Open tsjain opened 8 months ago
From the command line help, they both give p(s_i given backbone), but the latter option additionally states that the numbers come from the forward pass.
Can you explain what the difference is between the two options?
@tsjain my guess is that unconditional_probs_only ignores the mask that you provide, as e.g. in example 7.
unconditional_probs_only
From the command line help, they both give p(s_i given backbone), but the latter option additionally states that the numbers come from the forward pass.
Can you explain what the difference is between the two options?