Closed skirdey closed 2 weeks ago
Am I mistaken in assumption that the cond projection is actually not being used? According to the paper, it should be concatenated with the input. Maybe I am missing something.
https://github.com/lucidrains/e2-tts-pytorch/blob/e0326d0f5a32fecf68d97f95003ab2557a471f5a/e2_tts_pytorch/e2_tts.py#L743
@skirdey ah thanks for the catch, should be fixed
are you going to try out the latest improvisation with the dual text / audio transformer?
awesome, yeah going to give it a go!
Am I mistaken in assumption that the cond projection is actually not being used? According to the paper, it should be concatenated with the input. Maybe I am missing something.
https://github.com/lucidrains/e2-tts-pytorch/blob/e0326d0f5a32fecf68d97f95003ab2557a471f5a/e2_tts_pytorch/e2_tts.py#L743