Install and reproduce Case A

New to the project? Use the full eight-step guide for ZIP download, terminal setup, input definitions, expected output and troubleshooting. This page is the short ordered checklist.

1. Download and enter the repository root

git clone https://github.com/OlenaKompaniiets/tde-dust-echo.git
cd tde-dust-echo

Without Git, download and extract the full repository ZIP, then open a Conda-enabled terminal in the extracted folder containing environment.yml. Do not run from inside the ZIP or from docs/.

2. Create/activate the environment

conda env create -f environment.yml
conda activate ugc-dust
python --version

Create once; activate again in each new terminal. environment.yml targets Python3.11 with minimum package versions, not an exact lock.

3. Run two inexpensive checks first

python tools/check_install.py
python tools/check_saved_case.py

Both must pass. The stored archived χ² is58.95064072368716. No RT calculation is launched by these checks. Do not alter hashes, hosts or uncertainties to force a pass.

4. Compute one fixed archived Case A

python -u tools/run_case_a.py --plots

Wait for completion and retain the printed results/case_a_... folder name. The wrapper creates a new directory on every invocation. It does not run a grid or automatically resume a previous interrupted folder.

5. Check your new outputs

python tools/check_saved_case.py --folder results/PASTE_YOUR_CASE_A_FOLDER

Replace the placeholder. Inspect summary.json, epochs.csv, response.csv, provenance.json and corrected thermal_publication_v2 figures. Confirm the temperature-domain gate, actual numerics, energy ledger and fit before interpreting the model.

6. Figures from saved outputs

python tools/plot_thermal.py --folder examples/case_a

This regenerates corrected thermal figures from the archive without refitting. For your own completed plotted run, substitute its folder. Online visualisations remain the archived example.

7. Refined Case A and conditional variations

Follow steps7–8 of the walkthrough. A single evaluation of the refined saved configuration is distinct from regenerating conditional profile intervals. The latter requires original response caches, plans, baseline/checkpoint state, module layout and validated runtime. Do not launch a large staged search as an automatic next step or assume the dated source supplement is a turnkey inference package.

See what the conditional parameter ranges mean. More compute alone does not establish statistical coverage.