Running a deep learning thing on the cloud, but the amount of data is too large, in order to speed up.
I directly load all the datasets into memory. After all, for GPU cloud servers, memory and GPU are not worth the money. . .
However, running the top command found that the buff/cache ratio in linux is quite large, but I did not find the program that occupies this, so I decided to release it manually, so that my data set can be loaded into the memory perfectly.
Empty buff./cache code:
# echo 3 > /proc/sys/vm/drop_caches
This will continue to happily load the dataset =. =