models/florence2/base/tokenizer_config.json

{
    "model_max_length": 1024
}