Closed kubilay-atasu closed 1 year ago
@kubilay-atasu yes, those could be used, but is it ok to release the aml_e_small dataset used by these examples? if yes, then no need for a code snippet in the documentation.
There is also a very small dataset I had created for testing the features extracted. See amlcustom* here: https://github.ibm.com/ai-foundation/snapml/tree/jov-graph-preproc/test/graph-preprocessor/input
We are using this dataset already in the unittests: https://github.ibm.com/ai-foundation/snapml/blob/jov-graph-preproc/test/graph-preprocessor/UnitTests.py class GraphFeaturesCustomTest(unittest.TestCase):
The custom dataset is not created using Erik Altman's simulator. We can use it as we wish.
@kubilay-atasu Ok, great! Then let's use this custom dataset. I will create the notebook based on this one. And we don't add any more code snippets to the documentation. Would you agree?
Agreed, but note that the custom dataset is only for testing the feature extraction, and not for inference. There are no labels, and it is too small to train anything on it anyway.
Agreed, but note that the custom dataset is only for testing the feature extraction, and not for inference. There are no labels, and it is too small to train anything on it anyway.
Absolutely, just to show how to use API of the new Snap ML feature.
Forperformance, we will need to wait and see what big dataset to use publicly.
please do not merge until we have the new feature out, thanks!
@andreea-anghel: there is already some data and examples here: https://github.ibm.com/ai-foundation/snapml/tree/jov-graph-preproc/test/graph-preprocessor/examples