EvolveComplexity¶
Overview¶
EvolveComplexity is a singleton that takes an instruction and evolves it into a more complex version while maintaining the core meaning and intent. This is useful for creating variations of instructions with different complexity levels.
Inputs¶
| Field | Type | Description |
|---|---|---|
| instruction | str | The original instruction to be evolved |
Outputs¶
| Field | Type | Description |
|---|---|---|
| evolved_instruction | str | The evolved version of the instruction with increased complexity |
| instruction | str | Original instruction (echoed from input) |
| model | str | The AI model used for generation |
Usage¶
EvolveComplexity instance can be used in data generation as follows:
from dria.factory import EvolveComplexity
my_dataset = DriaDataset(
name="evolve_complexity",
description="A dataset for instruction evolution",
schema=EvolveComplexity.OutputSchema,
)
generator = DatasetGenerator(dataset=my_dataset)
Expected output¶
{
"evolved_instruction":"Write a short story about a cat who, unbeknownst to its human family, communicates with other cats in a secret language that revolves around solving mysteries within the neighborhood. The cat must navigate between two worlds: the simple life of domesticity and the complex web of feline intrigue, all while trying not to reveal their dual life to their human companions.",
"instruction":"Write a short story about a cat.",
"model":"qwen2.5:32b-instruct-fp16"
}