search

Fabian-Sc85 / non-invasive-bp-estimation-using-deep-learning

Assessment of non-invasive blood pressure prediction from PPG and rPPG signals using deep learning
GNU General Public License v3.0
151 stars 36 forks source link

Error about Slapnicar's model #4

Closed v3551G closed 2 years ago

v3551G commented 2 years ago

Hi, I try to reproduce your experiment, and successfully tests three models named Alexnet, ResNet and Lstm. However, when I run slapnicar's model, a bug is reported as follows: Screenshot from 2022-01-16 17-55-36 TypeError: Unable to serialize 125.0 to JSON. Unrecognized type <class 'tensorflow.python.framework.ops.EagerTensor'>.

It seems that the error is occurred when the first Epoch is finished and during the save of some tensor.

Can you help me?

Thanks.

Fabian-Sc85 commented 2 years ago

Hi, which Tensorflow version did you use and did you use your own script to train Slapnicar's model or did you use one from the repository?

The implementation of Slapnicar's model is quite old and relies on many outdated Python packages. I had some trouble myself getting it to work with recent versions of Tensorflow.

v3551G commented 2 years ago

Thanks for your reply: The experiment environment is Ubuntu 20.04, tensorflow-gpu2.4.1 with python-3.8.12. I directly used the code downloaded from your repository and made no changes, Besides, I tried tensorflow-gpu 2.7 and the same error is reported. Besides, The other three models are successfully tested.

Keke Qin

发件人: Fabian @.> 发送时间: 2022年1月17日 16:59 收件人: @.> 抄送: @.>; @.> 主题: Re: [Fabian-Sc85/non-invasive-bp-estimation-using-deep-learning] Error about Slapnicar's model (Issue #4)

Hi, which Tensorflow version did you use and did you use your own script to train Slapnicar's model or did you use one from the repository?

The implementation of Slapnicar's model is quite old and relies on many outdated Python packages. I had some trouble myself getting it to work with recent versions of Tensorflow.

― Reply to this email directly, view it on GitHubhttps://github.com/Fabian-Sc85/non-invasive-bp-estimation-using-deep-learning/issues/4#issuecomment-1014282081, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ADCP7RPCHFESBIT5VBPPNVLUWPK5LANCNFSM5MCSQJRQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub. You are receiving this because you authored the thread.Message ID: @.***>

Fabian-Sc85 commented 2 years ago

I believe I solved the problem. The error occurred when saving the model checkpoint after each epoch. Apparently, Tensorflow good upset when I converted the parameter of the diff-function I used with a lambda layer for the calculation of the first and second derivative to tf.constant. It is weird though because it used to work in the past.