README.md

---
base_model: Spestly/Athena-1-3B
tags:
- text-generation-inference
- transformers
- unsloth
- qwen2
- trl
- llama-cpp
- gguf-my-repo
license: other
license_name: qwen-research
license_link: https://huggingface.co/Spestly/Athena-1-3B/blob/main/LICENSE
language:
- en
---

# Triangle104/Athena-1-3B-Q5_K_S-GGUF
This model was converted to GGUF format from [`Spestly/Athena-1-3B`](https://huggingface.co/Spestly/Athena-1-3B) using llama.cpp via the ggml.ai's [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space.
Refer to the [original model card](https://huggingface.co/Spestly/Athena-1-3B) for more details on the model.

---
Model details:
-
Athena-1 3B is a fine-tuned, instruction-following large language model derived from Qwen/Qwen2.5-3B-Instruct. It is designed to provide efficient, high-quality text generation while maintaining a compact size. Athena 3B is optimized for lightweight applications, conversational AI, and structured data tasks, making it ideal for real-world use cases where performance and resource efficiency are critical.

Key Features
-
⚡ Lightweight and Efficient
-
Compact Size: At just 3.09 billion parameters, Athena-1 3B offers excellent performance with reduced computational requirements.

Instruction Following: Fine-tuned for precise and reliable adherence to user prompts.

Coding and Mathematics: Proficient in solving coding challenges and handling mathematical tasks.

📖 Long-Context Understanding
-
Context Length: Supports up to 32,768 tokens, enabling the processing of moderately lengthy documents or conversations.

Token Generation: Can generate up to 8K tokens of output.

🌍 Multilingual Support
-
Supports 29+ languages, including:
English, Chinese, French, Spanish, Portuguese, German, Italian, Russian
Japanese, Korean, Vietnamese, Thai, Arabic, and more.

📊 Structured Data & Outputs
-
Structured Data Interpretation: Processes structured formats like tables and JSON.

Structured Output Generation: Generates well-formatted outputs, including JSON and other structured formats.

Details
-
Base Model: Qwen/Qwen2.5-3B-Instruct

Architecture: Transformers with RoPE, SwiGLU, RMSNorm, Attention QKV bias, and tied word embeddings.

Parameters: 3.09B total (2.77B non-embedding).

Layers: 36

Attention Heads: 16 for Q, 2 for KV.

Context Length: Up to 32,768 tokens.

Applications
-
Athena 3B is designed for a variety of real-world applications:

Conversational AI: Build fast, responsive, and lightweight chatbots.

Code Generation: Generate, debug, or explain code snippets.

Mathematical Problem Solving: Assist with calculations and reasoning.

Document Processing: Summarize and analyze moderately large documents.

Multilingual Applications: Support for global use cases with diverse language requirements.

Structured Data: Process and generate structured data, such as tables and JSON.

Quickstart
-
Here’s how you can use Athena 3B for quick text generation:

---
# Use a pipeline as a high-level helper
from transformers import pipeline

messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe = pipeline("text-generation", model="Spestly/Athena-1-3B")
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Spestly/Athena-1-3B")
model = AutoModelForCausalLM.from_pretrained("Spestly/Athena-1-3B")

---
## Use with llama.cpp
Install llama.cpp through brew (works on Mac and Linux)

```bash
brew install llama.cpp

```
Invoke the llama.cpp server or the CLI.

### CLI:
```bash
llama-cli --hf-repo Triangle104/Athena-1-3B-Q5_K_S-GGUF --hf-file athena-1-3b-q5_k_s.gguf -p "The meaning to life and the universe is"
```

### Server:
```bash
llama-server --hf-repo Triangle104/Athena-1-3B-Q5_K_S-GGUF --hf-file athena-1-3b-q5_k_s.gguf -c 2048
```

Note: You can also use this checkpoint directly through the [usage steps](https://github.com/ggerganov/llama.cpp?tab=readme-ov-file#usage) listed in the Llama.cpp repo as well.

Step 1: Clone llama.cpp from GitHub.
```
git clone https://github.com/ggerganov/llama.cpp
```

Step 2: Move into the llama.cpp folder and build it with `LLAMA_CURL=1` flag along with other hardware-specific flags (for ex: LLAMA_CUDA=1 for Nvidia GPUs on Linux).
```
cd llama.cpp && LLAMA_CURL=1 make
```

Step 3: Run inference through the main binary.
```
./llama-cli --hf-repo Triangle104/Athena-1-3B-Q5_K_S-GGUF --hf-file athena-1-3b-q5_k_s.gguf -p "The meaning to life and the universe is"
```
or 
```
./llama-server --hf-repo Triangle104/Athena-1-3B-Q5_K_S-GGUF --hf-file athena-1-3b-q5_k_s.gguf -c 2048
```