For some molecules, openbabel will generate a large number of error/warning messages. When using weights and biases, which captures/reports the stderr, these error messages can cause a deadlock that brings training to a halt. This deadlock is avoided by suppressing error and warning messages from openbabel, which is what this PR does.
For some molecules, openbabel will generate a large number of error/warning messages. When using weights and biases, which captures/reports the stderr, these error messages can cause a deadlock that brings training to a halt. This deadlock is avoided by suppressing error and warning messages from openbabel, which is what this PR does.