Skip to content
This repository has been archived by the owner on May 15, 2023. It is now read-only.

What is a spurious disagreement in the quantization example? #4

Open
vv-ss opened this issue Feb 6, 2019 · 0 comments
Open

What is a spurious disagreement in the quantization example? #4

vv-ss opened this issue Feb 6, 2019 · 0 comments

Comments

@vv-ss
Copy link

vv-ss commented Feb 6, 2019

Hi,

I am interested in the accuracy loss due to quantization and was running the quantized_fuzzer.py example. In the script I see that we first get a "result" when the objective function is not met, namely argmax for logits and quantized_logits differ. And then, we check whether the disagreement is correct or spurious. Is this to capture non-determinism in floating point operation? I see that the loop runs 10 times for the same input. Is that intentional?

Thanks!

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant