Poor Accuracy of ADVI for Linear Regression

Questions
10 
t1 = time.time()
advi = pm.ADVI()
t2 = time.time()
approax = advi.fit(n=10000)
t3 = time.time()

I tried that but there is only 2-4 ms time difference between t3-t1 and t3-t2.

However, when t3-t2 = 5709 ms, the time shown by PyMC3 progress bar is 1s.

Average Loss = 24.219: 100%|██████████| 10000/10000 [00:01<00:00, 7736.26it/s]
Inference : ADVI, wall time : 5709ms, full time : 5712ms

@junpenglao are these the expected values? shouldn’t the t3-t2 should be something close to 1s?