Out of memory error when using Mirrored strategy in Tensorflow

Question

I read the data and processed it using the following code :

data = pd.read_csv(\'Step1_output.csv\')
data = data.sample(frac=1).reset_index(drop=True)
data1 =