Modelpublic

ronniealfaro/mythos-prime

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face

1likes24downloads

Model Card

🧿 Mythos Prime — A LoRA-Fused GGUF Model

Author: Ronnie Alfaro (@ronniealfaro) Base model: NousResearch/Hermes-2-Pro-Mistral-7B Fused using PEFT + `merge_and_unload`, converted with llama.cpp.

📜 Model Description

Mythos Prime is a fine-tuned fusion of the powerful Hermes 2 Pro (Mistral 7B) base model with a LoRA adapter trained on mythological, symbolic, and philosophical prompts across global traditions.

This model blends Hermes’ natural coherence with a deep narrative capacity focused on:

—🧠 Storytelling with archetypal structure
—📚 Myth interpretation and symbolic exploration
—💬 Philosophical discourse generation
—🎮 Lore-rich worldbuilding for fiction & games
—✍️ Creative writing inspired by ancient motifs

🛠️ Technical Specs

Feature	Value
Base model	`NousResearch/Hermes-2-Pro-Mistral-7B`
Adapter type	LoRA (`peft_type: LORA`)
Merge method	`merge_and_unload()` via PEFT
Output formats	`.gguf (f16)` + `.gguf (Q4_K_M)`
Tokenizer	Hermes original tokenizer
Context length	4096 tokens

🔍 Files Included

Filename	Description
`mythos.gguf`	Fused full-precision model (f16)
`mythos.Q4_K_M.gguf`	Quantized version (Q4KM)
`tokenizer.model/json`	Tokenizer for Hermes
`config.json`	Base model config

🧪 Intended Use

Mythos Prime was created for use in:

—Writing and storytelling based on mythic structures
—Narrative AI assistants with symbolic insight
—Philosophical or metaphysical exploration
—Games and RPG tools for generating lore, pantheons, or quests
—Worldbuilding that draws from multiple cultural traditions

⚠️ Limitations

—May produce abstract, symbolic, or obscure outputs — this is by design
—Not tuned for factual accuracy or real-time knowledge
—May reflect cultural biases from training data

🗿 Credits & Lore

Fused by Ronnie Alfaro, after wrestling with Axolotl, its examples.yaml minion, the import-loop cyclops, and the blue screen oracle of pain.

This model is a techno-mystical artifact of patience, recursion, and narrative passion.

🔮 License

Follows the base model license: Apache 2.0