Run training script on baseline mode.

CosmoStat / ShapeDeconv

Repo for Machine Learning-based deconvolution

MIT License

7 stars 1 forks source link

Run training script on baseline mode. #3

Closed EiffL closed 3 years ago

EiffL commented 3 years ago

We need a script to train the model. Fadi is working on it here: https://github.com/CosmoStat/ShapeDeconv/blob/master/scripts/tikhonet_train.py

Things to do there:

[x] : Test the script without constraints (in interactive session)
[x] : Write a job file for sbatch: http://www.idris.fr/jean-zay/gpu/jean-zay-gpu-exec_mono_batch.html
[x] : Train a model to the end on JZ

fadinammour commented 3 years ago

To be able to load the model, the option compile=False was used in the method tf.keras.load_model otherwise it would have raised an error of the form custom_loss: tikho_loss is not defined. Is this solution suitable for our case?