Etherll/Qwen2.5-CodeFIM-1.5B-v2

TEXT GENERATIONConcurrency Cost:1Model Size:1.5BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Nov 11, 2024Architecture:Transformer0.0K Cold

Etherll/Qwen2.5-CodeFIM-1.5B-v2 is a 1.5 billion parameter model, fine-tuned from Qwen/Qwen2.5-Coder-1.5B, specifically designed for Fill-in-the-Middle (FIM) code generation. It leverages a specialized dataset for FIM tasks, making it highly effective for code completion and infilling scenarios. With a context length of 32768 tokens, this model excels at generating code snippets within existing code structures.

Loading preview...

Etherll/Qwen2.5-CodeFIM-1.5B-v2: Code Fill-in-the-Middle Specialist

This model is a specialized 1.5 billion parameter language model, fine-tuned by Etherll from the Qwen/Qwen2.5-Coder-1.5B base. Its primary function is Fill-in-the-Middle (FIM) code generation, making it highly proficient at completing code within existing structures rather than generating entire code blocks from scratch.

Key Capabilities

  • Code Infilling: Excels at generating missing code segments based on surrounding context.
  • Specialized Training: Fine-tuned on the Etherll/code-fim-v2 dataset, specifically optimized for FIM tasks.
  • Integration with Development Tools: Designed for use with tools like Continue for enhanced developer workflows.
  • Context Length: Supports a substantial context window of 32768 tokens, allowing it to handle larger code files and maintain contextual awareness.

Usage Format

To utilize the model effectively for FIM, it requires a specific input format:

<|file_name|>{{{filename}}}<|fim_prefix|>{{{prefix}}}<|fim_suffix|>{{{suffix}}}<|fim_middle|>

This structure guides the model to understand where the prefix, suffix, and filename information are located for accurate infilling.

Good For

  • Code Autocompletion: Ideal for IDEs and code editors requiring intelligent code suggestions and completions.
  • Developer Productivity: Enhancing coding speed and reducing manual effort in repetitive coding tasks.
  • Contextual Code Generation: Generating code that seamlessly integrates with existing codebases.