I am running on Google Colab, which has one CPU (as well as a GPU), and thus runs jobs sequentially. During tuning, it runs the first 200 samples quickly. But on the second chain, it goes very slow and this often means the second chain takes much longer overall. Some examples from two similar but …

No early tuning in sequential second chain

byblian March 4, 2019, 11:18pm 5

But then do I have the diagnostics to compare the chains for convergence?

Topic		Replies	Views
`start` argument with multiple chains? Questions	1	544	November 26, 2018
Optimization suggestion for Hierarchical Model using NUTS on CPU/GPU Questions gpu , theano	7	1369	June 4, 2018
Chains parameter and Multiprocess sampling Questions	6	4461	August 30, 2018
Tt.shared variable making sampling very slow Questions	7	892	January 5, 2018
Initialize Once for Multiple Runs? Questions	2	722	January 10, 2020