interpretation, dream analysis, spiritual advice with logical structure.
	•	Lightweight Deployment: Runs great on phones, laptops, or embedded devices.
	•	Fun / Educational spiritual chatbots.
Limitations
	•	Very small model (135M) → Can be less coherent on complex multi-card spreads or highly nuanced queries compared to larger models.
	•	Knowledge is synthetic / training-data driven — treat outputs as inspirational rather than definitive divination.
	•	May occasionally hallucinate card meanings or astrological facts.
	•	Not a replacement for professional tarot/astrology consultation.
Training Procedure
	•	Framework: Hugging Face TRL (Transformers Reinforcement Learning library)
	•	Method: SFT on assistant responses in chat format.
	•	Framework Versions:
	◦	TRL: 1.8.0
	◦	Transformers: 5.13.1
	◦	PyTorch: 2.10.0+cu128
	◦	Datasets: 5.0.0
Citation
@misc{smollm-tarot-zodiac-reasoning,
  author = {11-47},
  title = {SmolLM-135M-Tarot-Zodiac-Reasoning},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/11-47/SmolLM-135M-Tarot-Zodiac-Reasoning}
}

# Cite TRL
@software{vonwerra2020trl,
  title   = {{TRL: Transformers Reinforcement Learning}},
  author  = {von Werra, Leandro and others},
  url     = {https://github.com/huggingface/trl},
  year    = {2020}
}