Triangle104/Athena-1-14B-Q4_K_M-GGUF

This model was converted to GGUF format from Spestly/Athena-1-14B using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model.

---

Model details:

Athena 1 is a state-of-the-art language model fine-tuned from Qwen/Qwen2.5-14B-Instruct. Designed to excel in instruction-following tasks, Athena 1 delivers advanced capabilities in text generation, coding, mathematics, and long-context understanding. It is optimized for a wide variety of use cases, including conversational AI, structured data interpretation, and multilingual applications. It outperforms Ava 1.5 in many aspects making Athena-1 the superior model.

    Key Features






    


    🚀 Enhanced Capabilities

Instruction Following: Athena 1 has been fine-tuned for superior adherence to user prompts, making it ideal for chatbots, virtual assistants, and guided workflows. Coding and Mathematics: Specialized fine-tuning enhances coding problem-solving and mathematical reasoning. Long-Context Understanding: Handles input contexts up to 128K tokens and generates up to 8K tokens.

    🌐 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: Understands and processes structured formats like tables and JSON. Structured Output Generation: Generates well-formatted outputs, including JSON, XML, and other structured formats.

    Model Details

Base Model: Qwen/Qwen2.5-14B-Instruct Architecture: Transformers with RoPE, SwiGLU, RMSNorm, and Attention QKV bias. Parameters: 14.7B total (13.1B non-embedding). Layers: 48 Attention Heads: 40 for Q, 8 for KV. Context Length: Up to 131,072 tokens.

    Applications

Athena 1 is designed for a wide range of use cases:

Conversational AI and chatbots. Code generation, debugging, and explanation. Mathematical problem-solving. Large-document summarization and analysis. Multilingual text generation and translation. Structured data processing (e.g., tables, JSON).

    Quickstart

Below is an example of how to use Athena 1 for text generation:

huggingface-cli login

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-1-14B") pipe(messages)

Load model directly

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Spestly/Athena-1-14B") model = AutoModelForCausalLM.from_pretrained("Spestly/Athena-1-14B")

    Performance

Athena 1 has been optimized for efficiency and performance on modern GPUs. For detailed evaluation metrics (e.g., throughput, accuracy, and memory requirements), refer to the Qwen2.5 performance benchmarks.

    Requirements

To use Athena 1, ensure the following:

Python >= 3.8 Transformers >= 4.37.0 (to support Qwen models) PyTorch >= 2.0 GPU with BF16 support for optimal performance.

    Citation

If you use Athena 1 in your research or projects, please cite its base model Qwen2.5 as follows:

@misc{qwen2.5, title = {Qwen2.5: A Party of Foundation Models}, url = {https://qwenlm.github.io/blog/qwen2.5/}, author = {Qwen Team}, month = {September}, year = {2024} }

---

Use with llama.cpp

Install llama.cpp through brew (works on Mac and Linux)

brew install llama.cpp

Invoke the llama.cpp server or the CLI.

CLI:

llama-cli --hf-repo Triangle104/Athena-1-14B-Q4_K_M-GGUF --hf-file athena-1-14b-q4_k_m.gguf -p "The meaning to life and the universe is"

Server:

llama-server --hf-repo Triangle104/Athena-1-14B-Q4_K_M-GGUF --hf-file athena-1-14b-q4_k_m.gguf -c 2048

Note: You can also use this checkpoint directly through the usage steps 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-1-14B-Q4_K_M-GGUF --hf-file athena-1-14b-q4_k_m.gguf -p "The meaning to life and the universe is"

or

./llama-server --hf-repo Triangle104/Athena-1-14B-Q4_K_M-GGUF --hf-file athena-1-14b-q4_k_m.gguf -c 2048