Pipeline max_length

🤗Transformers
1

What does the max_length argument in the pipeline function do?

    pipe = pipeline('text2text-generation', model=self.model, tokenizer=self.tokenizer)
    output = []

    for ele in dataset:
        pred = pipe(ele, max_length=1024)
        output.append({'input' : ele ,'output': pred[0]['generated_text']})
2

I think thats the max length in tokens the model can accept, could be wrong though.

3

Hi! The max_length here controls for maximum tokens that can be generated. The generation stops when we reach the maximum. Note that the model might generate incomplete sentences, if you specify max_length too short, by default it is 20 tokens.