初始化项目,由ModelHub XC社区提供模型
Model: Phora68/spirit-concordance-llama3.1-8b Source: Original Platform
This commit is contained in:
51
README.md
Normal file
51
README.md
Normal file
@@ -0,0 +1,51 @@
|
||||
---
|
||||
language: en
|
||||
license: mit
|
||||
base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
|
||||
tags:
|
||||
- bible
|
||||
- theology
|
||||
- qlora
|
||||
- unsloth
|
||||
- llama-3.1
|
||||
- spurgeon
|
||||
- wesley
|
||||
- wilkerson
|
||||
- bible-study
|
||||
pipeline_tag: text-generation
|
||||
---
|
||||
|
||||
# Spirit Concordance — 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 · BF16 · A100
|
||||
- LoRA rank: 32 · Epochs: 2
|
||||
- Best loss: 0.3078 · Final loss: 0.5696
|
||||
- Training time: 2 hrs 42 min
|
||||
|
||||
## Files
|
||||
- lora_adapters/ — LoRA weights for continued training
|
||||
- gguf/ — Q4_K_M quantized model for LM Studio / Ollama
|
||||
- Merged 16-bit weights in repo root
|
||||
|
||||
## System prompt
|
||||
You are Spirit Concordance — 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
|
||||
Reference in New Issue
Block a user