Update modeling_moss_audio_tokenizer.py

modeling_moss_audio_tokenizer.py

@@ -941,7 +941,7 @@ class MossAudioTokenizerPatchedPretransform(nn.Module):
         x = x.reshape(b, d, -1, h).permute(0, 1, 3, 2).reshape(b, d * h, -1)
         # We pad the input waveform to a multiple of `downsample_rate` before applying the encoder.
         # Use a ceil division to match that padding and avoid dropping the last (partially padded) frame.
-        output_lengths = (input_lengths + self.patch_size - 1) // self.patch_size
+        output_lengths = input_lengths // self.patch_size
         return x, output_lengths
 
     def decode(self, x, input_lengths):