Open rougejaw opened 8 months ago
Possible cause is in TensoRF/main.py:nerf_worker(). The pika connection and training occur on the same thread, and the pika conneciton sends a hearbeat every 300s. If training takes >300(s) the pika connection cannot send a im good and rabbitmq terminates the connection.
During end-to-end testing, nerf_worker is unable to connect to rabbitmq, timing out & causing the process to halt. Another notable issue in the screenshot is multiple progress bars being printed, when only one should appear.
The following log was from a run the only change to the TensoRF config file being "n_iters = 1000". Note that the test completes as expected when "n_iters = 1"