NL2Shell 0.8B — Natural Language to Shell Commands

Fine-tuned Qwen3.5-0.8B for converting natural language to shell commands.

Quick Start

ollama run hf.co/AryaYT/nl2shell-0.8b

Training

  • Method: QLoRA (r=16, alpha=32, 4-bit NF4, all linear layers)
  • Data: NL2Bash (8,130 pairs) + 40 macOS synthetic
  • Epochs: 3 | Loss: 0.6338 | Hardware: A100 40GB
  • Architecture: Hybrid DeltaNet (75% linear + 25% softmax attention)

Prompt Format (ChatML)

<|im_start|>system
You are an expert shell programmer. Given a natural language request, output ONLY the corresponding shell command. No explanations.<|im_end|>
<|im_start|>user
find all Python files modified today<|im_end|>
<|im_start|>assistant

License

MIT | Built by Arya Teja | CloudAGI

Downloads last month
88
Safetensors
Model size
0.9B params
Tensor type
F32
·
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for AryaYT/nl2shell-0.8b

Finetuned
Qwen/Qwen3.5-0.8B
Quantized
(60)
this model

Dataset used to train AryaYT/nl2shell-0.8b

Space using AryaYT/nl2shell-0.8b 1