Open lyzKF opened 5 years ago
I suggest that you can rewrite a lite version used tf.Session() , it is much easier to operate
@282866787 都是中国人,加微信详细聊聊?
libeiiebil 发送自 Windows 10 版邮件https://go.microsoft.com/fwlink/?LinkId=550986应用
发件人: lyz notifications@github.com 发送时间: Thursday, January 17, 2019 8:10:56 PM 收件人: tensorflow/tensor2tensor 抄送: LiBei; Mention 主题: Re: [tensorflow/tensor2tensor] Ensemble (#1375)
@282866787https://github.com/282866787 都是中国人,加微信详细聊聊?
― You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/tensorflow/tensor2tensor/issues/1375#issuecomment-455150541, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AZhGZhAxbXXz5igatHV3OPsfX4j3dHlSks5vEGhQgaJpZM4aCiVR.
I suggest that you can rewrite a lite version used tf.Session() , it is much easier to operate
Hi, @libeineu ! I am new to tf and t2t as i used pytorch before. Now i need to implement the model ensemble for tensor2tensor machine translation task, could you provided some more information about what you mentioned above "rewrite a lite version used tf.Session()" ? thanks !
Description
...Hi, I want to ensemble models on Tensor2tensor but found it a bit difficult with tf estimators. Maybe it would be better using raw session I guess. Does anyone can share me some suggestion to work out it?
Environment information
My t2t version is 1.11.0
Error logs:
...