metadata

language:
  - en
license: mit
task_categories:
  - time-series-forecasting
dataset_info:
  features:
    - name: t
      dtype:
        array2_d:
          shape:
            - 1001
            - 1
          dtype: float64
    - name: x
      dtype:
        array2_d:
          shape:
            - 1001
            - 900
          dtype: float64
    - name: args
      dtype:
        array2_d:
          shape:
            - 1001
            - 1
          dtype: float64
  splits:
    - name: train
      num_bytes: 4413496400
      num_examples: 610
    - name: valid
      num_bytes: 795876400
      num_examples: 110
    - name: test_fast
      num_bytes: 3617620000
      num_examples: 500
    - name: test_medium
      num_bytes: 3617620000
      num_examples: 500
    - name: test_slow
      num_bytes: 3617620000
      num_examples: 500
  download_size: 14279999184
  dataset_size: 16062232800
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: valid
        path: data/valid-*
      - split: test_fast
        path: data/test_fast-*
      - split: test_medium
        path: data/test_medium-*
      - split: test_slow
        path: data/test_slow-*

Dataset for polymer dynamics

References

Chen, X. et al. Constructing custom thermodynamics using deep learning. Nature Computational Science 4, 66–85 (2024).