This is an automated email from the ASF dual-hosted git repository.

github-bot pushed a change to branch nightly
in repository https://gitbox.apache.org/repos/asf/tvm.git


    from 63f9cd6523 [Relax] Alloc BYOC workspace with R.builtin.alloc_tensor 
(#17110)
     add 73cad19cfa [Relax][VM] Improved error messages for mismatched 
parameter count (#17118)
     add 3c6ca5d92b [Bugfix][Relax] Set purity=false for LazySetOutput (#17119)
     add a84adaf0ff [CudaGraph] Handle exceptions thrown while capturing cuda 
graph (#17113)
     add 4a5e22e869 [BugFix][MetaSchedule] MultiLevelTilingTensorCore generates 
inconsistent thread-binding sketch for batched matmul (#17012)

No new revisions were added by this update.

Summary of changes:
 .../schedule_rule/multi_level_tiling.cc            | 23 +++++-
 .../schedule_rule/multi_level_tiling.h             |  2 +-
 .../multi_level_tiling_tensor_core.cc              |  2 +-
 src/relax/backend/vm/exec_builder.cc               |  4 +
 src/relax/transform/lazy_transform_params.cc       |  3 +-
 src/runtime/relax_vm/cuda/cuda_graph_builtin.cc    | 81 +++++++++++++++----
 src/runtime/relax_vm/vm.cc                         | 20 ++++-
 .../test_meta_schedule_schedule_rule_mlt_tc.py     | 93 +++++++++++-----------
 .../relax/test_transform_lazy_transform_params.py  | 24 +++---
 tests/python/relax/test_vm_cuda_graph.py           | 77 +++++++++++++++++-
 10 files changed, 245 insertions(+), 84 deletions(-)

Reply via email to