ValueError: Unable to convert NumPy array to a Tensor (Unsupported object type float)

Question

I'm having a real problem diagnosing an issue I have with Tensorflow. I get this ValueError and it's not intuitive at all. I've tried converting my floats to np.float32 as specified in other SO questions. I've tried a basket of solutions, but I can't seem to get past it.

My data is in a Pandas dataframe. I joined a TF-IDF dataframe with some other wrangled features to get my final data set. The dimensions are (7176, 1006). I'm using Tensorflow and my model code is:

model = Sequential()

model.add(Dense(500, activation='relu'))
model.add(Conv1D(500, 10, activation='relu', input_shape=input_shape)) # Convultional Layer
model.add(MaxPooling1D(pool_size=10, strides=80)) # MaxPooling
model.add(LSTM(250, dropout=0.2, recurrent_dropout=0.2)) # LSTM Layer
model.add(Dense(111, activation='softmax')) #output layer
model.compile(
    loss='sparse_categorical_crossentropy',
    optimizer='adam',
    metrics=['accuracy']
)

model.fit(X_train, y_train, epochs=10, validation_data=(X_test, y_test))

I've used this data in an AutoKeras model so I'm reasonably sure that the issue is how I've constructed my model (that model never finished, it took way to long). After navigating the inevitable input dimension errors I've arrived here and I want to pull my hair out. Any help or suggestions are greatly appreciated!

Nazmul Hasan · Accepted Answer · 2020-07-02 01:03:27Z

1

You need to convert your final dataset to numpy array. You can use following method:

import numpy as np
input_data = np.array(your_final_data_array).astype(np.float32)

It should work fine now.

answered Jul 2, 2020 at 1:03

Nazmul Hasan

9406 silver badges19 bronze badges

Sign up to request clarification or add additional context in comments.

6 Comments

FuriouStyles Over a year ago

Some columns in my dataset are strings, but the rest of the values have already been converted.

NotAName Over a year ago

Strings will need to get encoded. Additionally, you might need to use some form of scaler to scale your data.

Nazmul Hasan Over a year ago

Your model doesn't support that input format. You will need encode as pavel already mentioned.

FuriouStyles Over a year ago

I was afraid I had forgotten something and I think that was definitely it! I will encode my columns and update you guys. Thanks!

Nazmul Hasan Over a year ago

after encoding, please convert the final data to numpy

|

Collectives™ on Stack Overflow

ValueError: Unable to convert NumPy array to a Tensor (Unsupported object type float)

1 Answer 1

6 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

6 Comments

Your Answer

Sign up or log in

Post as a guest

Related