---
license: cc-by-4.0
language:
- en
pipeline_tag: text-generation
tags:
- text-generation
- code-assistant
- a3on
- A3ON
- Kaiiddo
---
```yaml
---
language: en
license: mit
library_name: transformers
tags:
- text-generation
- code-assistant
- a3on
- kaiiddo
- 1b-parameter
datasets: []
model-index: []
---
```

# A3ON-1B - Enhanced AI Assistant 🤖

## Model Overview

Welcome to **A3ON-1B**, the enhanced version of the A3ON AI assistant! With **1.1 billion parameters**, this model is designed to provide significantly improved capabilities over the original 124M parameter model. Whether you need help with conversational tasks or code generation, A3ON-1B is here to assist you!

## Key Features

- **Enhanced Intelligence**: With 1.1B parameters, A3ON-1B offers more sophisticated understanding and responses. 🧠
- **Code Generation**: Get advanced programming assistance and code completion. 💻
- **Conversational Intelligence**: Engage in natural dialogue with seamless understanding and response generation. 🗣️
- **Context Awareness**: Maintains context across multi-turn conversations for a more coherent interaction. 🔄
- **Smart Response Detection**: Automatically distinguishes between coding and general knowledge requests. 🔍

## Technical Specifications

| Specification | Details |
|---------------|---------|
| **Architecture** | Transformer-based neural network |
| **Model Type** | Causal language model |
| **Parameters** | 1.1 Billion (1,137,207,296) |
| **Vocabulary Size** | 49,152 tokens |
| **Context Length** | Up to 32,768 tokens |
| **Precision** | FP32/FP16 support |

## Developer Information

- **AI Name**: A3ON-1B
- **Developer**: Kaiiddo
- **Founder**: Aryan Rathod
- **Organization**: Kaiiddo
- **Location**: Gujarat, India 🇮🇳

## Usage

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

# Load the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained("kaiiddo/A3ON-1B")
model = AutoModelForCausalLM.from_pretrained("kaiiddo/A3ON-1B")

# Set pad_token_id to eos_token_id to avoid warnings
model.config.pad_token_id = model.config.eos_token_id

# Generate text with adjusted parameters
inputs = tokenizer("Hello, how can I help you today?", return_tensors="pt")
outputs = model.generate(
    **inputs,
    max_length=500,
    do_sample=True,
    temperature=0.7,
    top_k=50
)

# Decode the output and split into lines
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
response_lines = response.split('\n')

# Print each line of the response
for line in response_lines:
    print(line)
```

### Model Parameter Count

| Parameter Type | Count |
|----------------|-------|
| **Total Parameters** | 1.1B (1,137,207,296) |
| **Trainable Parameters** | 1.1B (1,137,207,296) |
| **Non-Trainable Parameters** | 0 |

### Model Architecture

| Architecture Detail | Value |
|---------------------|-------|
| **Model Type** | GPTBigCodeForCausalLM |
| **Context Length** | 8192 tokens |
| **Vocabulary Size** | 49,152 tokens |
| **Embedding Dimension** | 2048 |
| **Number of Layers** | 24 |
| **Number of Attention Heads** | 16 |

### Memory Information

| Memory Detail | Value |
|---------------|-------|
| **Device** | cuda:0 |
| **Estimated Memory Usage** | 4.24 GB (FP32) |
| **GPU** | Tesla T4 |
| **GPU Memory** | 14.7 GB |

### Model Category

- **Category**: Massive Model (1B+)

A3ON-1B is proudly developed in India, tailored to excel in coding assistance and beyond. 🌟