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
460 stars 110 forks source link

About ConformerTransfucer structure #2

Closed alsm168 closed 3 years ago

alsm168 commented 3 years ago

Hello,I have another question. In your code ConformerTransfucer‘s TransducerPrediction is implemented with LSTM, CanTransducerPrediction be implemented with Conformer block? thank you

Z-yq commented 3 years ago

Maybe you wanne CTC structure.

------------------ 原始邮件 ------------------ 发件人: "Z-yq/TensorflowASR" <notifications@github.com>; 发送时间: 2020年9月27日(星期天) 下午4:25 收件人: "Z-yq/TensorflowASR"<TensorflowASR@noreply.github.com>; 抄送: "Subscribed"<subscribed@noreply.github.com>; 主题: [Z-yq/TensorflowASR] About ConformerTransfucer structure (#2)

Hello,I have another question. In your code ConformerTransfucer‘s TransducerPrediction is implemented with LSTM, CanTransducerPrediction be implemented with Conformer block? thank you

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

alsm168 commented 3 years ago

Thank you. I just read the paper of Conformer ,TransducerPrediction is implemented with LSTM. I think it can also be implemented with transformer.

Z-yq commented 3 years ago

That's a good idea, you can try it.

But it's hard for online streaming.

------------------ 原始邮件 ------------------ 发件人: "Z-yq/TensorflowASR" <notifications@github.com>; 发送时间: 2020年9月27日(星期天) 下午4:45 收件人: "Z-yq/TensorflowASR"<TensorflowASR@noreply.github.com>; 抄送: " Demon丶"<641242921@qq.com>;"Comment"<comment@noreply.github.com>; 主题: Re: [Z-yq/TensorflowASR] About ConformerTransfucer structure (#2)

Thank you. I just read the paper of Conformer ,TransducerPrediction is implemented with LSTM. I think it can also be implemented with transformer.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

alsm168 commented 3 years ago

thanks,I learned a lot from your code. It is great!