funasr/models/sense_voice/whisper_lib/decoding.py
@@ -63,8 +63,8 @@ else: x = x.to(mel.device) # FIX(funasr): sense vocie # logits = model.logits(x[:, :-1], mel)[:, -1] logits = model.logits(x[:, :], mel)[:, -1] logits = model.logits(x[:, :-1], mel)[:, -1] # logits = model.logits(x[:, :], mel)[:, -1] # collect detected languages; suppress all non-language tokens mask = torch.ones(logits.shape[-1], dtype=torch.bool)
