Corset | clustering de novo assembled transcripts | Genomics library

You need to change the basic_format and input_data, and you will get the desired output using the below-given code.

Note: this is just a code I ran for small reproducible input and output. You can change the data and play around to get the desired final output

I looked at the tensorflow/examples/image_retraining/retrain.py file which is linked from the tutorial in your post.

Part of the code that runs is a train/validation split that takes place in the function create_image_lists(image_dir, testing_percentage, validation_percentage)

The default value for the split is 10% and it's stated in the FLAGS.validation_percentage

Since you have classes that have less than 200 images, the split return classes with less than 20 photos for validation and that's where the error you are getting occur.

Try to run the code with only the classes with more than 200 images and see if it helps. if it does you can consider adding more images or manipulate the create_image_lists function to return at least 20 photos for validation.