About the utilization of CPU and GPU


(jackhuang) #1

When training the model, I input the command:
./DeepSpeech.py --train_files ./data/CV/cv-valid-train-chosed10h.csv --dev_files ./data/CV/cv-valid-dev.csv --test_files ./data/CV/cv-valid-test.csv --checkpoint_dir ../checkPoint1024-1/ --export_dir ./export_model/ --n_hidden 1024 --epoch 75 --validation_step 1 --display_step 1 --train_batch_size 1
The utilizaion of the GPU is about 60%
Howerver, when I set the --train_batch_size to 5, the utlizaiton of the GPU always be 0% and utilization of CPU is 100%. I worder if this is caused by CPU speed is not fast enough.


(kdavis) #2

This can usually be diagnosed by looking at the output of htop and nvidia-smi simultaneously.