From a8591060d3889cd7a72841fa32a7ee64b49db1d2 Mon Sep 17 00:00:00 2001
From: msgk <zxr935867802@outlook.com>
Date: 星期五, 14 二月 2025 14:16:51 +0800
Subject: [PATCH] fix(spk): 修复 speaker embedding 集群后的重新排序问题
---
funasr/models/ct_transformer_streaming/template.yaml | 11 ++++-------
1 files changed, 4 insertions(+), 7 deletions(-)
diff --git a/funasr/models/ct_transformer_streaming/template.yaml b/funasr/models/ct_transformer_streaming/template.yaml
index c20a098..2477ac2 100644
--- a/funasr/models/ct_transformer_streaming/template.yaml
+++ b/funasr/models/ct_transformer_streaming/template.yaml
@@ -27,13 +27,13 @@
- 1.0
sentence_end_id: 3
-encoder: SANMEncoder
+encoder: SANMVadEncoder
encoder_conf:
input_size: 256
output_size: 256
attention_heads: 8
linear_units: 1024
- num_blocks: 4
+ num_blocks: 3
dropout_rate: 0.1
positional_dropout_rate: 0.1
attention_dropout_rate: 0.0
@@ -41,13 +41,10 @@
pos_enc_class: SinusoidalPositionEncoder
normalize_before: true
kernel_size: 11
- sanm_shfit: 0
+ sanm_shfit: 5
selfattention_layer_type: sanm
padding_idx: 0
tokenizer: CharTokenizer
tokenizer_conf:
- unk_symbol: <unk>
-
-
-
+ unk_symbol: <unk>
\ No newline at end of file
--
Gitblit v1.9.1