The dhwanichande29/nl-to-bash model is a 0.5 billion parameter instruction-tuned causal language model, fine-tuned by dhwanichande29 from Qwen2.5-Coder-0.5B-Instruct. It specializes in translating natural language instructions into Bash commands, leveraging a context length of 32768 tokens. This model is optimized for generating accurate and semantically relevant Bash commands from English prompts, making it highly effective for command-line automation and developer tools.
Loading preview...
Overview
This model, dhwanichande29/nl-to-bash, is a fine-tuned version of the Qwen2.5-Coder-0.5B-Instruct model, specifically designed for natural language to Bash command translation. It was trained on 40,639 natural language-to-Bash command pairs from the NL2SH-ALFA dataset, making it highly specialized for this task. The model has 0.5 billion parameters and a context length of 32768 tokens.
Key Capabilities
- Natural Language to Bash Translation: Converts English instructions into executable Bash commands.
- Semantic Accuracy: Achieves a semantic match score of 60.33% (cosine similarity \u2265 0.8) on a held-out test set, indicating its ability to generate functionally equivalent commands.
- Lightweight: Based on a 0.5B parameter model, offering efficient performance for its specialized task.
Performance Metrics
Evaluated on 300 test examples from NL2SH-ALFA:
- Exact Match: 13.67%
- Semantic Match (cosine \u2265 0.8): 60.33%
- Average Similarity: 0.776
Training Details
The model was trained for 10 epochs using bfloat16 precision on an NVIDIA A100 GPU, taking approximately 2.09 hours. The training utilized the westenfelder/NL2SH-ALFA dataset.
Good for
- Developer Tools: Integrating natural language command generation into IDEs or custom scripts.
- Command-line Automation: Automating repetitive Bash tasks by describing them in plain English.
- Educational Purposes: Helping users learn Bash commands by providing natural language prompts.