This model has been optimized and converted to the .venus binary format for the EdgeFlow runtime. It is designed to run efficiently on any hardware (CPU, consumer GPU, Edge devices) using OpenCL.

Note: This is the v2 version with improved conversion optimizations compared to the original venus conversion.

Model Details

Property	Value
Original Model	Qwen/Qwen2.5-0.5B
Original Creator	Qwen Team (Alibaba)
Converted By	EdgeFlow
Architecture	Qwen2ForCausalLM
Parameters	0.5B (494M)
Context Length	32,768 tokens
License	Apache 2.0
Format	Venus Binary (.venus)
Quantization	Q8_0
Version	v2 (improved)

About the Original Model

Qwen2.5-0.5B is part of the Qwen2.5 series of large language models by Alibaba. Key features include:

Enhanced Knowledge: Improved capabilities in coding and mathematics
Long Context: Supports up to 128K tokens context, generates up to 8K tokens
Multilingual: Supports 29+ languages including Chinese, English, French, Spanish, and more
Efficient Architecture: Uses RoPE, SwiGLU, RMSNorm, and GQA (14 Q heads, 2 KV heads)

What's New in v2

Improved model conversion with better weight optimization
Enhanced inference performance
Better memory efficiency

Quick Start

You can run this model locally using the EdgeFlow CLI or integrate it into your app with the Venus SDK.

Method 1: EdgeFlow CLI

bash

# Install the runtime
pip install edgeflow

# Run instantly (auto-downloads from this repo)
edgeflow run edgeflow-ai/qwen2.5-0.5b-v2-venus --prompt "Hello world"

Method 2: Python SDK

python

from edgeflow import VenusModel

# Load the model
model = VenusModel.from_pretrained("edgeflow-ai/qwen2.5-0.5b-v2-venus")

# Generate text
response = model.generate("Explain quantum computing in simple terms")
print(response)

File Structure

qwen2.5-0.5b-v2-venus/
├── model.venus          # The optimized Venus binary model (v2)
├── tokenizer.json       # Tokenizer configuration
├── tokenizer_config.json
├── vocab.json
├── merges.txt
├── special_tokens_map.json
├── added_tokens.json
├── chat_template.jinja
└── README.md

Performance

The Venus format enables significant performance improvements:

Cross-platform: Runs on CPU, GPU, and edge devices via OpenCL
Optimized Memory: Reduced memory footprint with Q8_0 quantization
Fast Inference: Optimized binary format for quick model loading

Related Models

edgeflow-ai/qwen2.5-0.5b-venus - Original v1 conversion

Citation

If you use this model, please cite both the original work and EdgeFlow:

bibtex

@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}
}

License

This model inherits the Apache 2.0 license from the original Qwen2.5-0.5B model. See LICENSE for details.

Converted to Venus format by EdgeFlow