Rishabh Jain

Yes there is problem of over-fitting in the model if trained for whole 125 epochs. - Dropout is used to solve over-fitting but here it is not solving the problem...

No i have not preprocessed the data set. Yeah i think after preprocessing accuracy will certainly improve.

I had to update gen_requirements.py file to fix ml_dtypes version in https://github.com/apache/tvm/blob/main/python/gen_requirements.py ```sed -i '/("matplotlib", None),/a\ ("ml_dtypes", "==0.2.0"),' gen_requirements.py ```

@tqchen Should tvm install process fix the version of ml_dtypes https://github.com/apache/tvm/blob/main/python/gen_requirements.py to avoid installing higher versions which don't support Python 3.8?