From cf41c0aa3ac3f1a3517c2b3a1b6981386d4bc64b Mon Sep 17 00:00:00 2001
From: 志浩 <neo.dzh@alibaba-inc.com>
Date: 星期四, 09 三月 2023 17:24:41 +0800
Subject: [PATCH] modify unit test for speech_diarization_sond-en-us-callhome-8k-n16k4-pytorch
---
funasr/models/pooling/statistic_pooling.py | 7 +++++--
1 files changed, 5 insertions(+), 2 deletions(-)
diff --git a/funasr/models/pooling/statistic_pooling.py b/funasr/models/pooling/statistic_pooling.py
index dc8c98f..f493470 100644
--- a/funasr/models/pooling/statistic_pooling.py
+++ b/funasr/models/pooling/statistic_pooling.py
@@ -82,13 +82,16 @@
tt = xs_pad.shape[2]
num_chunk = int(math.ceil(tt / pooling_stride))
pad = pooling_size // 2
- features = F.pad(xs_pad, (0, 0, pad, pad), "reflect")
+ if len(xs_pad.shape == 4):
+ features = F.pad(xs_pad, (0, 0, pad, pad), "reflect")
+ else:
+ features = F.pad(xs_pad, (pad, pad), "reflect")
stat_list = []
for i in range(num_chunk):
# B x C
st, ed = i*pooling_stride, i*pooling_stride+pooling_size
- stat = statistic_pooling(features[:, :, st: ed, :], pooling_dim=pooling_dim)
+ stat = statistic_pooling(features[:, :, st: ed], pooling_dim=pooling_dim)
stat_list.append(stat.unsqueeze(2))
# B x C x T
--
Gitblit v1.9.1