Max Podkorytov

parametrize input lists to `fbgemm::tbe_input_combine_with_lengths` as templates

5

Summary: Prepare to supply lightweight wrappers instead of `std::vector` Differential Revision: D35620952

Unbackout "[static-runtime] add out variant for `fbgemm::tbe_input_combine_with_length`"

1

Differential Revision: D35750418

Support materialized views of joins of multiple tables

2

See issue https://github.com/prestodb/presto/issues/17359 Test plan - Added testMaterializedViewResultForThreeTableJoin() in TestHiveLogicalPlanner.java and made sure it passes ``` == RELEASE NOTES == Hive Changes * Support materialized views of joins of multiple...

add container for called fn sequence on cpp side of the model

2

Summary: we can iterate on the structure of the container if/when we want to it adds some flexibility to the order of the kernel calls during runtime Differential Revision: D43612905

reduce the number of groupnorm tests

2

Differential Revision: D43131944

refactor softmax test

11

Summary: created a module mimicking torch.nn interface separated iteration logic from inputs specification Differential Revision: D43547253

[Inductor][ROCm] Composable Kernel backend for Inductor

13

This PR adds an alternative backend for Inductor, adding Composable Kernel Universal GEMM instances to the autotune instance selection. The implementation is heavily influenced by the series of PRs which...

Model checkpoint files

1

Hi, thank you for sharing the code! I believe in order to run inference I would need the checkpoint files, could you add them as well? Thank you

[Inductor][ROCm][CK] add CK grouped conv2d fwd kernels to ROCm codegen

7

Plug into lowering and end to end test in a later PR Instance parsing companion PR https://github.com/ROCm/composable_kernel/pull/1585 cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225...

## Proposed changes The current implementation of static_ford uses recursion over the sequence of dimensions and instantiates lambdas at each internal step. This can be avoided and should improve compilation...