How to fix LSTM Model that returns value for epoch size during training different from that in code.

Question

MBUNYA NERVILLE ANYANG 2023-5-25

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1973114-how-to-fix-lstm-model-that-returns-value-for-epoch-size-during-training-different-from-that-in-code

编辑： Ashutosh 2023-5-25

I have this LSTM model constructed for time series analysis. with parameters as such

numFeatures = 6
numHiddenLayers = 100
numResponse = 1
Layers = [...
     sequenceInputLayer(numfeatures);...
     lstmLayer(numHiddenLayer, "OutputMode", "last")
     fullyConnectedLayer(numResponses)
     reluLayer
     regressionLayer
     ];
miniBatchSize = 24
Epoch = 50
 options  = trainingOptions("adam",...
      "ExecutionEnvironment", "auto",...
      "MaxEpoch", Epoch, ...
      "MiniBatchSize", miniBatchSize,...
        "Plots","training-progress");

Now during trainig the Epoch size is kept constant at 30, even though I set it to 50 in code. please how can I fix this and what is the cause?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Ashutosh 2023-5-25

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1973114-how-to-fix-lstm-model-that-returns-value-for-epoch-size-during-training-different-from-that-in-code#answer_1244524

编辑：Ashutosh 2023-5-25

在 MATLAB Online 中打开

Hi,

I am assuming that you are using the R2023a distribution of MATLAB. In that case, your method of assigning name-value arguments ("MaxEpoch", Epoch,"MiniBatchSize", miniBatchSize etc) may not be valid, instead you may have to use Name=Value format. Refer to the MATLAB Documentation for trainingOptions, more specifically the section for Name-Value Arguments. Note that the comma method is used for pre-R2021a as mentioned. Note also, that the default value for epochs is 30 which is being assigned in your case. It should work as intended if you give the options as below:

options  = trainingOptions("adam",...
    ExecutionEnvironment = "auto",...
    MaxEpochs = Epoch, ...
    MiniBatchSize = miniBatchSize,...
    Plots = "training-progress");