This is an automated email from the ASF dual-hosted git repository.
syfeng pushed a change to branch test_all_cases_on_unity
in repository https://gitbox.apache.org/repos/asf/tvm.git
from bb57ff97a6 upd ci
add 44d80e5107 [Unity][Bugfix] Reset window cache current pos when
clearing (#16132)
add 8a2ffee38e [Unity] Do not import SciPy by default (#16136)
add 90edf76716 [Unity][LLM] Add NaN checks during sampling for better
error reporting (#16141)
add 5e61adc0a3 [Unity][DistIR] Enhance PropagateSharding pass (#16094)
add 165b84bc7f Always use int64 in JSON parser (#16145)
add 4c07f6af43 [Runtime] Introduce Type-Checked `TVMArgs::At<T>(i)`
(#16147)
add 4e70c28217 [Runtime] Allowing Packed Arguments in TVM Module VTable
(#16148)
add 29450b927e [Unity][MSC] Enable add attributes while fuse ops (#16128)
add bafd49d4b8 [Unity] Flash infer integration (#16146)
add 9a985714f3 [Unity] Migrate Relax Executable/VM to `TVM_MODULE_VTABLE`
Convention (#16149)
add 925cb2bbaa [Unity][BYOC] Add cutlass finegrained decode matmul (#16144)
add aae1112a65 [Unity] Support constant args in `nn.ExternModule` (#16130)
add 756ce9917f [Unity][3rdparty] Remove TVM in 3rdparty of FlashInfer
(#16155)
add 1de8b347d1 [Unity][DistIR] LowerGlobalViewToLocalView (#16095)
add 2dcb8716e8 [Unity][BlockBuilder] Depracate `BlockBuilder.get()` and
change it to `BlockBuilder.finalize()` (#16090)
add 8f24a272a0 [Unity][MSC][M2.1] Add Manager for compile pipeline (#16163)
add af803cf7b4 [Unity][DLight] Fix `general_reduction` for GroupNorm
(#16161)
add 64fe5a8a89 [Unity][DistIR] Add DTensor struct info propagation rule
for stop_lift_params (#16170)
add c640d0a3c9 [Unity][Web] Fix missing function NVTXScopedRange for web
(#16177)
add 8a6184ccfa [Unity, BYOC] Add check for leaking intemediate variables
for cublas and cudnn (#16175)
add a6adaae5ef [Unity][DistIR] LowerDistIR (#16169)
add 85389efa2c [Unity][BYOC] Fix Flash var_len attention with sliding
window (#16185)
add 68443482c9 [Unity][Bugfix] Handle symbolic matching with
non-structural match (#15994)
add d52a9bf388 [Unity][Transform] Implement RemoveUnusedOutputs (#16117)
add d6015c5643 [Unity][BugFix] Fix a bug in relax gelu_tanh computation
(#16188)
add fe9d2fe57d [Unity][Transform] Implement ExpandTupleArguments (#16115)
add fc324d0f2c [Unity][Transform] Implement RemoveUnusedParameters (#16116)
add 74667b97f0 [Unity] Enable ccache for `nn.SourceModule` (#16189)
add ed2772f9c8 [Unity][MSC][M2.1] Add pruner for model pruning (#16186)
add 9e4e17ca88 [Unity][WebGPU] Get params from cache by name (#16198)
add a2f55a8812 [WEBGPU] Update to latest compilationHints API (#16197)
new c1523ab50a [Unity] Fix Rust IRModule Constructor
new f905b83011 Merge commit 'c1523ab50a035909e7dcf4283be60c08846c0d39'
into test_all_cases_on_unity
The 2 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:
.gitmodules | 3 +
3rdparty/cutlass_fpA_intB_gemm | 2 +-
3rdparty/flashinfer | 1 +
3rdparty/picojson/picojson.h | 5 +-
CMakeLists.txt | 21 +
include/tvm/relax/block_builder.h | 9 +
include/tvm/relax/distributed/axis_group_graph.h | 181 +-
include/tvm/relax/distributed/transform.h | 13 +
include/tvm/relax/transform.h | 41 +-
include/tvm/runtime/packed_func.h | 47 +-
include/tvm/runtime/relax_vm/executable.h | 24 +-
include/tvm/runtime/relax_vm/vm.h | 2 -
python/tvm/autotvm/tuner/__init__.py | 7 +-
python/tvm/contrib/cc.py | 2 +-
python/tvm/contrib/cutlass/attention_operation.py | 3 +-
python/tvm/contrib/cutlass/build.py | 4 +
python/tvm/contrib/cutlass/gemm_operation.py | 16 +-
python/tvm/contrib/cutlass/gen_tensor_op.py | 7 +-
python/tvm/contrib/msc/core/codegen/sources.py | 56 +-
python/tvm/contrib/msc/core/frontend/translate.py | 25 +-
python/tvm/contrib/msc/core/ir/graph.py | 166 +-
python/tvm/contrib/msc/core/runtime/runner.py | 274 +-
.../msc/core/tools}/__init__.py | 6 +-
python/tvm/contrib/msc/core/tools/execute.py | 386 +++
.../msc/core/tools/prune}/__init__.py | 5 +-
python/tvm/contrib/msc/core/tools/prune/method.py | 118 +
python/tvm/contrib/msc/core/tools/prune/pruner.py | 433 +++
python/tvm/contrib/msc/core/tools/tool.py | 1465 ++++++++++
python/tvm/contrib/msc/core/transform/pattern.py | 115 +-
python/tvm/contrib/msc/core/utils/dataset.py | 415 ++-
python/tvm/contrib/msc/core/utils/file.py | 104 +-
python/tvm/contrib/msc/core/utils/info.py | 120 +-
python/tvm/contrib/msc/core/utils/message.py | 48 +-
python/tvm/contrib/msc/core/utils/namespace.py | 14 +-
python/tvm/contrib/msc/core/utils/register.py | 134 +-
.../msc/framework/tensorflow/runtime/runner.py | 36 +-
.../msc/framework/tensorflow/tools}/__init__.py | 4 +-
.../framework/tensorflow/tools/prune}/__init__.py | 4 +-
.../msc/framework/tensorflow/tools/prune/pruner.py | 55 +
.../msc/framework/tensorrt/codegen/codegen.py | 40 +-
.../msc/framework/tensorrt/frontend/translate.py | 38 +-
.../msc/framework/tensorrt/runtime/runner.py | 37 +-
.../msc/framework/tensorrt/tools}/__init__.py | 4 +-
.../framework/tensorrt/tools/prune}/__init__.py | 4 +-
.../msc/framework/tensorrt/tools/prune/pruner.py | 55 +
.../msc/framework/tensorrt/transform/pattern.py | 83 +-
.../contrib/msc/framework/torch/runtime/runner.py | 2 +
.../msc/framework/torch/tools}/__init__.py | 4 +-
.../msc/framework/torch/tools/prune}/__init__.py | 4 +-
.../msc/framework/torch/tools/prune/pruner.py | 55 +
.../contrib/msc/framework/tvm/runtime/runner.py | 34 +-
.../msc/framework/tvm/tools}/__init__.py | 4 +-
.../msc/framework/tvm/tools/prune}/__init__.py | 4 +-
.../msc/framework/tvm/tools/prune/pruner.py | 55 +
.../transform => contrib/msc/pipeline}/__init__.py | 4 +-
python/tvm/contrib/msc/pipeline/manager.py | 845 ++++++
python/tvm/dlight/gpu/general_reduction.py | 19 +-
python/tvm/relax/backend/contrib/cublas.py | 3 +
python/tvm/relax/backend/contrib/cudnn.py | 3 +
python/tvm/relax/backend/contrib/cutlass.py | 36 +-
python/tvm/relax/backend/utils.py | 43 +
python/tvm/relax/block_builder.py | 46 +-
python/tvm/relax/distributed/transform/__init__.py | 7 +-
.../tvm/relax/distributed/transform/transform.py | 22 +
python/tvm/relax/dpl/pattern.py | 4 +-
python/tvm/relax/frontend/nn/core.py | 46 +-
python/tvm/relax/frontend/nn/op.py | 11 +-
python/tvm/relax/frontend/nn/spec.py | 42 +-
python/tvm/relax/ir/instrument.py | 2 +-
python/tvm/relax/op/distributed/__init__.py | 7 +-
python/tvm/relax/op/distributed/distributed.py | 53 +-
python/tvm/relax/transform/__init__.py | 4 +
python/tvm/relax/transform/legalize_ops/nn.py | 14 +-
python/tvm/relax/transform/transform.py | 53 +-
python/tvm/relay/analysis/sparse_conv2d.py | 11 +-
python/tvm/relay/analysis/sparse_dense.py | 11 +-
python/tvm/relay/frontend/tensorflow_ops.py | 6 +-
python/tvm/relay/qnn/op/legalizations.py | 19 +-
.../tvm/script/ir_builder/relax/distributed/ir.py | 1 +
python/tvm/script/parser/relax/dist.py | 1 +
python/tvm/script/parser/relax/parser.py | 3 +
python/tvm/testing/utils.py | 5 +-
python/tvm/topi/arm_cpu/qnn_legalize.py | 8 +-
python/tvm/topi/cuda/sparse.py | 7 +-
python/tvm/topi/nn/conv3d_transpose.py | 6 +-
rust/tvm/src/ir/module.rs | 8 +-
src/contrib/msc/core/codegen/base_codegen.h | 24 +-
src/contrib/msc/core/codegen/code_stack.h | 8 +-
src/contrib/msc/core/codegen/codegen_utils.cc | 2 +-
src/contrib/msc/core/codegen/codegen_utils.h | 75 +-
src/contrib/msc/core/codegen/cpp_codegen.h | 90 +-
src/contrib/msc/core/codegen/py_codegen.h | 59 +-
src/contrib/msc/core/ir/graph.cc | 430 ++-
src/contrib/msc/core/ir/graph.h | 199 +-
src/contrib/msc/core/printer/python_printer.cc | 2 +-
src/contrib/msc/core/transform/fuse_tuple.cc | 2 +
src/contrib/msc/core/transform/set_expr_name.cc | 110 +-
src/contrib/msc/core/utils.cc | 20 +-
src/contrib/msc/framework/tensorflow/codegen.cc | 4 +-
src/contrib/msc/framework/tensorrt/codegen.cc | 96 +-
src/contrib/msc/framework/tensorrt/codegen.h | 6 +
src/contrib/msc/framework/tensorrt/codegen_utils.h | 17 +-
.../msc/framework/tensorrt/tensorrt_opcode.cc | 15 +-
.../msc/framework/tensorrt/tensorrt_opcode.h | 2 +-
src/contrib/msc/framework/torch/codegen.cc | 22 +-
src/contrib/msc/framework/torch/codegen_utils.h | 4 +-
src/contrib/msc/framework/torch/torch_opcode.cc | 46 +
src/contrib/msc/framework/torch/torch_opcode.h | 5 +-
src/contrib/msc/framework/tvm/codegen.cc | 40 +-
src/contrib/msc/framework/tvm/relax_opcode.cc | 17 +-
src/contrib/msc/framework/tvm/relax_opcode.h | 2 +-
src/relax/analysis/well_formed.cc | 5 +-
src/relax/distributed/axis_group_graph.cc | 195 +-
src/relax/distributed/transform/lower_distir.cc | 271 ++
.../transform/lower_global_view_to_local_view.cc | 442 +++
.../distributed/transform/propagate_sharding.cc | 285 +-
src/relax/distributed/transform/utils.cc | 81 +
src/relax/distributed/transform/utils.h | 67 +
src/relax/ir/block_builder.cc | 6 +
src/relax/op/distributed/binary.h | 5 +-
src/relax/op/distributed/{op.cc => ccl.cc} | 23 +-
src/relax/op/distributed/distributed.cc | 55 +
src/relax/op/distributed/linear_algebra.cc | 5 +-
src/relax/op/distributed/manipulate.cc | 10 +-
src/relax/op/distributed/nn.cc | 5 +-
src/relax/op/distributed/op.cc | 10 +
src/relax/op/distributed/statistical.cc | 5 +-
src/relax/op/distributed/utils.cc | 47 +-
src/relax/op/distributed/utils.h | 13 +-
src/relax/transform/expand_tuple_arguments.cc | 187 ++
src/relax/transform/fuse_ops.cc | 37 +-
src/relax/transform/fuse_tir.cc | 98 +-
src/relax/transform/normalize.cc | 103 +
src/relax/transform/remove_unused_outputs.cc | 326 +++
src/relax/transform/remove_unused_parameters.cc | 260 ++
src/runtime/contrib/msc/tensorrt_runtime.cc | 71 +-
src/runtime/relax_vm/executable.cc | 45 +-
src/runtime/relax_vm/lm_support.cc | 15 +-
src/runtime/relax_vm/ndarray_cache_support.cc | 8 +
src/runtime/relax_vm/vm.cc | 299 +-
src/script/printer/relax/call.cc | 7 +-
src/tir/schedule/transform.cc | 13 +
src/tir/schedule/transform.h | 9 +
tests/python/contrib/test_msc/test_graph_build.py | 25 +-
tests/python/contrib/test_msc/test_manager.py | 263 ++
tests/python/contrib/test_msc/test_runner.py | 4 +-
tests/python/contrib/test_msc/test_tools.py | 191 ++
.../contrib/test_msc/test_translate_relax.py | 2 +-
.../contrib/test_msc/test_translate_tensorrt.py | 36 +
tests/python/dlight/test_gpu_general_reduction.py | 256 +-
.../test_distributed_transform_lower_distir.py | 396 +++
...ributed_transform_lower_global_to_local_view.py | 1579 +++++++++++
...est_distributed_transform_propagate_sharding.py | 2988 +++++++++++++-------
tests/python/relax/test_blockbuilder_core.py | 103 +-
tests/python/relax/test_codegen_cutlass.py | 218 +-
.../python/relax/test_frontend_nn_extern_module.py | 64 +-
.../relax/test_transform_expand_tuple_args.py | 79 +
.../relax/test_transform_fuse_ops_by_pattern.py | 23 +
tests/python/relax/test_transform_fuse_tir.py | 157 +-
.../python/relax/test_transform_legalize_ops_nn.py | 75 +-
.../relax/test_transform_normalize_global_var.py | 98 +
.../relax/test_transform_remove_unused_outputs.py | 123 +
.../test_transform_remove_unused_parameters.py | 101 +
web/Makefile | 2 +-
web/emcc/wasm_runtime.cc | 1 +
web/package-lock.json | 12 +-
web/package.json | 2 +-
web/src/runtime.ts | 17 +
web/src/webgpu.ts | 7 +-
169 files changed, 14890 insertions(+), 2381 deletions(-)
create mode 160000 3rdparty/flashinfer
copy python/tvm/{relax/distributed/transform =>
contrib/msc/core/tools}/__init__.py (87%)
create mode 100644 python/tvm/contrib/msc/core/tools/execute.py
copy python/tvm/{relax/distributed/transform =>
contrib/msc/core/tools/prune}/__init__.py (87%)
create mode 100644 python/tvm/contrib/msc/core/tools/prune/method.py
create mode 100644 python/tvm/contrib/msc/core/tools/prune/pruner.py
create mode 100644 python/tvm/contrib/msc/core/tools/tool.py
copy python/tvm/{relax/distributed/transform =>
contrib/msc/framework/tensorflow/tools}/__init__.py (87%)
copy python/tvm/{relax/distributed/transform =>
contrib/msc/framework/tensorflow/tools/prune}/__init__.py (87%)
create mode 100644
python/tvm/contrib/msc/framework/tensorflow/tools/prune/pruner.py
copy python/tvm/{relax/distributed/transform =>
contrib/msc/framework/tensorrt/tools}/__init__.py (87%)
copy python/tvm/{relax/distributed/transform =>
contrib/msc/framework/tensorrt/tools/prune}/__init__.py (87%)
create mode 100644
python/tvm/contrib/msc/framework/tensorrt/tools/prune/pruner.py
copy python/tvm/{relax/distributed/transform =>
contrib/msc/framework/torch/tools}/__init__.py (87%)
copy python/tvm/{relax/distributed/transform =>
contrib/msc/framework/torch/tools/prune}/__init__.py (87%)
create mode 100644 python/tvm/contrib/msc/framework/torch/tools/prune/pruner.py
copy python/tvm/{relax/distributed/transform =>
contrib/msc/framework/tvm/tools}/__init__.py (87%)
copy python/tvm/{relax/distributed/transform =>
contrib/msc/framework/tvm/tools/prune}/__init__.py (87%)
create mode 100644 python/tvm/contrib/msc/framework/tvm/tools/prune/pruner.py
copy python/tvm/{relax/distributed/transform =>
contrib/msc/pipeline}/__init__.py (87%)
create mode 100644 python/tvm/contrib/msc/pipeline/manager.py
create mode 100644 python/tvm/relax/backend/utils.py
create mode 100644 src/relax/distributed/transform/lower_distir.cc
create mode 100644
src/relax/distributed/transform/lower_global_view_to_local_view.cc
create mode 100644 src/relax/distributed/transform/utils.cc
create mode 100644 src/relax/distributed/transform/utils.h
copy src/relax/op/distributed/{op.cc => ccl.cc} (59%)
create mode 100644 src/relax/transform/expand_tuple_arguments.cc
create mode 100644 src/relax/transform/remove_unused_outputs.cc
create mode 100644 src/relax/transform/remove_unused_parameters.cc
create mode 100644 tests/python/contrib/test_msc/test_manager.py
create mode 100644 tests/python/contrib/test_msc/test_tools.py
create mode 100644
tests/python/relax/distributed/test_distributed_transform_lower_distir.py
create mode 100644
tests/python/relax/distributed/test_distributed_transform_lower_global_to_local_view.py
create mode 100644 tests/python/relax/test_transform_expand_tuple_args.py
create mode 100644 tests/python/relax/test_transform_normalize_global_var.py
create mode 100644 tests/python/relax/test_transform_remove_unused_outputs.py
create mode 100644
tests/python/relax/test_transform_remove_unused_parameters.py
