From 8ec8c071dc18a5f4c6e5f34e6a2ebb3b28d58546 Mon Sep 17 00:00:00 2001
From: 雾聪 <wucong.lyb@alibaba-inc.com>
Date: 星期二, 23 一月 2024 15:49:40 +0800
Subject: [PATCH] rm LoadWav2Char for clients
---
funasr/auto/auto_model.py | 8 ++++----
1 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/funasr/auto/auto_model.py b/funasr/auto/auto_model.py
index 107c78e..f724650 100644
--- a/funasr/auto/auto_model.py
+++ b/funasr/auto/auto_model.py
@@ -377,7 +377,7 @@
result[k] = restored_data[j][k]
else:
result[k] = torch.cat([result[k], restored_data[j][k]], dim=0)
- elif k == 'text':
+ elif k == 'raw_text':
if k not in result:
result[k] = restored_data[j][k]
else:
@@ -398,7 +398,7 @@
if self.spk_model is not None:
all_segments = sorted(all_segments, key=lambda x: x[0])
spk_embedding = result['spk_embedding']
- labels = self.cb_model(spk_embedding, oracle_num=self.preset_spk_num)
+ labels = self.cb_model(spk_embedding.cpu(), oracle_num=self.preset_spk_num)
del result['spk_embedding']
sv_output = postprocess(all_segments, None, labels, spk_embedding.cpu())
if self.spk_mode == 'vad_segment':
@@ -406,12 +406,12 @@
for res, vadsegment in zip(restored_data, vadsegments):
sentence_list.append({"start": vadsegment[0],\
"end": vadsegment[1],
- "sentence": res['text'],
+ "sentence": res['raw_text'],
"timestamp": res['timestamp']})
else: # punc_segment
sentence_list = timestamp_sentence(punc_res[0]['punc_array'], \
result['timestamp'], \
- result['text'])
+ result['raw_text'])
distribute_spk(sentence_list, sv_output)
result['sentence_info'] = sentence_list
--
Gitblit v1.9.1