The unit test on the recent dev branch failed because our tests for TestNormalChol.test_mean_and_var and TestNormal.test_mean_and_var were just outside the tolerance threshold. We can fix this by increasing the number of samples drawn from the distribution, but rather than specifying this value for each test in test_distributions.py, we can follow webPPL's example where the precision is configurable (say, at 0.99 for unit tests), and we can calculate the number of samples needed. Refer to the parent task - #16
neerajprad
commented
7 years ago
The unit test on the recent dev branch failed because our tests for
TestNormalChol.test_mean_and_varandTestNormal.test_mean_and_varwere just outside the tolerance threshold. We can fix this by increasing the number of samples drawn from the distribution, but rather than specifying this value for each test intest_distributions.py, we can follow webPPL's example where the precision is configurable (say, at 0.99 for unit tests), and we can calculate the number of samples needed. Refer to the parent task - #16