HI: all I do some cv models cpu inference experiments recently. I have some questions about the results of cpu experiments.
- when I auto tuning with different threads.I found as long as I use the same threads to inference.The speed of the model is the same. Is it normal? 2.I inference model with different threads.I found when I use the default threads (half of the total cpu threads) the model fastest. Why not use all threads the model fastest? Forward for you answer