Ghaythfd/gherkin-deepseek-lora

Gherkin Scenario Generator (DeepSeek Coder LoRA)

A fine-tuned LoRA adapter for generating Gherkin BDD test scenarios, built on top of DeepSeek Coder 6.7B Instruct.

Model Description

This model generates Gherkin/Cucumber test scenarios for data management systems. It was fine-tuned on real-world BDD test cases covering:

• Data import/export (CSV, JSON, Excel)
• REST and SOAP API testing
• UI navigation and search
• Job scheduling and reporting
• IBOR and financial data operations

Usage

With Unsloth (Recommended)

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="Ghaythfd/gherkin-deepseek-lora",
    max_seq_length=2048,
    load_in_4bit=True,
)
FastLanguageModel.for_inference(model)

prompt = "### Instruction:\nWrite a Gherkin scenario for testing CSV file import\n\n### Response:\n"
inputs = tokenizer([prompt], return_tensors="pt").to("cuda")

outputs = model.generate(
    **inputs,
    max_new_tokens=512,
    temperature=0.5,
    top_p=0.9,
    repetition_penalty=1.15,
)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

With PEFT/Transformers

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base_model = AutoModelForCausalLM.from_pretrained(
    "deepseek-ai/deepseek-coder-6.7b-instruct",
    device_map="auto",
    load_in_4bit=True,
)
model = PeftModel.from_pretrained(base_model, "Ghaythfd/gherkin-deepseek-lora")
tokenizer = AutoTokenizer.from_pretrained("Ghaythfd/gherkin-deepseek-lora")

Prompt Format

### Instruction:
{your request here}

### Response:

Example Output

Prompt: "Write a Gherkin scenario for testing CSV file import"

Output:

Scenario Outline: Testing CSV file import
    Given I am logged as TAV_standard.user on <screen_name>
    And I go to detail screen <detail_screen> of <object_type>
    When I select the tab <tab>
    Then The fields <fields> should be displayed with values <values>
    
    Examples:
      | screen_name   | object_type | detail_screen | tab      | fields | values |
      | My Securities | Equity      | Detail        | Overview | Name   | Test   |

Training Details

• Base Model: deepseek-ai/deepseek-coder-6.7b-instruct
• Method: LoRA (Low-Rank Adaptation)
• LoRA Rank: 16
• LoRA Alpha: 16
• Training Data: 726 examples from Gherkin feature files
• Epochs: 1
• Framework: Unsloth + TRL

Limitations

• Generates scenarios in the style of the training data (data management domain)
• May hallucinate specific field names or values
• Works best for scenarios similar to the training examples

License

Apache 2.0