Dear Prof. Ganapathysubramanian and co-authors,
I hope this message finds you well.
I am currently working with the MPF-Bench dataset and carefully reviewing both the paper and the released files. According to Section 3.3 (Metadata) and Table 3 of the paper, the dimensionless parameters (ρ∗, µ∗, Bo, Re) are described as being included in the input tensor X for each sample.
However, while inspecting the downloaded .npz files (including the train/test splits for S1–S6), I was not able to clearly identify where these four scalar parameters are stored. It appears that the provided tensors primarily contain the spatiotemporal solution fields.
Could you kindly clarify:
Whether the dimensionless parameters are included in the released HuggingFace dataset, and if so, in which file or array they are stored?
If they are embedded as constant fields within the input channels, how they are arranged in the channel dimension?
Whether there is a separate metadata file containing these per-sample parameters?
Understanding this detail is important for my intended use of the dataset in parametric modeling and comparison with physics-informed approaches.
Thank you very much for your time and for making this valuable dataset publicly available. I greatly appreciate your assistance.
Best regards,
XUNgongzi509
Dear Prof. Ganapathysubramanian and co-authors,
I hope this message finds you well.
I am currently working with the MPF-Bench dataset and carefully reviewing both the paper and the released files. According to Section 3.3 (Metadata) and Table 3 of the paper, the dimensionless parameters (ρ∗, µ∗, Bo, Re) are described as being included in the input tensor X for each sample.
However, while inspecting the downloaded .npz files (including the train/test splits for S1–S6), I was not able to clearly identify where these four scalar parameters are stored. It appears that the provided tensors primarily contain the spatiotemporal solution fields.
Could you kindly clarify:
Whether the dimensionless parameters are included in the released HuggingFace dataset, and if so, in which file or array they are stored?
If they are embedded as constant fields within the input channels, how they are arranged in the channel dimension?
Whether there is a separate metadata file containing these per-sample parameters?
Understanding this detail is important for my intended use of the dataset in parametric modeling and comparison with physics-informed approaches.
Thank you very much for your time and for making this valuable dataset publicly available. I greatly appreciate your assistance.
Best regards,
XUNgongzi509