Run trained kaggle model in jupyter notebook

Question

I have build and trained the model in kaggle and have downloaded it's output. How can I now run the trained model locally in jupyter notebook to make prediction ?

Possible duplicate of [Make predictions using a tensorflow graph from a keras model](https://stackoverflow.com/questions/44274701/make-predictions-using-a-tensorflow-graph-from-a-keras-model) — David Parks, May 06 '19 at 15:57

score 1 · Answer 1 · answered Sep 19 '19 at 23:53

You can save your model in a file using:

model.save('model.hdf5')

Then you can run this in your kaggle kernel:

from IPython.display import FileLink
FileLink(r'model.h5')

It will then generate a link so you can download your hdf5 file.

In your local jupyter notebook run the following:

from keras.models import load_model
model = load_model('model.h5')

score 0 · Answer 2 · answered May 06 '19 at 15:58

I guess you're trying to say How to load pretrained model weights in jupyter.

All you want to do is make a new model in jupyter as the same pretrained model layers and parameters from kaggle. Then use loadweights method. You're ready now to predict now cheers :D

for exmaple:

weights_path="donwloaded_weights.hdf5"
model = YourModel()
model.loadweights(weights_path)
model.predict(your_data)

Run trained kaggle model in jupyter notebook

2 Answers2