Thanks for the checkpoint!
Got a question on batched inference. So the inference input if shaped [B, 80, seq_len]. Now if inputs within the same batch have different effective seq_len, what do you expect us to pad the shorter inputs with?
· Sign up or log in to comment