---
title: thoshan_Flash
emoji: 🐨
colorFrom: yellow
colorTo: purple
sdk: gradio
sdk_version: 5.45.0
app_file: app.py
pinned: true
license: mit
---

# 💕 thoshan_Flash - Complete Offline Package

A conversational Large language model with a fun, flirty personality. This repository includes everything you need for offline development and training - no cloud dependencies required!

## 🎁 What's Included

- **Ready-to-run Gradio app** (`app.py`) - No modifications needed
- **Flirty tech dataset** (`flirt_dataset.jsonl`) - 20 Q&A pairs for training
- **Complete offline setup** - All dependencies clearly listed
- **Zero cloud dependencies** - Works entirely on your local machine
- **Copy-paste ready** - All commands tested and ready to use

## 📦 Download & Quick Start

### Option 1: Clone Everything

```bash
# Clone the complete repository
git clone https://huggingface.co/spaces/lingadevaruhp/thoshan_Flash_mini
cd thoshan_Flash_mini

# Install dependencies (one command, no conflicts)
pip install transformers torch gradio accelerate

# Run immediately
python app.py
```

### Option 2: Download Just the Dataset

Get the dataset file directly for your own projects:

- **Raw JSONL**: [flirt_dataset.jsonl](https://huggingface.co/spaces/lingadevaruhp/thoshan_Flash_mini/raw/main/flirt_dataset.jsonl) 
- **Direct download**: Right-click → "Save as" or use `wget`

```bash
# Download dataset only
wget https://huggingface.co/spaces/lingadevaruhp/thoshan_Flash_mini/raw/main/flirt_dataset.jsonl
```

## 🔥 Using the Dataset (Offline Training)

The `flirt_dataset.jsonl` file contains 20 flirty tech Q&A pairs in standard format:

```json
{"instruction": "Hey gorgeous, explain machine learning to me", "response": "Aww, you're so cute when you're curious! 💕 Think of machine learning like..."}
```

### Load in Python (Copy-Paste Ready)

```python
import json

# Load the dataset
with open('flirt_dataset.jsonl', 'r') as f:
    dataset = [json.loads(line) for line in f]

print(f"Loaded {len(dataset)} flirty tech examples!")
for item in dataset[:2]:  # Show first 2
    print(f"Q: {item['instruction']}")
    print(f"A: {item['response'][:100]}...\n")
```

### Use with Popular Training Libraries

```python
# With Hugging Face datasets
from datasets import load_dataset
dataset = load_dataset('json', data_files='flirt_dataset.jsonl')

# With pandas
import pandas as pd
df = pd.read_json('flirt_dataset.jsonl', lines=True)

# Direct training format
training_data = []
with open('flirt_dataset.jsonl', 'r') as f:
    for line in f:
        data = json.loads(line)
        training_data.append({
            'input': data['instruction'],
            'output': data['response']
        })
```

## 💻 Model Information

- **Base Model**: `thoshan_Flash`
- **No fine-tuning required** - Works out of the box
- **Public model** - No API keys or special access needed
- **Lightweight** - Runs on consumer GPUs (4GB+ VRAM recommended)
- **Fast inference** - Optimized for real-time chat

## 🚀 Advanced Usage

### Custom Training with Your Data

```python
# Combine with your own data
my_data = []
with open('flirt_dataset.jsonl', 'r') as f:
    my_data.extend([json.loads(line) for line in f])

# Add your own examples
my_data.append({
    "instruction": "Your custom question",
    "response": "Your custom flirty response"
})

# Save combined dataset
with open('my_custom_dataset.jsonl', 'w') as f:
    for item in my_data:
        f.write(json.dumps(item) + '\n')
```

### Fine-tune Locally (Optional)

```python
# Example fine-tuning setup (requires additional setup)
from transformers import AutoTokenizer, AutoModelForCausalLM
from transformers import TrainingArguments, Trainer

model_name = "thoshan_Flash"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Your fine-tuning code here...
```

## 🛠️ Troubleshooting

### Common Issues & Solutions

**"Model not found"** → Check internet connection for first download

```bash
pip install --upgrade transformers torch
```

**"Out of memory"** → Reduce batch size or use CPU mode

```python
# In app.py, add:
device = "cpu"  # Force CPU usage
```

**"Gradio won't start"** → Update Gradio

```bash
pip install --upgrade gradio
```

**"Dataset won't load"** → Verify file format

```bash
# Check if file is valid JSON lines
python -c "import json; [json.loads(line) for line in open('flirt_dataset.jsonl')]; print('Valid!')"
```

## 📝 Dataset Details

**Content**: 20 technology-themed Q&A pairs with flirty, fun responses
**Topics covered**: Machine learning, web development, databases, DevOps, security, and more
**Format**: Standard JSONL (one JSON object per line)
**Size**: ~10.4KB - Perfect for quick experiments
**Style**: Educational but playful - explains complex tech concepts with personality

## 🎯 Perfect For

- **Learning AI development** - Complete, working example
- **Chatbot experimentation** - Ready-made personality dataset
- **Offline development** - No API dependencies
- **Educational projects** - Fun way to learn tech concepts
- **Fine-tuning practice** - Small, manageable dataset

## 🔄 Updates & Versions

- **Latest**: Added complete offline dataset (flirt_dataset.jsonl)
- **Improved**: Zero-dependency local setup
- **Fixed**: All dependency conflicts resolved
- **Added**: Copy-paste ready code examples

---
**Ready to get flirty with AI? Download, run, and start chatting! 💕**