What parameters would you like me to run this with? Running it with 8 chains and parallel (which is what Iβm doing for my model) just immediately fails the test. Running with 8 chains and sequential I get:
sample: 100%|β| 20/20 [00:00<00:00, 21.07it/s, 3 steps of size 1.22e+00. acc. pr
sample: 100%|β| 20/20 [00:00<00:00, 2108.27it/s, 1 steps of size 9.51e-01. acc.
sample: 100%|β| 20/20 [00:00<00:00, 1872.75it/s, 1 steps of size 7.83e-01. acc.
sample: 100%|β| 20/20 [00:00<00:00, 1970.17it/s, 1 steps of size 7.35e-01. acc.
sample: 100%|β| 20/20 [00:00<00:00, 1900.97it/s, 1 steps of size 1.02e+00. acc.
sample: 100%|β| 20/20 [00:00<00:00, 1866.91it/s, 7 steps of size 8.82e-01. acc.
sample: 100%|β| 20/20 [00:00<00:00, 1889.20it/s, 3 steps of size 6.76e-01. acc.
sample: 100%|β| 20/20 [00:00<00:00, 2200.46it/s, 3 steps of size 8.39e-01. acc.
warmup: 100%|β| 10/10 [00:00<00:00, 2073.51it/s, 7 steps of size 1.22e+00. acc.
warmup: 100%|β| 10/10 [00:00<00:00, 2028.59it/s, 1 steps of size 9.51e-01. acc.
warmup: 100%|β| 10/10 [00:00<00:00, 1961.33it/s, 15 steps of size 7.83e-01. acc.
warmup: 100%|β| 10/10 [00:00<00:00, 1619.55it/s, 15 steps of size 7.35e-01. acc.
warmup: 100%|β| 10/10 [00:00<00:00, 2067.48it/s, 3 steps of size 1.02e+00. acc.
warmup: 100%|β| 10/10 [00:00<00:00, 1734.47it/s, 9 steps of size 8.82e-01. acc.
warmup: 100%|β| 10/10 [00:00<00:00, 1946.76it/s, 7 steps of size 6.76e-01. acc.
warmup: 100%|β| 10/10 [00:00<00:00, 1571.49it/s, 1 steps of size 8.39e-01. acc.
sample: 100%|β| 10/10 [00:00<00:00, 2087.65it/s, 3 steps of size 1.22e+00. acc.
sample: 100%|β| 10/10 [00:00<00:00, 1859.18it/s, 1 steps of size 9.51e-01. acc.
sample: 100%|β| 10/10 [00:00<00:00, 1908.15it/s, 1 steps of size 7.83e-01. acc.
sample: 100%|β| 10/10 [00:00<00:00, 1735.69it/s, 1 steps of size 7.35e-01. acc.
sample: 100%|β| 10/10 [00:00<00:00, 1572.02it/s, 1 steps of size 1.02e+00. acc.
sample: 100%|β| 10/10 [00:00<00:00, 2007.23it/s, 7 steps of size 8.82e-01. acc.
sample: 100%|β| 10/10 [00:00<00:00, 1611.71it/s, 3 steps of size 6.76e-01. acc.
sample: 100%|β| 10/10 [00:00<00:00, 2008.77it/s, 3 steps of size 8.39e-01. acc.
sample: 100%|β| 20/20 [00:00<00:00, 22.55it/s, 7 steps of size 6.33e-01. acc. pr
---------------------------------------------------------------------------
AssertionError Traceback (most recent call last)
Input In [28], in <cell line: 1>()
----> 1 test_compile_warmup_run(8, 'sequential', True)
Input In [23], in test_compile_warmup_run(num_chains, chain_method, progress_bar)
42 mcmc.run(rng_key)
43 first_chain_samples = mcmc.get_samples()["x"]
---> 44 assert_allclose(actual_samples[:num_samples], first_chain_samples, atol=1e-5)
[... skipping hidden 1 frame]
File ~/miniconda3/envs/bcore/lib/python3.9/site-packages/numpy/testing/_private/utils.py:844, in assert_array_compare(comparison, x, y, err_msg, verbose, header, precision, equal_nan, equal_inf)
840 err_msg += '\n' + '\n'.join(remarks)
841 msg = build_err_msg([ox, oy], err_msg,
842 verbose=verbose, header=header,
843 names=('x', 'y'), precision=precision)
--> 844 raise AssertionError(msg)
845 except ValueError:
846 import traceback
AssertionError:
Not equal to tolerance rtol=1e-07, atol=1e-05
Mismatched elements: 10 / 10 (100%)
Max absolute difference: 2.4036353
Max relative difference: 37.921455
x: array([-1.116398, -1.116398, 0.089071, 0.841793, 0.886094, 0.648453,
1.747691, -0.288125, -0.721614, 0.112518], dtype=float32)
y: array([ 0.030237, -0.993181, -1.382091, -0.197607, -0.769425, -0.806987,
0.407783, 0.505296, -0.859371, -2.291117], dtype=float32)
I cant see if there is any compiling going on here as it is so quick. It seems like there is no delay though.