Krishnasri2027/qwen25-coder-1.5b-instruct-fine-tuned-merged

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 6, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Krishnasri2027/qwen25-coder-1.5b-instruct-fine-tuned-merged is a 1.5 billion parameter Qwen2.5-Coder-Instruct model fine-tuned by Krishnasri2027. This standalone merged causal language model specializes in Python code modernization, specifically targeting compatibility with Python 3.14. It excels at updating legacy Python syntax and refactoring code to leverage newer language features, making it ideal for automated Python version migration workflows.

Loading preview...

Model Overview

This model, developed by Krishnasri2027, is a specialized 1.5 billion parameter Qwen2.5-Coder-Instruct variant, fine-tuned for Python code modernization. Unlike adapter-based models, this repository contains the standalone merged model weights (FP16), eliminating the need for separate LoRA adapters during inference. It was fine-tuned using QLoRA with a 4-bit NF4 quantized base model and LoRA adapters, then merged for direct use.

Key Capabilities

  • Python 3.14 Modernization: Specifically designed to update Python code to be compatible with Python 3.14 syntax and practices.
  • Legacy Code Refactoring: Capable of modernizing older Python syntax and refactoring implementations to use newer language features.
  • Direct Code Generation: Intended to directly generate modernized Python source code.
  • Integration Ready: Can be integrated into developer assistants, IDE tooling, Python migration pipelines, and automated refactoring systems.

Performance & Training Highlights

  • Fine-tuned with QLoRA, achieving a significant reduction in validation loss from 0.252825 to 0.037586 over 500 steps.
  • Achieved a final mean token accuracy of 98.6229% on the evaluation dataset.
  • Trained on a chat-formatted dataset for Python code modernization, with a maximum sequence length of 256 tokens.

Important Considerations

While highly effective, the model does not guarantee full compatibility or semantic equivalence. Generated code should always be reviewed, parsed, tested, and validated before production use. It is not a replacement for human code review or automated testing.

Finetunes
1 models