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

Use more attention heads in slowest layer.

Browse Source
This commit is contained in:
Daniel Povey 2022-11-11 22:56:14 +08:00
parent f7aff4f507
commit 4988c815c9
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
egs/librispeech/ASR/pruned_transducer_stateless7/train.py
View File

@ -127,7 +127,7 @@ def add_model_arguments(parser: argparse.ArgumentParser):
parser.add_argument( parser.add_argument(
"--num-heads", "--num-heads",
type=str, type=str,
default="8", default="8,8,8,16,8,8",
help="Number of attention heads in the zipformer encoder layers: a single int or comma-separated list.", help="Number of attention heads in the zipformer encoder layers: a single int or comma-separated list.",
) )