---
license: apache-2.0
library_name: transformers
pipeline_tag: text-generation
task: text-generation
inference:
  parameters:
    max_new_tokens: 400
    temperature: 0.7
    top_p: 0.9
    do_sample: true
language:
- en
tags:
- llama-3.1
- personal-assistant
- book-advisor
- merged-lora
base_model: meta-llama/Llama-3.1-8B-Instruct
---

# 📚 Jacob's Personal Book Advisor (Merged Model)

This is a **merged model** combining Llama-3.1-8B-Instruct with a LoRA adapter trained on Jacob's personal book library.

**✅ Ready for Inference API** - This merged model works directly with HuggingFace Inference API.

## Features
- Personalized book recommendations from Jacob's library
- Content questions and summaries
- Reading advice based on actual book collection

## Usage

### With Inference API
```python
from huggingface_hub import InferenceClient

client = InferenceClient(model="jacobpmeyer/book-advisor-merged")
response = client.text_generation(
    "### Instruction:\nRecommend a science fiction book\n\n### Response:\n",
    max_new_tokens=300,
    temperature=0.7
)
print(response)
```

### Direct Usage
```python
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("jacobpmeyer/book-advisor-merged")
model = AutoModelForCausalLM.from_pretrained("jacobpmeyer/book-advisor-merged")

prompt = "### Instruction:\nWhat's a good book for vacation reading?\n\n### Response:\n"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=200, temperature=0.7)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
```

## Training Details
- **Base Model**: meta-llama/Llama-3.1-8B-Instruct
- **Method**: LoRA fine-tuning + merging
- **Training Data**: Personal epub book collection
- **Format**: Instruction-following (Alpaca style)

## Model Performance
This merged model combines the instruction-following capabilities of Llama-3.1-8B-Instruct with personalized knowledge from Jacob's book library, providing relevant and personalized book recommendations and insights.