提交 08b68e4b 编写于 作者: H huangyuxin

change the deepspeech2_online.yaml

上级 0309c36a
......@@ -11,7 +11,7 @@ data:
max_output_input_ratio: .inf
collator:
batch_size: 32 # one gpu
batch_size: 64 # one gpu
mean_std_filepath: data/mean_std.json
unit_type: char
vocab_filepath: data/vocab.txt
......@@ -41,12 +41,12 @@ model:
rnn_direction: forward # [forward, bidirect]
num_fc_layers: 1
fc_layers_size_list: 512,
use_gru: True
use_gru: False
training:
n_epoch: 50
lr: 2e-3
lr_decay: 0.83 # 0.83
lr_decay: 0.9 # 0.83
weight_decay: 1e-06
global_grad_clip: 3.0
log_interval: 100
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册