Hi there,
Recently I detected a problem that when I run the tutorial of the VTA deployment.
VTA_deplot_vision_tutorial Instead of locating exact categories of the cat prediction, I save the exact value of “tvm_output” to check if there is any different between “arm_cpu” and “vta” device. (“sim” mod)
I ran this program and surprisingly found that it shows 0.0 on a Linux server but 1.0917945e-05 on my VM and docker on my Mac OS laptop. I checked the result carefully and found that only the output of the “device = vta” on Mac performs different than others. This happens not only on “resnet18_v1” but all the networks given in the dictionary.
I am kind of confused why their values come different, can someone help me 
As well, I really would like to know if I work on any other parts of TVM, would my laptop be ok?
Thank you 
*I find that not all networks shows the difference, some of my local network can get the output 0.0 *My CPU is Intel(R) Core™ i9-8950HK