Text generation pipeline - output_scores parameter

[bala1802]

In text-generation pipeline, I am looking for a parameter which calculates the confidence score of the generated text. Source: here

I am assuming that, output_scores (from here) parameter is not returned while prediction,

Code:

predictedText = pipeline('text-generation',model=checkpoint_path, tokenizer=gpt2_tokenizer, config={'max_length':20, 'output_scores':True})

predictedText('This is a ')

Output:

Setting pad_token_idtoeos_token_id:50256 for open-end generation. [{'generated_text': 'This is a Generated Text'}]

In the output, I am looking for a confidence score of the predicted text to be displayed

[valhalla]

the text-generation pipeline doesn’t return scores, however you could the generate method directly, to get the scores, this should help

Generation Probabilities: How to compute probabilities of output scores for GPT2

Now that it is possible to return the logits generated at each step, one might wonder how to compute the probabilities for each generated sequence accordingly. The following code snippet showcases how to do so for generation with do_sample=True for GPT2: import torch from transformers import AutoModelForCausalLM from transformers import AutoTokenizer gpt2 = AutoModelForCausalLM.from_pretrained("gpt2", return_dict_in_generate=True) tokenizer = AutoTokenizer.from_pretrained("gpt2") input_ids …