Hi, I saw the link from your study and tried to use your code and followed the instruction in the readme file, but the output was not as I expected.
Firstly, I tested the best model with precompiled file and the output in all 3 models are all very low around 52~55%. Secondly, I found that the deposit version in the link and the latest submit in this master are different especially in test file. I'm not sure the bad output result I got is because of the different precompiled file. I also tried to recompile the sequence file but the SCRATCH took unacceptable runtime.
I'm curious how the author complete SCRATCH generating features of all sequence (because it really take a lot of time on my platform that I couldn't wait for the program finish). Also, could you please help me double check the result of your precompile file performed on your original platform with your latest submit file?
Thanks a lot.
Hi, I saw the link from your study and tried to use your code and followed the instruction in the readme file, but the output was not as I expected. Firstly, I tested the best model with precompiled file and the output in all 3 models are all very low around 52~55%. Secondly, I found that the deposit version in the link and the latest submit in this master are different especially in test file. I'm not sure the bad output result I got is because of the different precompiled file. I also tried to recompile the sequence file but the SCRATCH took unacceptable runtime. I'm curious how the author complete SCRATCH generating features of all sequence (because it really take a lot of time on my platform that I couldn't wait for the program finish). Also, could you please help me double check the result of your precompile file performed on your original platform with your latest submit file? Thanks a lot.