generated_response = lm.generate(
    input_ids=inputs["input_ids"],
    attention_mask=inputs["attention_mask"],
    max_new_tokens=max_new_tokens, # The max number of tokens to generate
    do_sample=True,                # Probabilistic sampling
    top_p=0.95,                    # Nucleus sampling
    num_return_sequences=G,        # Number of sequences per question
    temperature=1,                 # Increase randomness
    eos_token_id=eos_token_id,
    pad_token_id=eos_token_id,
)