This is an automated email from the ASF dual-hosted git repository.
junrushao pushed a change to branch unity-staging
in repository https://gitbox.apache.org/repos/asf/tvm.git
from 8a98ffecc9 Merge branch 'main' into unity
add 25c6218e64 [Unity] Skip shape checking on transformed params (#15736)
add 1965315108 [Unity][BugFix] FuseTIR not keeping ExternFunc in IRModule
(#15782)
add 9fd45741e8 [nn.Module] Add Missing Import nn.MultiLinear (#15784)
add 14f06adca2 [Unity] Keep num_input attribute after ModelBundleParams
(#15789)
add 0fe3ed4424 [Unity] Mark result of LazyTransformParams as impure
function (#15697)
add 2fdedf1ea8 [Disco] Integrate RCCL (#15776)
add 9613385fd0 [Codegen][ROCm] Mismatched Dtype of Workgroup/Workitem
(#15777)
add 17adc1e722 [Hotfix] Fix Windows Pipe (#15778)
new cfed9c2354 Merge remote-tracking branch 'origin/main' into
unity-staging
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "add" were already present in the repository and have only
been added to this reference.
Summary of changes:
CMakeLists.txt | 11 +-
include/tvm/relax/expr.h | 7 +
python/tvm/relax/frontend/nn/__init__.py | 9 +-
.../tvm/relax/transform/lazy_transform_params.py | 8 +-
python/tvm/target/detect_target.py | 2 +-
src/relax/backend/vm/vm_shape_lower.cc | 62 +++++---
src/relax/transform/bundle_model_params.cc | 5 +-
src/relax/transform/fuse_tir.cc | 18 ++-
src/relax/transform/lift_transform_params.cc | 3 +-
src/relax/transform/rewrite_cuda_graph.cc | 3 +-
src/runtime/disco/{nccl/nccl.cc => ccl/ccl.cc} | 127 ++++++++++-----
src/runtime/disco/{nccl => ccl}/utils.h | 23 ++-
src/runtime/rocm/rocm_device_api.cc | 4 +
src/support/pipe.h | 8 +-
src/target/llvm/codegen_amdgpu.cc | 3 +-
src/target/llvm/intrin_rule_rocm.cc | 2 +-
tests/python/disco/{test_nccl.py => test_ccl.py} | 98 ++++++------
.../relax/test_backend_transform_shape_lower.py | 177 +++++++++++++++++++++
.../relax/test_transform_bundle_model_params.py | 2 +
tests/python/relax/test_transform_fuse_tir.py | 8 +
.../relax/test_transform_lazy_transform_params.py | 7 +-
21 files changed, 427 insertions(+), 160 deletions(-)
rename src/runtime/disco/{nccl/nccl.cc => ccl/ccl.cc} (66%)
rename src/runtime/disco/{nccl => ccl}/utils.h (77%)
rename tests/python/disco/{test_nccl.py => test_ccl.py} (88%)