Meet a 'ValueError: Invalid filepath extension for saving' in training

[lindawillli]

Hello, I got a 'ValueError: Invalid filepath extension for saving' when I used 'scaden train' function. I had use Scaden successfully before, but this time I met this error. How could I solve this problem? I am looking forward to your reply. Thank you very much.

My code is shown below.
pip install scaden
scaden simulate --data example_data/ -n 100 --pattern "*_counts.txt"
scaden process data.h5ad example_data/example_bulk_data.txt
scaden train processed.h5ad --steps 5000 --model_dir model

Then I got an error.
ValueError: Invalid filepath extension for saving. Please add either a .keras extension for the native Keras format (recommended) or a .h5 extension. Use model.export(filepath) if you want to export a SavedModel for use with TFLite/TFServing/etc. Received: filepath=model/m256.

[sxu509]

Hi, I got same issues too, I figure out by:

Open the Scaden script (scaden.py) or locate the line in scaden/model/scaden.py: self.model.save(self.model_dir)
(in python/lib/site-packages/scaden)

And change to:
os.makedirs(self.model_dir, exist_ok=True)
self.model.save(self.model_dir + "model.keras")

And the example works!

[lindawillli]

Thank you very much for your reply! Your method directly solved my problem!
I also made some attempts and just confirmed that they would be effective as well.
I used Docker. It's worth noting that the scaden version in Docker is v1.1.1. To avoid new errors, I updated pip, scaden (from 1.1.1 to 1.1.2), and anndata (from 0.7.6 to 0.8.0) after running the container.