Ubuntu-tensorflow: The program ends the problem of not releasing GPU video memory

The author runs TensorFlow program on Ubuntu, midway using the Win+c key to end the program, but the GPU's video memory is not released, has been in the occupied state.
Using commands
Watch-n 1 Nvidia-smi
Shows the following

Two GPU programs are in execution, in fact, gpu:0 has been stopped by the author, but the GPU is not released, the process continues, so only the use of violence, the process manually shut down, the process number as the Red line part of the figure, because the author in the two GPU run program, it is difficult to find themselves from the program name, But you can find your own PID from the Gpu:num.
The close command is as follows:
sudo kill-9 PID

Good, perfect solution GPU video Memory release problem