Z-yq / TensorflowASR

一个执着于让CPU\端侧-Model逼近GPU-Model性能的项目,CPU上的实时率(RTF)小于0.1
Apache License 2.0
461 stars 111 forks source link

关于使用Tester 来大批量测试 #42

Closed oneappleya closed 2 years ago

oneappleya commented 2 years ago

请问可以直接下载您训练好的模型,如ConformerCTC(S),用来测试吗? 我使用您训练好的ConformerCTC(S),在准备好的test_list下,当运行python eval_am.py --data_config ./asr/configs/am_data.yml --model_config ./asr/configs/ConformerS.yml 的时候会出现一下错误: outputs = call_fn(inputs, *args, **kwargs) TypeError: call() got multiple values for argument 'training' 想请问一下这怎么解决

Z-yq commented 2 years ago

可能有bug。。。。我看一下

Z-yq commented 2 years ago

./asr/tester/am_tester.py line:41
translator_out = self.translator(ctc_decode, enc_output, training=False) ====>>>> translator_out = self.translator([ctc_decode, enc_output], training=False)

oneappleya commented 2 years ago

感谢感谢