Search code examples
pythontensorflowlstmvalueerror

Error in LSTM nlp multiclass model :- ValueError: Shapes (None, 1) and (None, 3) are incompatible

I am working on an email-classifier (using Tensorflow and keras) for whom I created a dataset myself. It has 3 classes - spam, feedback/advice and complaint. It is giving an error: ValueError: Shapes (None, 1) and (None, 3) are incompatible when I fit the model. I read a few answers on it which said that I need to use to_categorical() on my training and validation data (X_train, X_val, Y_train, Y_val). But that also gives a new errors like axis error and more commonly my session on colab crashes. I am totally confused please help:

My code: https://colab.research.google.com/drive/1GWMx1p9XhQjbC-CD5CQsVpVIn4jvxG6m?usp=sharing

My dataset: https://www.kaggle.com/ayaanmustafa/email-classification

Solution

  • See my answer here. In short, convert your target variable to one hot. It's shape is (batch_size, 1) and it should be (batch_size, 3)