DeepSeek React Reviewer

A fine-tuned DeepSeek-Coder model focused on reviewing React and JavaScript code, identifying issues, and suggesting best practices.

This model is optimized to act as a code reviewer, not just a code generator.


Base Model

  • DeepSeek-Coder (deepseek-ai/deepseek-coder)
  • Architecture: Causal Language Model

Fine-Tuning

  • Method: LoRA (Low-Rank Adaptation)
  • Framework: Unsloth + Transformers
  • Training: Supervised fine-tuning (SFT)
  • Precision: FP16 (LoRA merged)
  • Training quantization: 4-bit (QLoRA-style)

The model uploaded here is fully merged, so it can be used directly for inference without PEFT dependencies.


Training Data

  • Format: Chat-style JSONL
  • Focus areas:
    • React code review
    • JavaScript / JSX patterns
    • Hooks misuse
    • Common bugs and anti-patterns
    • Performance and best practices

⚠️ The dataset may contain synthetic or curated examples and should not be treated as authoritative.


Intended Use

Recommended for:

  • Reviewing React and JavaScript code
  • Explaining bugs and anti-patterns
  • Suggesting improvements and best practices
  • Educational feedback for frontend developers

Not intended for:

  • Security audits
  • Production-critical code review
  • Executing untrusted code

Author

Fine-tuned by Nabin Raj Pandey

Downloads last month
71
Safetensors
Model size
7B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Space using nabinwell/deepseek-react-reviewer 1