Closed ZJG0 closed 3 years ago
The data was just handpicked 8 samples from MNIST and hence has been removed. The repo provides more complete testing datasets in the preload
folder.
I assume you're asking about this line (please use links/references in your future questions). There is probably a bug there but that will load zeros into the variables and hence will still help you get performance overhead. The data i/o pipeline will have to be done and I am happy to pull in any improvements on that end if you would like to contribute.
Thank you for your answer. I am happy to solve this bug, but i hava no time now. After a while, if i solve this bug, i will submit PR.
I don't find the
files/train_data_A
etc. why do every party get the same data?