CodeQwen1.5-7B-AWQ
| Property | Value |
|---|---|
| Parameter Count | 1.8B |
| License | tongyi-qianwen-research |
| Paper | Research Paper |
| Quantization | AWQ (4-bit precision) |
| Context Length | 64K tokens |
What is CodeQwen1.5-7B-AWQ?
CodeQwen1.5-7B-AWQ is a specialized, quantized version of the base CodeQwen1.5 model, specifically designed for code generation tasks. It's a transformer-based decoder-only language model that has been trained on 3 trillion tokens of code data, implementing group query attention (GQA) for efficient inference.
Implementation Details
This model represents an optimized implementation using AWQ quantization, requiring transformers>=4.37.0 for proper functionality. It maintains the architectural benefits of the base model while reducing the computational requirements through 4-bit precision.
- Transformer-based decoder-only architecture
- AWQ quantization for efficient deployment
- Group Query Attention (GQA) implementation
- 64K token context length support
Core Capabilities
- Supports 92 different programming languages
- Specialized in code generation and understanding
- Excellent performance in text-to-SQL tasks
- Advanced bug fixing capabilities
- Long context understanding and generation
Frequently Asked Questions
Q: What makes this model unique?
This model stands out due to its specialized focus on code generation, extensive language support (92 programming languages), and efficient implementation through AWQ quantization while maintaining high performance.
Q: What are the recommended use cases?
The model is best suited for code infilling, code generation, and finetuning tasks. It's important to note that this is the base model, not the chat model, so proper stopping criteria should be implemented for generation tasks.
