dataset_info:
- config_name: large_100
features:
- name: lrs
sequence:
array4_d:
shape:
- 3
- 16
- 16
- 16
dtype: float32
- name: hr
dtype:
array4_d:
shape:
- 3
- 64
- 64
- 64
dtype: float32
splits:
- name: train
num_bytes: 268237120
num_examples: 80
- name: validation
num_bytes: 33529640
num_examples: 10
- name: test
num_bytes: 33529640
num_examples: 10
download_size: 329464088
dataset_size: 335296400
- config_name: large_50
features:
- name: lrs
sequence:
array4_d:
shape:
- 3
- 16
- 16
- 16
dtype: float32
- name: hr
dtype:
array4_d:
shape:
- 3
- 64
- 64
- 64
dtype: float32
splits:
- name: train
num_bytes: 134118560
num_examples: 40
- name: validation
num_bytes: 16764820
num_examples: 5
- name: test
num_bytes: 16764820
num_examples: 5
download_size: 164732070
dataset_size: 167648200
- config_name: small_50
features:
- name: lrs
sequence:
array4_d:
shape:
- 3
- 4
- 4
- 4
dtype: float32
- name: hr
dtype:
array4_d:
shape:
- 3
- 16
- 16
- 16
dtype: float32
splits:
- name: train
num_bytes: 2220320
num_examples: 40
- name: validation
num_bytes: 277540
num_examples: 5
- name: test
num_bytes: 277540
num_examples: 5
download_size: 2645696
dataset_size: 2775400
Super-resolution of Velocity Fields in Three-dimensional Fluid Dynamics
This dataset loader attempts to reproduce the data of Wang et al. (2024)'s experiments on Super-resolution of 3D Turbulence.
References:
- Wang et al. (2024): "Discovering Symmetry Breaking in Physical Systems with Relaxed Group Convolution"
Usage
For a given configuration (e.g. large_50
):
>>> ds = datasets.load_dataset("dl2-g32/jhtdb", name="large_50")
>>> ds
DatasetDict({
train: Dataset({
features: ['lrs', 'hr'],
num_rows: 40
})
validation: Dataset({
features: ['lrs', 'hr'],
num_rows: 5
})
test: Dataset({
features: ['lrs', 'hr'],
num_rows: 5
})
})
Each split contains the input lrs
which corresponds on a sequence of low resolution samples from time t - ws/2, ..., t, ... ts + ws/2
(ws = window size) and hr
corresponds to the high resolution sample at time t
. All the parameters per data point are specified in the corresponding metadata_*.csv
.
Specifically, for the default configuration, for each datapoint we have 3
low resolution samples and 1
high resolution sample. Each of the former have shapes (3, 16, 16, 16)
and the latter has shape (3, 64, 64, 64)
.
Replication
This dataset is entirely generated by scripts/generate.py
and each configuration is fully specified in their corresponding scripts/*.yaml
.
Usage
python -m scripts.generate --config scripts/small_100.yaml --token edu.jhu.pha.turbulence.testing-201311
This will create two folders on datasets/jhtdb
:
- A
tmp
folder that will store all samples accross runs to serve as a cache. - The corresponding subset,
small_50
for example. This folder will contain ametadata_*.csv
and data*.zip
for each split.
Note:
- For the small variants, the default token is enough, but for the large variants a token has to be requested. More details here.
- For reference, the
large_100
takes ~15 minutes to generate for a total of ~300MB.