Open JoshVarty opened 4 years ago
Although it's unlikely, we should guard against the possibility that there may be 50+ faces in a given video. For this reason we should ensure that our batch size is never larger than 128 during inference.
Although it's unlikely, we should guard against the possibility that there may be 50+ faces in a given video. For this reason we should ensure that our batch size is never larger than 128 during inference.