TIMPS-Coder-0.5B ๐ ๏ธ
Fine-tuned coding model for bug fixing by Sandeep Reddy ยท TIMPS Brand ยท Made in India ๐ฎ๐ณ
Benchmark
| Model | Score |
|---|---|
| Base (Qwen/Qwen2.5-Coder-0.5B-Instruct) | 88.0% |
| TIMPS-Coder (this) | 92.0% |
| 10-task bug-fix benchmark: NullPointer, KeyError, AsyncBug, ScopeBug, RecursionError, etc. |
Quickstart โ MLX (Mac M1/M2/M3)
pip install mlx-lm
mlx_lm generate --model sandeeprdy1729/TIMPS-Coder-0.5B --max-tokens 500 --temp 0.1 \
--prompt '<|im_start|>system
You are TIMPS-Coder. Explain the bug cause then show fixed code.<|im_end|>
<|im_start|>user
Fix null_pointer: My Spring @Autowired field is null<|im_end|>
<|im_start|>assistant'
Training
- Base:
Qwen/Qwen2.5-Coder-0.5B-Instruct| Method: LoRA rank=16 | HW: Mac M2 Air 8GB - Dataset: 30,000+ clean coding instruction pairs (Python, Java, JS, C++, Go, Rust)
- Framework: MLX-LM | LR: 5e-6 | Iters: 3000
Links
- GitHub: Sandeeprdy1729
- YouTube: @sandeepreddythummala
Apache 2.0 License
- Downloads last month
- 418
Model size
0.5B params
Tensor type
BF16
ยท
Hardware compatibility
Log In to add your hardware
Quantized
Model tree for sandeeprdy1729/TIMPS-Coder-0.5B
Base model
Qwen/Qwen2.5-0.5B Finetuned
Qwen/Qwen2.5-Coder-0.5B Finetuned
Qwen/Qwen2.5-Coder-0.5B-Instruct