From fce4e1d1b48f23cd8332e60afce3df8d6209a6a7 Mon Sep 17 00:00:00 2001
From: gaochangfeng <54253717+gaochangfeng@users.noreply.github.com>
Date: 星期四, 11 四月 2024 14:59:22 +0800
Subject: [PATCH] SenseVoice对富文本解码的参数 (#1608)
---
funasr/models/lcbnet/model.py | 5 ++---
1 files changed, 2 insertions(+), 3 deletions(-)
diff --git a/funasr/models/lcbnet/model.py b/funasr/models/lcbnet/model.py
index e83f8d7..3ac319c 100644
--- a/funasr/models/lcbnet/model.py
+++ b/funasr/models/lcbnet/model.py
@@ -426,12 +426,11 @@
tokenizer=tokenizer)
time2 = time.perf_counter()
meta_data["load_data"] = f"{time2 - time1:0.3f}"
- pdb.set_trace()
audio_sample_list = sample_list[0]
if len(sample_list) >1:
ocr_sample_list = sample_list[1]
else:
- ocr_sample_list = [294, 0]
+ ocr_sample_list = [[294, 0]]
speech, speech_lengths = extract_fbank(audio_sample_list, data_type=kwargs.get("data_type", "sound"),
frontend=frontend)
time3 = time.perf_counter()
@@ -445,7 +444,7 @@
encoder_out, encoder_out_lens = self.encode(speech, speech_lengths)
if isinstance(encoder_out, tuple):
encoder_out = encoder_out[0]
- pdb.set_trace()
+
ocr_list_new = [[x + 1 if x != 0 else x for x in sublist] for sublist in ocr_sample_list]
ocr = torch.tensor(ocr_list_new).to(device=kwargs["device"])
ocr_lengths = ocr.new_full([1], dtype=torch.long, fill_value=ocr.size(1)).to(device=kwargs["device"])
--
Gitblit v1.9.1