During RidgeRegression.fit(), we group on ['sample_block', 'label'] but not on 'alpha'. This becomes a limiting factor on our scalability due to PyArrow limits, which constrain the number of 8-byte float values in a vector to 132,152,839; see the limits described in https://github.com/projectglow/glow/pull/282.
During
RidgeRegression.fit()
, we group on['sample_block', 'label']
but not on'alpha'
. This becomes a limiting factor on our scalability due to PyArrow limits, which constrain the number of 8-byte float values in a vector to 132,152,839; see the limits described in https://github.com/projectglow/glow/pull/282.