How to do data parallelism for num_return_sequences in generation pipeline

🤗Accelerate
1

is there an easy way to parallelize:

from transformers import pipeline

# Load the text generation pipeline with a small model
generator = pipeline('text-generation', model='distilgpt2')

# Define the prompt
prompt = "Once upon a time"

# Generate text
generated_text = generator(prompt, max_length=50, num_return_sequences=100)

# Print the generated text
print(generated_text[0]['generated_text'])