openai-community/gpt2 · model output

foxsilverfox

Apr 3

Hello,

Is it normal in text generation that the output prediction contains also the input?

Thx

lysandre

OpenAI community org Apr 8

Hey @foxsilverfox , that depends of the code that you use to generate text

foxsilverfox

Apr 8

Hi @lysandre . Well i have a dataset with inputs and one column output.I prepared my prompts and I trained the model.i am using text generation.am using the following for text generation and am replacing the input with empty string is this the normal thing to do?
def generate_predictions(model_name, tokenizer, prompts):
generator = pipeline('text-generation', model='./mygpt')
set_seed(42)
predictions = []

for prompt in prompts:
    generated_text = generator(prompt, max_length=250, num_return_sequences=1)
    modified_text = generated_text[0]['generated_text'].replace(prompt, '')
    predictions.append(modified_text)

return predictions