Question about the time performance

[coder-chenzhi]

On our environment, each run of the 10-fold takes nearly 3.75 hours.

So, the total time used for training is 3.75 hours or 37.5 hours?

In fact, I have ran classification.py on GCJ dataset to train the model. However, it has taken about 140 hours for training, which is a significant performance degradation. Below is the configuration of my hardware environment.

• Intel i7 6700 3.4GHz
• NVIDIA Titan XP 11G
• DDR4 2400MHz, 32GB

The main frequency of CPU and clock rate of memory are slightly lower than yours. In my opinion, it is unreasonable to observe such degradation of time performance.

Can you give more details about the calculation of the reported training time? Thanks!