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/paraformer/template.yaml | 11 ++---------
1 files changed, 2 insertions(+), 9 deletions(-)
diff --git a/funasr/models/paraformer/template.yaml b/funasr/models/paraformer/template.yaml
index 94eebf7..249e88c 100644
--- a/funasr/models/paraformer/template.yaml
+++ b/funasr/models/paraformer/template.yaml
@@ -6,7 +6,6 @@
# tables.print()
# network architecture
-#model: funasr.models.paraformer.model:Paraformer
model: Paraformer
model_conf:
ctc_weight: 0.0
@@ -87,14 +86,8 @@
accum_grad: 1
grad_clip: 5
max_epoch: 150
- val_scheduler_criterion:
- - valid
- - acc
- best_model_criterion:
- - - valid
- - acc
- - max
keep_nbest_models: 10
+ avg_nbest_model: 10
log_interval: 50
optim: adam
@@ -107,7 +100,7 @@
dataset: AudioDataset
dataset_conf:
index_ds: IndexDSJsonl
- batch_sampler: DynamicBatchLocalShuffleSampler
+ batch_sampler: BatchSampler
batch_type: example # example or length
batch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;
max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,
--
Gitblit v1.9.1