From 3c349ac0531b07239f37b81254f8568ab80e3f6a Mon Sep 17 00:00:00 2001
From: Han Zhang <45134013+holazzer@users.noreply.github.com>
Date: 星期二, 18 三月 2025 11:45:37 +0800
Subject: [PATCH] fix: use converted token_ids for alignment for sensevoice model with timestamp output (#2429)
---
funasr/datasets/audio_datasets/index_ds.py | 6 ++++++
1 files changed, 6 insertions(+), 0 deletions(-)
diff --git a/funasr/datasets/audio_datasets/index_ds.py b/funasr/datasets/audio_datasets/index_ds.py
index 385218a..39ef409 100644
--- a/funasr/datasets/audio_datasets/index_ds.py
+++ b/funasr/datasets/audio_datasets/index_ds.py
@@ -118,6 +118,12 @@
text_language = data.get("text_language", None)
if text_language is not None:
contents_i["text_language"] = text_language
+ if "emo_target" in data:
+ contents_i["emo_target"] = data["emo_target"]
+ if "event_target" in data:
+ contents_i["event_target"] = data["event_target"]
+ if "with_or_wo_itn" in data:
+ contents_i["with_or_wo_itn"] = data["with_or_wo_itn"]
# audio_language = data.get("audio_language", None)
# if audio_language is not None:
# contents_i["audio_language"] = audio_language
--
Gitblit v1.9.1