torch-sensitivity

Repository for the development of multi-fidelity torch uncertainty quantification methods and analysis

We recommend installing this repository as a package by running:

pip3.11 install -e .

at the top level directory, since it contains useful utility files for the entire torch UQ pipeline.

We also recommend symbolically linking torch1d.py at the top level directory for execution of samples

ln -s /path/to/torch1d/torch1d.py .

Directories

tuq_main: Contains the primary UQ pipeline scripts tuq_misc: Miscellaneous scripts for examining sample runs and meshes tuq_util: Useful utility classes and functions

Main UQ Pipelines in tuq_main

Scripts to automate/parallelize these processes for many samples are in dane_scripts

Building reaction rate KL models and sampling them

build_rate_kl_models.py
resample_rates.py

Sampling Torch1D

rate_data_to_torch1d_input.py
run_torch_samples.py

Post-processing

sf_torch1d_uq_post.py

Optional

compare_qoi_dists.py
reaction_sensitivity.py
reaction_sens_plots.py

Sampling TPS (Axisymmetric)

rate_data_to_tps2d_input.py
manage_bigcomp_runs.py
manage_bigcomp_runs.py --run

Post-processing

sf_tps2d_uq_post.py