[Question]Failed while test a pretrained model based on chronos-bolt-small in gift-eval dataset

[jianbinl]

Hi, I refer this page to pretrain a model based on chronos-bolt-small, pretrainning succesed but failed while test in gift-eval dataset refer this page gift-eval , please help me , how should I fix it.

Processing dataset: electricity/15T (1 of 55)
/mnt/data_2/home_aiagent/miniconda3/envs/ljb56/lib/python3.10/site-packages/gluonts/time_feature/seasonality.py:47: FutureWarning: 'T' is deprecated and will be removed in a future version, please use 'min' instead.
  offset = pd.tseries.frequencies.to_offset(freq)
/mnt/data_2/home_aiagent/projects/causal_generator/cauker_exp/gift_eval/evaluation/../src/gift_eval/data.py:151: FutureWarning: 'T' is deprecated and will be removed in a future version, please use 'min' instead.
  freq = norm_freq_str(to_offset(self.freq).name)
Dataset size: 7400
prediction_length: 48
Unknown architecture: T5ForConditionalGeneration, defaulting to ChronosBoltModelForForecasting
Traceback (most recent call last):
  File "/mnt/data_2/home_aiagent/projects/causal_generator/cauker_exp/gift_eval/evaluation/eval_chronos_cauker.py", line 312, in <module>
    main()
  File "/mnt/data_2/home_aiagent/projects/causal_generator/cauker_exp/gift_eval/evaluation/eval_chronos_cauker.py", line 265, in main
    predictor = ChronosPredictor(
  File "/mnt/data_2/home_aiagent/projects/causal_generator/cauker_exp/gift_eval/evaluation/eval_chronos_cauker.py", line 126, in __init__
    self.pipeline = BaseChronosPipeline.from_pretrained(
  File "/mnt/data_2/home_aiagent/miniconda3/envs/ljb56/lib/python3.10/site-packages/chronos/base.py", line 380, in from_pretrained
    return class_.from_pretrained(  # type: ignore[attr-defined]
  File "/mnt/data_2/home_aiagent/miniconda3/envs/ljb56/lib/python3.10/site-packages/chronos/chronos_bolt.py", line 629, in from_pretrained
    model = class_.from_pretrained(pretrained_model_name_or_path, *args, **kwargs)
  File "/mnt/data_2/home_aiagent/miniconda3/envs/ljb56/lib/python3.10/site-packages/transformers/modeling_utils.py", line 277, in _wrapper
    return func(*args, **kwargs)
  File "/mnt/data_2/home_aiagent/miniconda3/envs/ljb56/lib/python3.10/site-packages/transformers/modeling_utils.py", line 4971, in from_pretrained
    model = cls(config, *model_args, **model_kwargs)
  File "/mnt/data_2/home_aiagent/miniconda3/envs/ljb56/lib/python3.10/site-packages/chronos/chronos_bolt.py", line 161, in __init__
    self.chronos_config = ChronosBoltConfig(**config.chronos_config)
TypeError: ChronosBoltConfig.__init__() got an unexpected keyword argument 'eos_token_id'

the config I used is modified by chronos-t5-small.yaml, I modify the model_id "amazon/chronos-bolt-small"

[shchur]

Looks like you are trying to load the original Chronos model (e.g. https://huggingface.co/amazon/chronos-t5-small) using the Chronos-Bolt pipeline (e.g. https://huggingface.co/autogluon/chronos-bolt-base). I cannot provide more detailed guidance without seeing the exact code that reproduces this issue.