part4.1_HG_quantization.ipynb reproducibility problem.

[shakeeluetp1041]

Non-deterministic accuracy of hls_model.predict() in part4.1_HG_quantization.ipynb

When running part4.1_HG_quantization.ipynb, I noticed that the accuracy of the hls_model varies drastically across multiple runs on the same input data. For example, running the same code multiple times produces accuracy values like 72%, 57% 45%, and other intermediate values. I can not figure out the exact reason.