In remora tests/data, there is a levels.txt file. I know ‘AAAAAAAGA’ is 9-mer, but what does the numerical value mean? In metrics_api.ipynb's graph, I can see that it is related to "model_levels". What is "model levels"? In comments, it explains "First the expected levels are extracted using the basecalled sequence (io_read.seq)." And I could see from code that extract_levels function utilize this levels.txt file. So is this something like the expected value getting from training data? Or am i entirely wrong? Also, what exactly is the input to neural network during training, where can I get this information? In the github readme file, it says "Finally each k-mer is one-hot encoded for input into the neural network. " but the process resulting in those numberical values is still a mistery to me. Could someone give me some hints and point me in the right direction?
AAAAAAAAA -1.8424464464187622
AAAAAAAAC -1.6519798040390015
AAAAAAAAG -1.7665722370147705
AAAAAAAAT -1.6588099002838135
AAAAAAACA -1.4318406581878662
...
TTTTTTTGT 1.1797282695770264
TTTTTTTTA 0.5989069938659668
TTTTTTTTC 0.5715355277061462
TTTTTTTTG 0.6644539833068848
TTTTTTTTT 0.5237446427345276
