README.md

---
library_name: transformers
pipeline_tag: translation
license: mit
datasets:
- westenfelder/NL2SH-ALFA
language:
- en
base_model: Qwen/Qwen2.5-Coder-3B-Instruct
model-index:
- name: Qwen2.5-Coder-3B-Instruct-NL2SH
  results:
  - task:
      type: translation
      name: Natural Language to Bash Translation
    dataset:
      type: translation
      name: NL2SH-ALFA
      split: test
    metrics:
      - type: accuracy
        value: 0.51
        name: InterCode-ALFA
    source:
      name: InterCode-ALFA
      url: https://arxiv.org/abs/2502.06858
---

# Model Card for Qwen2.5-Coder-3B-Instruct-NL2SH
This model translates natural language (English) instructions to Bash commands.

## Model Details
### Model Description
This model is a fine-tuned version of the [Qwen2.5-Coder-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-3B-Instruct) model trained on the [NL2SH-ALFA](https://huggingface.co/datasets/westenfelder/NL2SH-ALFA) dataset for the task of natural language to Bash translation (NL2SH). For more information, please refer to the [paper](https://arxiv.org/abs/2502.06858).
- **Developed by:** [Anyscale Learning For All (ALFA) Group at MIT-CSAIL](https://alfagroup.csail.mit.edu/)
- **Language:** English
- **License:** MIT License
- **Finetuned from model:** [Qwen/Qwen2.5-Coder-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-3B-Instruct)

### Model Sources
- **Repository:** [GitHub Repo](https://github.com/westenfelder/NL2SH)
- **Paper:** [LLM-Supported Natural Language to Bash Translation](https://arxiv.org/abs/2502.06858)

## Uses
### Direct Use
This model is intended for research on machine translation. The model can also be used as an educational resource for learning Bash.

### Out-of-Scope Use
This model should not be used in production or automated systems without human verification.  

**Considerations for use in high-risk environments:** This model should not be used in high-risk environments due to its low accuracy and potential for generating harmful commands.

## Bias, Risks, and Limitations
This model has a tendency to generate overly complex and incorrect Bash commands. It may produce harmful commands that delete data or corrupt a system. This model is not intended for natural languages other than English, scripting languages or than Bash, or multi-line Bash scripts. 

### Recommendations
Users are encouraged to use this model as Bash reference tool and should not execute commands without verification.

## How to Get Started with the Model
Use the code below to get started with the model.
```python
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

def translate(prompt):
    model_name = "westenfelder/Qwen2.5-Coder-3B-Instruct-NL2SH"
    tokenizer = AutoTokenizer.from_pretrained(model_name, clean_up_tokenization_spaces=False)
    model = AutoModelForCausalLM.from_pretrained(model_name, device_map="cuda", torch_dtype=torch.bfloat16)
        
    messages = [
        {"role": "system", "content": "Your task is to translate a natural language instruction to a Bash command. You will receive an instruction in English and output a Bash command that can be run in a Linux terminal."},
        {"role": "user", "content": f"{prompt}"},
    ]

    tokens = tokenizer.apply_chat_template(
        messages,
        add_generation_prompt=True,
        tokenize=True,
        return_tensors="pt"
    ).to(model.device)

    attention_mask = torch.ones_like(tokens)

    outputs = model.generate(
        tokens,
        attention_mask=attention_mask,
        max_new_tokens=100,
        do_sample=False,
        temperature=None,
        top_p=None,
        top_k=None,
    )
    
    response = outputs[0][tokens.shape[-1]:]
    return tokenizer.decode(response, skip_special_tokens=True)


nl = "List files in the /workspace directory that were accessed over an hour ago."
sh = translate(nl)
print(sh)
```

## Training Details
### Training Data
This model was trained on the [NL2SH-ALFA](https://huggingface.co/datasets/westenfelder/NL2SH-ALFA) dataset.

### Training Procedure
Please refer to section 4.1 and 4.3.4 of the [paper](https://arxiv.org/abs/2502.06858) for information about data pre-processing, training hyper-parameters and hardware.

## Evaluation
This model was evaluated on the [NL2SH-ALFA](https://huggingface.co/datasets/westenfelder/NL2SH-ALFA) test set using the [InterCode-ALFA](https://github.com/westenfelder/InterCode-ALFA) benchmark.

### Results
This model achieved an accuracy of **0.51** on the InterCode-ALFA benchmark.

## Environmental Impact
Experiments were conducted using a private infrastructure, which has a approximate carbon efficiency of 0.432 kgCO2eq/kWh. A cumulative of 12 hours of computation was performed on hardware of type RTX A6000 (TDP of 300W). Total emissions are estimated to be 1.56 kgCO2eq of which 0 percents were directly offset. Estimations were conducted using the [Machine Learning Emissions Calculator](https://mlco2.github.io/impact#compute).

## Citation
**BibTeX:**
```
@misc{westenfelder2025llmsupportednaturallanguagebash,
      title={LLM-Supported Natural Language to Bash Translation}, 
      author={Finnian Westenfelder and Erik Hemberg and Miguel Tulla and Stephen Moskal and Una-May O'Reilly and Silviu Chiricescu},
      year={2025},
      eprint={2502.06858},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2502.06858}, 
}
```

## Model Card Authors
Finn Westenfelder

## Model Card Contact
Please email finnw@mit.edu or make a pull request.