mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-12-11 06:55:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

modify max_len to allow longer input audio

Browse Source
This commit is contained in:
danqing fu 2023-06-06 11:09:00 +08:00
parent 8f8b09498f
commit 09c4946c15
1 changed files with 6 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
egs/librispeech/ASR/zipformer/zipformer.py
View File

@ -1305,6 +1305,12 @@ class CompactRelPositionalEncoding(torch.nn.Module):
) -> None: ) -> None:
"""Construct a CompactRelPositionalEncoding object.""" """Construct a CompactRelPositionalEncoding object."""
super(CompactRelPositionalEncoding, self).__init__() super(CompactRelPositionalEncoding, self).__init__()
if torch.jit.is_tracing:
# 10k frames correspond to ~100k ms, e.g., 100 seconds, i.e.,
# It assumes that the maximum input won't have more than
# 10k frames.
#
max_len = 10000
self.embed_dim = embed_dim self.embed_dim = embed_dim
assert embed_dim % 2 == 0 assert embed_dim % 2 == 0
self.dropout = Dropout2(dropout_rate) self.dropout = Dropout2(dropout_rate)