Hao Wu

The sample doesn't seem to trigger the sparse path

4

It looks the default ml20-all sample doesn't call sparse kernels. All I see is cublasSgemm. Log says "NNDataSet::CalculateSparseDatapointCounts: Maximum sparse datapoints (9254) per example in dataset gl_input too large for...

rules_cuda example fails to run

1

When I run ` bazel run --cuda //examples:hello_cuda`, it fails with the following error gcc-8: error: unrecognized command line option '-Wthread-safety'; did you mean '-fthread-jumps'? gcc-8: error: unrecognized command line...

Update tp support in muon

3

# What does this PR do ? Improve partition_dim checker, i.e. explicitly check if it "is None". Get rid of some out of date fallback. pg_collection.use_mpu_process_group now has all the...