README.md

metadata

license: apache-2.0
base_model: unsloth/llama-3-8b-bnb-4bit
tags:
  - llama-3
  - lora
  - multi-persona
  - nlp-tutor
  - emotional-mentor

Llama-3-8B Multi-Persona (Ray's Edition)

This model is a fine-tuned version of Meta's Llama-3-8B using Unsloth LoRA. It is specialized in three distinct personas:

1. NLP Assistant: Provides structured technical knowledge about NLP and Deep Learning.
2. Emotional Mentor: Offers rational and gentle advice for relationship issues (style inspired by PTT Boy-Girl).
3. Extreme Simp (舔狗): Responds with extreme humbleness and dedication.

💻 Hardware Environment

• GPU: NVIDIA GeForce RTX 3080 Ti (12GB VRAM)
• Power Usage: Peak 391W
• Framework: Unsloth (4-bit quantization)

🧪 Training Details

• Epochs: 3 (Version 2.0)
• Batch Size: 2 with Gradient Accumulation 4
• Learning Rate: 1e-4 (Cosine Decay)
• Dataset: Mixed heterogeneous corpora from PTT and academic NLP samples.

🚀 How to use

from unsloth import FastLanguageModel
model, tokenizer = FastLanguageModel.from_pretrained("你的帳號/Llama-3-8B-Multi-Persona-Ray")