Time series Timeseries Deep Learning Machine Learning Python Pytorch fastai | State-of-the-art Deep Learning library for Time Series and Sequences in Pytorch / fastai
The automated decoding functionality in get_X_preds does not work reliably for me. Sometimes it decodes the probabilities in a way that the decoding for an individual sample is an array with 3 labels instead of a single label. With some datasets it works correctly with others it gives multiple labels. See this screenshot:
The workaround is simple as you can just decode the probabilities yourself as in the screenshot; but it is still a weird bug.
Hi @lwachowiak,
Thanks for sharing this. It's a bit difficult to debug it if I don't understand what's the input and expected output.
Could you please share a code snippet to reproduce the issue?
The automated decoding functionality in get_X_preds does not work reliably for me. Sometimes it decodes the probabilities in a way that the decoding for an individual sample is an array with 3 labels instead of a single label. With some datasets it works correctly with others it gives multiple labels. See this screenshot:
The workaround is simple as you can just decode the probabilities yourself as in the screenshot; but it is still a weird bug.
My setup: