Version dependant slowing down of Gaussian Mixture sampling in Ubuntu 20.04

43

Well you got it, Now 5.9.1 roughly the same speed as 5.9.0 and 5.8.2.

5.9.1:

pymc version: 5.9.1, random_seed: 111, nclusters 5, n_samples: 1000
using advi: False
Auto-assigning NUTS sampler...
Initializing NUTS using jitter+adapt_diag...
Multiprocess sampling (4 chains in 4 jobs)
NUTS: [w, x_coord, y_coord, sigma]
Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 100 seconds.[8000/8000 01:39<00:00 Sampling 4 chains, 5 divergences]
The rhat statistic is larger than 1.01 for some parameters. This indicates problems during sampling. See https://arxiv.org/abs/1903.08008 for details
The effective sample size per chain is smaller than 100 for some parameters.  A higher number is needed for reliable rhat and ess computation. See https://arxiv.org/abs/1903.08008 for details
There were 5 divergences after tuning. Increase `target_accept` or reparameterize.

5.9.0

pymc version: 5.9.0, random_seed: 111, nclusters 5, n_samples: 1000
using advi: False
Auto-assigning NUTS sampler...
Initializing NUTS using jitter+adapt_diag...
Multiprocess sampling (4 chains in 4 jobs)
NUTS: [w, x_coord, y_coord, sigma]
Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 101 seconds.[8000/8000 01:40<00:00 Sampling 4 chains, 5 divergences]
The rhat statistic is larger than 1.01 for some parameters. This indicates problems during sampling. See https://arxiv.org/abs/1903.08008 for details
The effective sample size per chain is smaller than 100 for some parameters.  A higher number is needed for reliable rhat and ess computation. See https://arxiv.org/abs/1903.08008 for details
There were 5 divergences after tuning. Increase `target_accept` or reparameterize.

5.8.2

pymc version: 5.8.2, random_seed: 111, nclusters 5, n_samples: 1000
using advi: False
Auto-assigning NUTS sampler...
Initializing NUTS using jitter+adapt_diag...
Multiprocess sampling (4 chains in 4 jobs)
NUTS: [w, x_coord, y_coord, sigma]
Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 106 seconds.[8000/8000 01:45<00:00 Sampling 4 chains, 5 divergences]
The rhat statistic is larger than 1.01 for some parameters. This indicates problems during sampling. See https://arxiv.org/abs/1903.08008 for details
The effective sample size per chain is smaller than 100 for some parameters.  A higher number is needed for reliable rhat and ess computation. See https://arxiv.org/abs/1903.08008 for details
There were 5 divergences after tuning. Increase `target_accept` or reparameterize.
2 Likes