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 89c40611b3 [Unity][MSC][M0.4 && M0.5] Codegen && Test (#15645)
add c5b7afc721 [Unity] Implemented BundleModelParams transform (#15657)
add ec4a8b3011 [Unity] Implement relax.Function.bind_params (#15626)
add f0869fecc4 [Unity][Dlight] Matmul rule on int32 workloads (#15486)
add e1725ac01d [Unity][Dlight] Fallback rule supporting more spatial
workloads (#15687)
add 8820ce48d1 [Unity][MSC][special] Change special names (#15691)
add 755af1fec7 [Unity] Added known tir.Expr to relax.PrimValue (#15577)
add 98ba395b6f [Disco][Op] gather_to_worker0 (#15690)
add 60e90a96a8 [Unity] Update remaining calls to `export_library` with
keyword arguments (#15695)
add bec2b8f48d [Unity] Preserve ShapeExpr in EliminateCommonSubexpr
transform (#15701)
add 7df7b7b1b0 [Unity][Analysis] Check for usage of DataflowVar in
all_vars() (#15698)
add 9e7227e084 [Disco][Op] scatter_from_worker0 (#15680)
add e9aef6a479 [Unity] Add `has_function` to RelaxVM (#15674)
add 4956e4f260 [Disco] Add LoadAll method to Disco Shard Loader (#15673)
add 32ed4f00de [Unity] Fix CUTLASS tests following LiftTransformParams
signature change (#15707)
add 70d23100f6 [Unity] Clear slots before each shape lowering (#15712)
add c0b8953a1f [Unity] Dynamic-shape param support in LazyTransformParams
(#15713)
add 1c90ca7c7c [Unity][Frontends][Onnx] fixup resize2d dynamic input
support (#15692)
add 35164b379f [Module] Implement custom imported modules serialization
(#15666)
add 5d3f3dd644 [microTVM] Check the output of microNPU demos in CI (#15667)
add 738c2e9e90 [VM][Adreno] Fix using buffers for weights in VM (#15671)
add 666bd14d82 [Runtime] Make `CSourceModule` and `StaticLibraryModule`
Binary Serializable (#15693)
add d5a4f66fdc [FFI] Propagate Python errors across FFI boundaries (#15596)
add cd7d64e914 Fix a bug caused by PyTorch instance_norm when the input
shape is [1,1,1,2] (#15683)
add ff45cfbee8 [TVMScript][FIX] Disable `show_object_address` in printing
by default (#15705)
add 7322769474 [TVMScript] Disable `black_format` by default (#15706)
add f23d6b2434 [Relay][Bugfix] fix the wrong implementation of Softplus in
OneFlow (#15717)
add f719151486 [Bugfix][Strategy] Fix `arm_cpu` int8 conv2d strategy for
dotprod and i8mm targets (#15711)
add 2032b44f42 [TOPI] Ensure vectorization of input padding in `arm_cpu`
int8 conv2d interleaved schedule (#15710)
add 25b8a0798e [Hopper TMA] Add intrinsic to create barriers for
synchronization (#15684)
new 6d1b58dc0a Merge remote-tracking branch 'main' into `unity`
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:
include/tvm/relax/attrs/ccl.h | 12 ++
include/tvm/relax/struct_info.h | 15 +-
include/tvm/relax/transform.h | 2 +-
include/tvm/relax/utils.h | 22 +++
include/tvm/runtime/c_runtime_api.h | 7 +
include/tvm/runtime/registry.h | 45 ++++++
include/tvm/tir/builtin.h | 21 ++-
python/tvm/_ffi/_ctypes/packed_func.py | 26 ++--
python/tvm/_ffi/_cython/base.pxi | 5 +-
python/tvm/_ffi/_cython/packed_func.pxi | 19 ++-
python/tvm/_ffi/base.py | 149 +++++++++++++++++++-
python/tvm/contrib/hexagon/session.py | 2 +
python/tvm/contrib/hexagon/tools.py | 130 ++++++++++++++++-
python/tvm/contrib/msc/core/ir/graph.py | 16 +--
python/tvm/dlight/gpu/fallback.py | 13 +-
python/tvm/dlight/gpu/matmul.py | 4 +-
python/tvm/relax/expr.py | 50 +++++++
python/tvm/relax/frontend/onnx/onnx_frontend.py | 8 +-
python/tvm/relax/op/ccl/ccl.py | 19 +++
python/tvm/relax/struct_info.py | 63 ++++++++-
.../tvm/relax/transform/lazy_transform_params.py | 13 +-
python/tvm/relax/transform/legalize_ops/ccl.py | 34 ++++-
python/tvm/relax/transform/transform.py | 28 +++-
python/tvm/relay/frontend/oneflow.py | 7 +-
python/tvm/relay/frontend/pytorch.py | 2 +-
python/tvm/relay/op/strategy/arm_cpu.py | 49 +++++--
python/tvm/runtime/disco/session.py | 13 ++
python/tvm/runtime/module.py | 24 +++-
python/tvm/runtime/script_printer.py | 12 +-
python/tvm/script/highlight.py | 4 +-
python/tvm/script/ir_builder/tir/ir.py | 2 +
python/tvm/script/parser/relax/entry.py | 43 ++++--
python/tvm/tir/__init__.py | 1 +
python/tvm/tir/op.py | 75 +++++-----
python/tvm/tir/schedule/schedule.py | 4 +-
python/tvm/tir/schedule/trace.py | 4 +-
python/tvm/topi/arm_cpu/conv2d_gemm.py | 7 +-
src/contrib/msc/core/ir/graph.cc | 28 ++--
src/contrib/msc/core/ir/graph.h | 28 ++--
src/contrib/msc/core/ir/graph_builder.cc | 8 +-
src/contrib/msc/core/transform/set_expr_name.cc | 8 +-
src/ir/transform.cc | 88 ++++++++----
src/relax/analysis/analysis.cc | 2 +
src/relax/analysis/struct_info_analysis.cc | 3 +
src/relax/backend/vm/vm_shape_lower.cc | 2 +
src/relax/ir/expr.cc | 2 +-
src/relax/ir/expr_functor.cc | 3 +
src/relax/ir/struct_info.cc | 15 +-
src/relax/ir/struct_info_functor.cc | 17 ++-
src/relax/op/ccl/ccl.cc | 48 +++++++
src/relax/op/ccl/ccl.h | 3 +
src/relax/transform/bind_params.cc | 116 ++++++++++-----
src/relax/transform/bundle_model_params.cc | 119 ++++++++++++++++
src/relax/transform/eliminate_common_subexpr.cc | 2 +
src/relax/transform/lift_transform_params.cc | 106 ++++++++------
src/relax/utils.cc | 56 ++++++++
src/relay/analysis/type_solver.cc | 2 -
src/relay/transforms/annotate_texture_storage.cc | 8 +-
src/runtime/c_runtime_api.cc | 94 ++++++++++++-
src/runtime/disco/builtin.cc | 5 +
src/runtime/disco/builtin.h | 7 +
src/runtime/disco/loader.cc | 52 ++++++-
src/runtime/disco/nccl/nccl.cc | 41 ++++++
src/runtime/logging.cc | 137 ++++++++++++------
src/runtime/registry.cc | 62 +++++++-
src/runtime/relax_vm/executable.cc | 4 +
src/runtime/static_library.cc | 30 +++-
src/script/printer/relax/struct_info.cc | 23 ++-
src/support/ffi_testing.cc | 12 ++
src/target/codegen.cc | 78 ++++++-----
src/target/source/codegen_cuda.cc | 61 +++++---
src/target/source/codegen_cuda.h | 8 ++
src/target/source/ptx.cc | 30 ++--
src/target/source/ptx.h | 36 ++---
src/target/source/source_module.cc | 41 +++++-
src/tir/op/builtin.cc | 7 +
src/tir/schedule/transform.cc | 6 +-
tests/python/disco/test_loader.py | 55 +++++++-
tests/python/disco/test_nccl.py | 55 +++++---
tests/python/dlight/test_gpu_fallback.py | 70 +++++++++
tests/python/dlight/test_gpu_matmul.py | 92 ++++++++++++
tests/python/dlight/test_gpu_matmul_tensorize.py | 26 ++--
tests/python/frontend/oneflow/test_forward.py | 2 +-
tests/python/frontend/pytorch/test_forward.py | 14 ++
tests/python/relax/test_analysis.py | 9 ++
.../relax/test_analysis_struct_info_analysis.py | 43 +++++-
.../relax/test_backend_transform_shape_lower.py | 103 ++++++++++++++
tests/python/relax/test_bind_params.py | 156 +++++++++++++++++++++
tests/python/relax/test_bind_symbolic_vars.py | 87 +++++++++++-
tests/python/relax/test_codegen_cutlass.py | 12 +-
tests/python/relax/test_expr.py | 17 +++
tests/python/relax/test_op_ccl.py | 57 ++++++++
tests/python/relax/test_relay_translator.py | 6 +-
tests/python/relax/test_struct_info.py | 18 ++-
tests/python/relax/test_transform_bind_params.py | 52 +++++++
.../relax/test_transform_bundle_model_params.py | 104 ++++++++++++++
tests/python/relax/test_transform_cse.py | 15 ++
tests/python/relax/test_transform_fold_constant.py | 3 +-
.../relax/test_transform_lazy_transform_params.py | 120 ++++++++++++++--
.../relax/test_transform_legalize_ops_ccl.py | 21 +++
.../relax/test_transform_lift_transform_params.py | 90 ++++++------
tests/python/relax/test_tvmscript_parser.py | 46 +++++-
tests/python/relax/test_tvmscript_printer_relax.py | 4 +-
tests/python/relax/test_vm_build.py | 4 +-
.../opencl_texture/test_conv2d_nchw_texture.py | 77 ++++++++--
.../relay/opencl_texture/utils/adreno_utils.py | 18 +--
.../relay/strategy/test_select_implementation.py | 8 ++
tests/python/relay/test_pass_instrument.py | 16 +--
tests/python/relay/test_type_infer.py | 2 +-
...eta_schedule_schedule_rule_apply_custom_rule.py | 2 +-
.../unittest/test_roundtrip_runtime_module.py | 12 +-
tests/python/unittest/test_runtime_error.py | 102 ++++++++++++--
.../unittest/test_runtime_module_property.py | 2 +-
tests/python/unittest/test_tir_op_types.py | 20 +--
tests/python/unittest/test_tir_ptx_cp_async.py | 44 ++----
.../test_tir_transform_inject_ptx_async_copy.py | 14 +-
tests/scripts/task_demo_microtvm.sh | 31 +++-
web/tests/python/prepare_test_libs.py | 2 +-
web/tests/python/relax_rpc_test.py | 2 +-
119 files changed, 3318 insertions(+), 663 deletions(-)
create mode 100644 src/relax/transform/bundle_model_params.cc
create mode 100644 tests/python/relax/test_bind_params.py
create mode 100644 tests/python/relax/test_transform_bundle_model_params.py
