search

yeyupiaoling / PPASR

基于PaddlePaddle实现端到端中文语音识别，从入门到实战，超简单的入门案例，超实用的企业项目。支持当前最流行的DeepSpeech2、Conformer、Squeezeformer模型
Apache License 2.0
804 stars 128 forks source link

An error occurred when use pretrained model to train WenetSpeech dataset #30

Closed tcexeexe closed 2 years ago

tcexeexe commented 2 years ago

Hello~ l copy the pretrained model form "PPASR_大数据集/models/deepspeech2/best_model" to the "/PPASR/models/deepspeech2/last_model", then l start trainning, but an error occurred. l wonder what's the reason of this problem?

AssertionError: Variable Shape not match, Variable [ linear_0.w_0_moment1_0 ] need tensor with shape (1024, 5451) but load set tensor with shape (1024, 6436)

微信图片_20220205190959
yeyupiaoling commented 2 years ago

Because dictionaries are different sizes. It is recommended to use all the files in the pre-training dataset.

tcexeexe commented 2 years ago

Because dictionaries are different sizes. It is recommended to use all the files in the pre-training dataset.

好的，谢谢!

tcexeexe commented 2 years ago

Because dictionaries are different sizes. It is recommended to use all the files in the pre-training dataset. 你好，那如果按照你的意思，我要在原有模型上，添加新的词语，做迁移学习的话，是不是就不行了

yeyupiaoling commented 2 years ago

那个大数据的字典基本已经包括常用的汉字了

tcexeexe commented 2 years ago

那个大数据的字典基本已经包括常用的汉字了

好的