I've corrected an omission from the run_ogusa_example.py script that raised and error after the baseline policy was run. The change was to place the following lines just below line 90 in the script:
start_time = time.time()
runner(**kwargs)
print('run time = ', time.time()-start_time)
The issue was that no reform policy was run and then when deltas between the baseline and reform were calculated, that function computing the differences couldn't see the reform file.
The OG-USA repo have been updated to include this change. So you can git fetch and git merge upstream master from that repo and you should be able to run the baseline and reform and get some output to explore with the script @rickecon put in this repo: WB-India\Code\FinalDay\OGUSA_output.py. With 4 cores this seemed to be taking about 1.5 hours for the tax function estimation (if doing age-specific functions) then about 1 hour for the time path solution (SS solves in about 10 seconds). So to do the baseline and reform together you are looking at about 5 hours of time (longer with less ability to parallelize and shorter if you have more cores).
I've corrected an omission from the
run_ogusa_example.pyscript that raised and error after the baseline policy was run. The change was to place the following lines just below line 90 in the script:The issue was that no reform policy was run and then when deltas between the baseline and reform were calculated, that function computing the differences couldn't see the reform file.
The OG-USA repo have been updated to include this change. So you can
git fetchandgit merge upstream masterfrom that repo and you should be able to run the baseline and reform and get some output to explore with the script @rickecon put in this repo:WB-India\Code\FinalDay\OGUSA_output.py. With 4 cores this seemed to be taking about 1.5 hours for the tax function estimation (if doing age-specific functions) then about 1 hour for the time path solution (SS solves in about 10 seconds). So to do the baseline and reform together you are looking at about 5 hours of time (longer with less ability to parallelize and shorter if you have more cores).