microsoft

/

Florence-2-large

like 1.44k

Follow

Microsoft 9.79k

I would like to know how the scores are calculated when setting "output_scores=True", please provide a reference for the score's meaning.

def run_example_with_score(task_prompt, text_input=None):
    if text_input is None:
        prompt = task_prompt
    else:
        prompt = task_prompt + text_input
    inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, torch_dtype)
    generated_ids = model.generate(
      input_ids=inputs["input_ids"],
      pixel_values=inputs["pixel_values"],
      max_new_tokens=1024,
      num_beams=3,
      return_dict_in_generate=True,
      output_scores=True,
    )
    generated_text = processor.batch_decode(generated_ids.sequences, skip_special_tokens=False)[0]
    prediction, scores, beam_indices = generated_ids.sequences, generated_ids.scores, generated_ids.beam_indices
    transition_beam_scores = model.compute_transition_scores(
        sequences=prediction,
        scores=scores,
        beam_indices=beam_indices,
    )
    parsed_answer = processor.post_process_generation(sequence=generated_ids.sequences[0], 
        transition_beam_score=transition_beam_scores[0],
        task=task_prompt, image_size=(image.width, image.height)
    )

    print(parsed_answer)

here is the resource: https://huggingface.co/microsoft/Florence-2-large/discussions/56/files