That’s plausible. is this overhead consistent with the slowing down after the NUTS % counting started? I mean its not just in the beginning of the program execution that suffers from the cpu sampler overhead.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| GPU utilization is high but memory usage is very low leading to subpar sampling performance | 2 | 6227 | August 26, 2017 | |
| Sampling time GPU vs CPU | 1 | 2877 | February 7, 2023 | |
| PyMC3 Pickle Issue with GPU | 3 | 2585 | July 24, 2018 | |
| Gpu and cpu showing different log prob numbers | 2 | 604 | May 21, 2018 | |
| Optimization suggestion for Hierarchical Model using NUTS on CPU/GPU | 7 | 1369 | June 4, 2018 |