Yash Datta

cc @isnotinvain cc @liancheng request your thoughts on this ? I have tested with latest spark master , and was able to eliminate schema read/ reconciliation for spark-sql purposes (where...

@danielcweeks this is possible to do in spark as well (have tried, works for projections) but how do you resolve column names in filter objects which are pushed down in...

@julienledem request your thoughts on this ...

Reopened this https://github.com/apache/parquet-mr/pull/82 after rebasing to changed code base

I think I get the same error ``` [rank22]: File "/opt/conda/lib/python3.11/site-packages/flash_attn/bert_padding.py", line 212, in pad_input [rank22]: output = index_put_first_axis(hidden_states, indices, batch * seqlen) [rank22]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank22]: File "/opt/conda/lib/python3.11/site-packages/torch/autograd/function.py", line 598,...