gistfile1.txt

from dataformer.components.evol_instruct import EvolInstruct
from dataformer.llms import AsyncLLM
from datasets import load_dataset

# Take any sample dataset
dataset = load_dataset("dataformer/self-knowledge")
datasetsub = dataset["train"].select(range(2))
instructions = [example["question"] for example in datasetsub]


llm = AsyncLLM(api_provider="together",
			   model="meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo",
			   sampling_params={"temperature": 0.6, "top_p": 1}) 

evol_instruct = EvolInstruct(
    llm=llm,
    num_evolutions=2,  # Number of times to evolve each instruction
    store_evolutions=True,  # Store all evolutions
    generate_answers=True,  # Generate answers for the evolved instructions
    # include_original_instruction=True  # Include the original instruction in the results
)

results = evol_instruct.generate(instructions)
print(results)