autonomio / talos

Hyperparameter Experiments with TensorFlow and Keras
https://autonom.io
MIT License
1.62k stars 270 forks source link

the new released talos 1.0 don't run the example code #520

Closed zzhulm closed 3 years ago

zzhulm commented 3 years ago

in my old install with pip install --use-feature=2020-resolver git+https://github.com/autonomio/talos@1.0 working with tensorflow 2.1, the talso example code is running fine, but when I installed the new released talos 1.0 by pip install talos, with the tensorflow 2.31 , the tensorflow runs fine, by when run the same talso example code it got error message:

ValueError: Layer sequential_1 expects 1 inputs, but it received 2 input tensors. Inputs received: [<tf.Tensor 'IteratorGetNext:0' shape=(None, 8) dtype=float32>, <tf.Tensor 'ExpandDims:0' shape=(None, 1) dtype=float32>]

can you concur on this ? or anything wrong with the installation? or a bug? Thank you. talos_test.ipynb.zip

ghost commented 3 years ago

This mistake is due to an error in validation_data. You should change validation_data = [x_val, y_val]into validaton_data = (x_val, y_val), and it works perfectly for me.

zzhulm commented 3 years ago

seems to be the problem, it is in model.fit() so it is not talos 's problem, it should be from the tensorflow update inconsistency issue, right?