Spirit Concordance โ€” Llama 3.1 8B

Fine-tuned on the King James Bible, Greek NT (NA28 + Strong's), Hebrew OT (BHS + Strong's), Spurgeon, Wesley, and Wilkerson.

Training

  • Base: meta-llama/Meta-Llama-3.1-8B-Instruct
  • Method: QLoRA (4-bit) with Unsloth ยท BF16 ยท A100
  • LoRA rank: 32 ยท Epochs: 2
  • Best loss: 0.3078 ยท Final loss: 0.5696
  • Training time: 2 hrs 42 min

Files

  • lora_adapters/ โ€” LoRA weights for continued training
  • gguf/ โ€” Q4_K_M quantized model for LM Studio / Ollama
  • Merged 16-bit weights in repo root

System prompt

You are Spirit Concordance โ€” a Bible study companion trained on the KJV, Greek NT (NA28 Strong's), Hebrew OT (BHS Strong's). You draw on Spurgeon (sovereign grace), Wesley (holiness), and Wilkerson (prophetic urgency). Quote KJV accurately. Give Strong's numbers with transliteration and definition.

LM Studio settings

  • Temperature: 0.3
  • Stop strings: <|eot_id|> <|end_of_text|>
  • Max tokens: 500

Capabilities

  • Quotes KJV verses word for word
  • Greek word studies with Strong's G numbers
  • Hebrew word studies with Strong's H numbers
  • Topical concordance searches
  • Spurgeon, Wesley and Wilkerson theological voices
Downloads last month
109
Safetensors
Model size
8B params
Tensor type
BF16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for Phora68/spirit-concordance-llama3.1-8b

Quantized
(615)
this model