This is an automated email from the ASF dual-hosted git repository.
junrushao pushed a change to branch v0.8
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from e807743 [ETHOSU] Add early simplify to fix LoopPartition (#9387)
add 54019b9 [microTVM] Add microTVM Template Projects to tlcpack pip
Package (#9309)
add 36e216b [IR] Minor cleanup to tvm.ir.instrument.PassInstrument (#9392)
add 1a3dac7 [VMCompiler] Support shape func lowering for nested function
call (#9405)
add 56c4d07 BUG: Make sure FoldConstant can inline constants underneath
on_device annotations (#9367)
add c83dfa6 fix (#9412)
add d1aebcb [TVMC] Keep quantized weights when importing PyTorch model
(#9417)
add aaae58f [BYOC][NPU] Fix integration tests not running (#9415)
add 7cfcad9 [BYOC][ACL] Update installation docs (#9426)
add aeff3ea BUG: alloc_tensor offset and reshape shape should be on the
CPU (#9421)
add 95bab00 Bump the CMake version in ubuntu_install_cmake_source.sh to
3.14.7. (#9424)
add 5ed8e01 Support quantized ABS operator in TFLite frontend (#9411)
add 01f90e0 [Frontend][PaddlePaddle] Support more common operators (#9428)
add 5d0ce8e Fix GetQmin and GetQmax from relay.qnn (#9427)
add 054fff5 [DOC] Add tip on mitigation for symbol conflict with PyTorch
(#9433)
add 17e4ad6 change id (#9431)
add 32f829c add reviewer (#9430)
add 60b21b9 [CUTLASS] Initial support for dynamic shape dense (#9419)
add 90b946c [ONNX][Converter] Add dynamic nodes support (#9380)
add 8ada2b1 [TIR] Fix VerifyGPUCode for vectorized halfx8 store (#9420)
add bff9884 [Keras] Add l2_normalize support (#9383)
add a6c948a Adds SEScope (Storage/Execution Scope) for use as new unit of
planning in 'device' planning. (#9313)
add c99f55f [TVMScript] Use // and % for FloorDiv/FloorMod (#9437)
add e0c1256 Removed a manual file handler pitfall (#9435)
add 19b23b9 Arm(R) Ethos(TM)-U NPU Pooling operators support (#9384)
add c7a01a4 [CUTLASS] Support batch_matmul (#9439)
add b8fb438 Fix several typos in pytest_target_parameterization.rst
(#9447)
add 85fd85f [MetaSchedule] Task Extraction (#9382)
add e24c075 [MetaSchedule] Sample-Perfect-Tile (#9449)
add e3a3043 [CI] Pin setuptools to v58.4.0 in CI to circumvent breaking
change in v58.5 (#9446)
add 5527cbf Update TVM ci-cpu docker image to v.079 (#9454)
add 048994b [Bug][Meta Schedule] Fix Infinite Loop Caused When Calling
Methods Not Overrided In PyClass. (#9451)
add e7024fb [PyTorch]Add PyTorchTVM: compile torchscript to tvm and
export as pytorch_op (#8777)
add 4973cd3 [TIR][Schedule] Add get-child-blocks primitive (#9434)
add efb367a Fix typo. (#9462)
add 374e15b [TensorIR] Print TVMScript with prefix T instead of tir
(#9422)
add 7b58e16 [TensorIR] GetProducer, GetConsumer (#506) (#9464)
add 3a9f40c Fix an infinite recompilation loop in tvm-sys. (#9450)
add 5a4526d [TVMC] Re-enable PyTorch test (#9441)
add 01141d4 Support quantised SQRT operator in TFLite (#9258)
add a644e29 [Frontend][PaddlePaddle] Add operators of interploate/flatten
and modify try_infer_value (#9459)
add 811312c [TIR] Make compact buffer and get access region aware of
conditions (#9372)
add bc505fc add is_entry_func tag for device function (#9436)
add 6aa5ba2 Add back-to-back conv2d Hexagon test for stripe scheduling
(#9390)
add 028f4fa [Relay] Non-recursive Dtor for Let (#9461)
add 7ced071 [Frontend][TFlite] Cast MirrorPad paddings to int32 (#9468)
add 7c3283c Fix function annotation (#9474)
add b86aedd [CMSIS-NN] Convert CMSIS-NN to use Target Hooks (#9397)
add 5147442 [TVMC] Add test for quantized pytorch model (#9467)
add 00b0a81 [TIR] Add type hint for TIR (#9432)
No new revisions were added by this update.
Summary of changes:
.github/CODEOWNERS | 20 +-
.gitignore | 3 +-
CMakeLists.txt | 6 +
CONTRIBUTORS.md | 5 +-
Jenkinsfile | 6 +-
.../template_project/microtvm_api_server.py | 9 +-
apps/{tf_tvmdsoop => pt_tvmdsoop}/CMakeLists.txt | 10 +-
.../prepare_and_test_pt_tvm_class.sh} | 19 +-
apps/pt_tvmdsoop/tests/test_torch_compile_cpu.py | 68 ++
apps/pt_tvmdsoop/tests/test_torch_compile_gpu.py | 63 ++
apps/pt_tvmdsoop/tests/test_torch_graph_module.py | 129 +++
apps/pt_tvmdsoop/tests/test_torch_script.py | 116 +++
apps/pt_tvmdsoop/tests/test_torch_vm_module.py | 122 +++
apps/pt_tvmdsoop/tests/test_trace_tvm_module.py | 58 ++
cmake/config.cmake | 3 +
cmake/modules/Arduino.cmake | 78 ++
cmake/modules/LibInfo.cmake | 1 +
cmake/modules/StandaloneCrt.cmake | 15 +-
cmake/modules/Zephyr.cmake | 78 ++
cmake/modules/contrib/PT_TVMDSOOP.cmake | 59 ++
cmake/utils/Utils.cmake | 13 +
docker/install/ubuntu1804_install_python.sh | 4 +-
docker/install/ubuntu1804_install_python_venv.sh | 4 +-
docker/install/ubuntu_install_cmake_source.sh | 4 +-
docker/install/ubuntu_install_python.sh | 4 +-
docker/install/ubuntu_install_vela.sh | 1 -
docs/dev/how_to/pytest_target_parametrization.rst | 10 +-
docs/how_to/deploy/arm_compute_lib.rst | 27 +-
docs/install/from_source.rst | 3 +
.../how_to/work_with_microtvm/micro_autotune.py | 15 +-
gallery/how_to/work_with_microtvm/micro_tflite.py | 7 +-
include/tvm/arith/int_set.h | 2 +-
include/tvm/ir/attrs.h | 4 +-
include/tvm/meta_schedule/builder.h | 1 +
include/tvm/meta_schedule/database.h | 23 +-
include/tvm/meta_schedule/integration.h | 214 +++++
include/tvm/meta_schedule/runner.h | 5 +-
include/tvm/meta_schedule/search_strategy.h | 12 +-
include/tvm/meta_schedule/space_generator.h | 4 +
include/tvm/meta_schedule/task_scheduler.h | 68 +-
include/tvm/relay/expr.h | 50 ++
include/tvm/runtime/vm/executable.h | 7 +
include/tvm/runtime/vm/vm.h | 4 +-
include/tvm/target/compilation_config.h | 170 ++++
include/tvm/target/se_scope.h | 349 ++++++++
include/tvm/target/target.h | 18 +-
include/tvm/tir/function.h | 7 +
include/tvm/tir/schedule/schedule.h | 34 +
python/setup.py | 7 +
python/tvm/contrib/cutlass/__init__.py | 2 +-
python/tvm/contrib/cutlass/build.py | 239 +++--
python/tvm/contrib/cutlass/gemm_operation.py | 15 +-
python/tvm/contrib/cutlass/gen_gemm.py | 69 +-
python/tvm/contrib/cutlass/library.py | 2 +
python/tvm/contrib/target/onnx.py | 10 +-
python/tvm/contrib/torch/__init__.py | 51 ++
python/tvm/contrib/torch/module.py | 121 +++
python/tvm/contrib/torch/pytorch_tvm.py | 249 ++++++
python/tvm/driver/tvmc/common.py | 4 +-
python/tvm/driver/tvmc/frontends.py | 4 +-
python/tvm/driver/tvmc/target.py | 14 +-
python/tvm/ir/instrument.py | 158 +++-
python/tvm/ir/module.py | 2 +-
python/tvm/meta_schedule/__init__.py | 1 +
python/tvm/meta_schedule/builder/builder.py | 5 +-
python/tvm/meta_schedule/database/database.py | 18 +-
python/tvm/meta_schedule/integration.py | 250 ++++++
python/tvm/meta_schedule/runner/runner.py | 5 +-
.../search_strategy/search_strategy.py | 27 +-
.../space_generator/space_generator.py | 10 +-
.../meta_schedule/task_scheduler/task_scheduler.py | 122 ++-
.../_ctypes => meta_schedule/testing}/__init__.py | 4 +-
.../{testing.py => testing/local_rpc.py} | 2 +-
python/tvm/meta_schedule/testing/relay_workload.py | 88 ++
python/tvm/meta_schedule/utils.py | 40 +
python/tvm/micro/__init__.py | 1 +
python/tvm/micro/build.py | 46 +
.../tvm/relay/backend/contrib/ethosu/legalize.py | 136 ++-
.../relay/backend/contrib/ethosu/op/__init__.py | 1 +
.../relay/backend/contrib/ethosu/op/convolution.py | 6 +-
.../relay/backend/contrib/ethosu/op/depthwise.py | 8 +-
.../contrib/ethosu/op/{depthwise.py => pooling.py} | 109 +--
.../relay/backend/contrib/ethosu/te/__init__.py | 1 +
.../relay/backend/contrib/ethosu/te/convolution.py | 2 +-
.../relay/backend/contrib/ethosu/te/depthwise.py | 2 +-
.../contrib/ethosu/te/{depthwise.py => pooling.py} | 90 +-
.../relay/backend/contrib/ethosu/tir/compiler.py | 2 +-
.../tvm/relay/backend/contrib/ethosu/tir/passes.py | 2 +
.../ethosu/tir/{depthwise.py => pooling.py} | 62 +-
.../backend/contrib/ethosu/tir_to_cs_translator.py | 112 ++-
python/tvm/relay/backend/executor_factory.py | 8 +-
python/tvm/relay/backend/vm.py | 12 +-
python/tvm/relay/frontend/common.py | 5 +-
python/tvm/relay/frontend/keras.py | 104 ++-
python/tvm/relay/frontend/paddlepaddle.py | 425 ++++++++-
python/tvm/relay/frontend/tflite.py | 6 +-
python/tvm/relay/op/contrib/cmsisnn.py | 15 +-
python/tvm/relay/op/contrib/cutlass.py | 11 +-
python/tvm/relay/op/contrib/ethosu.py | 139 ++-
python/tvm/runtime/vm.py | 7 +
python/tvm/script/tir/__init__.pyi | 359 ++++++++
python/tvm/script/tir/intrin.py | 5 +
python/tvm/script/tir/ty.py | 11 +-
python/tvm/target/__init__.py | 2 +
.../bound.py => target/compilation_config.py} | 28 +-
.../tvm/target/se_scope.py | 6 +-
python/tvm/te/__init__.py | 2 +-
python/tvm/te/operation.py | 26 +-
python/tvm/tir/function.py | 2 +-
python/tvm/tir/schedule/schedule.py | 78 ++
python/tvm/tir/schedule/testing.py | 9 +-
rust/tvm-sys/build.rs | 3 +-
src/arith/int_set.cc | 2 +
src/contrib/torch/pt_call_tvm/tvm_class.cc | 686 ++++++++++++++
src/contrib/torch/utils.h | 264 ++++++
src/driver/driver_api.cc | 5 +-
src/ir/module.cc | 9 +-
src/meta_schedule/integration.cc | 151 ++++
src/meta_schedule/task_scheduler/round_robin.cc | 4 +-
src/meta_schedule/task_scheduler/task_scheduler.cc | 51 +-
src/node/serialization.cc | 34 +-
src/parser/parser.cc | 4 +-
src/printer/relay_text_printer.cc | 4 +-
src/printer/tvmscript_printer.cc | 35 +-
src/relay/backend/aot_executor_codegen.cc | 4 +-
src/relay/backend/build_module.cc | 11 +-
.../backend/contrib/cmsisnn/codegen_cmsisnn.cc | 51 --
src/relay/backend/contrib/cmsisnn/relay_to_tir.cc | 127 +--
.../{example_target_hooks => cmsisnn}/target.cc | 15 +-
.../backend/contrib/cmsisnn/tir_to_runtime.cc | 104 +--
src/relay/backend/contrib/codegen_c/codegen_c.h | 11 +-
src/relay/backend/contrib/cutlass/codegen.cc | 219 +++--
src/relay/backend/graph_executor_codegen.cc | 2 +-
src/relay/backend/te_compiler.cc | 56 +-
src/relay/backend/te_compiler.h | 2 +-
src/relay/backend/te_compiler_cache.cc | 70 +-
src/relay/backend/te_compiler_cache.h | 11 +-
src/relay/backend/utils.h | 11 +-
src/relay/backend/vm/compiler.cc | 41 +-
src/relay/backend/vm/compiler.h | 7 +-
src/relay/backend/vm/inline_primitives.cc | 2 +-
src/relay/ir/expr.cc | 33 +-
src/relay/op/annotation/annotation.cc | 10 +
src/relay/op/contrib/ethosu/convolution.cc | 6 +-
src/relay/op/contrib/ethosu/depthwise.cc | 4 +-
src/relay/op/contrib/ethosu/pooling.cc | 195 ++++
src/relay/qnn/utils.h | 6 +-
src/relay/transforms/device_aware_visitors.cc | 2 +-
src/relay/transforms/device_planner.cc | 2 +-
src/relay/transforms/fold_constant.cc | 398 +++++----
src/relay/transforms/memory_alloc.cc | 16 +-
src/relay/transforms/partition_graph.cc | 6 +-
src/runtime/cuda/cuda_device_api.cc | 10 +-
src/runtime/vm/executable.cc | 32 +-
src/runtime/vm/memory_manager.cc | 8 +-
src/runtime/vm/pooled_allocator.h | 6 +-
src/runtime/vm/serialize_utils.h | 4 +-
src/runtime/vm/vm.cc | 36 +-
src/target/compilation_config.cc | 229 +++++
src/target/se_scope.cc | 192 ++++
src/target/target.cc | 15 +-
src/te/operation/create_primfunc.cc | 38 +-
src/tir/analysis/block_access_region_detector.cc | 34 +
src/tir/analysis/verify_gpu_code.cc | 8 +-
src/tir/schedule/analysis/analysis.cc | 4 +-
src/tir/schedule/concrete_schedule.cc | 42 +
src/tir/schedule/concrete_schedule.h | 30 +-
src/tir/schedule/error.cc | 2 +-
src/tir/schedule/primitive.h | 62 +-
src/tir/schedule/primitive/get_block_loop.cc | 132 +++
src/tir/schedule/primitive/sampling.cc | 315 ++++++-
src/tir/schedule/schedule.cc | 18 +
src/tir/schedule/traced_schedule.cc | 60 ++
src/tir/schedule/traced_schedule.h | 15 +-
src/tir/schedule/utils.h | 42 +-
src/tir/transforms/compact_buffer_region.cc | 234 +++--
src/tir/transforms/ir_utils.cc | 84 ++
src/tir/transforms/ir_utils.h | 39 +
src/tir/transforms/split_host_device.cc | 1 +
tests/cpp/relay_dismantler_test.cc | 19 +
tests/cpp/target/compilation_config_test.cc | 184 ++++
tests/cpp/target/se_scope_test.cc | 119 +++
tests/lint/pylint.sh | 1 +
tests/micro/arduino/conftest.py | 14 +-
tests/micro/zephyr/test_utils.py | 11 +-
.../python/contrib/test_cmsisnn/test_binary_ops.py | 2 +-
tests/python/contrib/test_cmsisnn/test_softmax.py | 2 +-
tests/python/contrib/test_cutlass.py | 212 ++++-
tests/python/contrib/test_ethosn/infrastructure.py | 16 +-
tests/python/contrib/test_ethosn/test_networks.py | 16 +-
tests/python/contrib/test_ethosu/infra.py | 49 +-
tests/python/contrib/test_ethosu/test_codegen.py | 89 ++
tests/python/contrib/test_ethosu/test_legalize.py | 101 ++-
.../contrib/test_ethosu/test_replace_pooling.py | 158 ++++
.../test_ethosu/test_tir_to_cs_translator.py | 79 +-
.../contrib/test_ethosu/test_type_inference.py | 91 +-
tests/python/contrib/test_hexagon/README.md | 486 +---------
.../{README.md => test_conv2d_blocked.md} | 23 +-
.../contrib/test_hexagon/test_conv2d_conv2d.md | 986 +++++++++++++++++++++
.../contrib/test_hexagon/test_conv2d_conv2d.py | 341 +++++++
tests/python/contrib/test_onnx.py | 34 +-
tests/python/contrib/test_tensorrt.py | 5 +-
tests/python/driver/tvmc/conftest.py | 17 +
tests/python/driver/tvmc/test_compiler.py | 10 +-
tests/python/driver/tvmc/test_frontends.py | 16 +-
tests/python/driver/tvmc/test_target.py | 13 +
tests/python/driver/tvmc/test_target_options.py | 10 +
tests/python/frontend/keras/test_forward.py | 21 +
tests/python/frontend/paddlepaddle/test_forward.py | 339 ++++++-
tests/python/frontend/tflite/test_forward.py | 145 ++-
tests/python/relay/test_ir_nodes.py | 12 +
tests/python/relay/test_pass_fold_constant.py | 100 ++-
tests/python/relay/test_pass_instrument.py | 18 +-
tests/python/relay/test_pass_partition_graph.py | 43 +
tests/python/relay/test_prng.py | 7 +-
tests/python/relay/test_vm.py | 87 ++
tests/python/target/test_se_scope.py | 52 ++
.../unittest/test_meta_schedule_integration.py | 120 +++
.../unittest/test_meta_schedule_space_generator.py | 8 +-
.../unittest/test_meta_schedule_task_scheduler.py | 75 +-
.../python/unittest/test_target_codegen_vulkan.py | 5 +-
.../test_tir_analysis_get_block_access_region.py | 47 +
.../unittest/test_tir_analysis_verify_gpu_code.py | 29 +
.../python/unittest/test_tir_schedule_sampling.py | 43 +-
.../python/unittest/test_tir_schedule_utilities.py | 59 +-
.../test_tir_transform_compact_buffer_region.py | 136 +++
.../test_tir_transform_convert_blocks_to_opaque.py | 12 +-
.../test_tir_transform_split_host_device.py | 52 ++
.../python/unittest/test_tvmscript_error_report.py | 16 +-
tests/python/unittest/test_tvmscript_roundtrip.py | 21 +
tests/python/unittest/test_tvmscript_type.py | 86 ++
tests/scripts/task_mypy.sh | 4 +
tests/scripts/task_python_ethosn_tests.sh | 2 +-
tests/scripts/task_python_integration.sh | 4 +-
version.py | 27 +-
235 files changed, 12318 insertions(+), 2013 deletions(-)
copy apps/{tf_tvmdsoop => pt_tvmdsoop}/CMakeLists.txt (84%)
copy apps/{tf_tvmdsoop/prepare_and_test_tfop_module.sh =>
pt_tvmdsoop/prepare_and_test_pt_tvm_class.sh} (72%)
mode change 100644 => 100755
create mode 100644 apps/pt_tvmdsoop/tests/test_torch_compile_cpu.py
create mode 100644 apps/pt_tvmdsoop/tests/test_torch_compile_gpu.py
create mode 100644 apps/pt_tvmdsoop/tests/test_torch_graph_module.py
create mode 100644 apps/pt_tvmdsoop/tests/test_torch_script.py
create mode 100644 apps/pt_tvmdsoop/tests/test_torch_vm_module.py
create mode 100644 apps/pt_tvmdsoop/tests/test_trace_tvm_module.py
create mode 100644 cmake/modules/Arduino.cmake
create mode 100644 cmake/modules/Zephyr.cmake
create mode 100644 cmake/modules/contrib/PT_TVMDSOOP.cmake
create mode 100644 include/tvm/meta_schedule/integration.h
create mode 100644 include/tvm/target/compilation_config.h
create mode 100644 include/tvm/target/se_scope.h
create mode 100644 python/tvm/contrib/torch/__init__.py
create mode 100644 python/tvm/contrib/torch/module.py
create mode 100644 python/tvm/contrib/torch/pytorch_tvm.py
create mode 100644 python/tvm/meta_schedule/integration.py
copy python/tvm/{_ffi/_ctypes => meta_schedule/testing}/__init__.py (87%)
rename python/tvm/meta_schedule/{testing.py => testing/local_rpc.py} (97%)
create mode 100644 python/tvm/meta_schedule/testing/relay_workload.py
copy python/tvm/relay/backend/contrib/ethosu/op/{depthwise.py => pooling.py}
(58%)
copy python/tvm/relay/backend/contrib/ethosu/te/{depthwise.py => pooling.py}
(59%)
copy python/tvm/relay/backend/contrib/ethosu/tir/{depthwise.py => pooling.py}
(61%)
create mode 100644 python/tvm/script/tir/__init__.pyi
copy python/tvm/{arith/bound.py => target/compilation_config.py} (58%)
copy cmake/modules/contrib/CODEGENC.cmake => python/tvm/target/se_scope.py
(78%)
create mode 100644 src/contrib/torch/pt_call_tvm/tvm_class.cc
create mode 100644 src/contrib/torch/utils.h
create mode 100644 src/meta_schedule/integration.cc
delete mode 100644 src/relay/backend/contrib/cmsisnn/codegen_cmsisnn.cc
copy src/relay/backend/contrib/{example_target_hooks => cmsisnn}/target.cc
(77%)
create mode 100644 src/relay/op/contrib/ethosu/pooling.cc
create mode 100644 src/target/compilation_config.cc
create mode 100644 src/target/se_scope.cc
create mode 100644 tests/cpp/target/compilation_config_test.cc
create mode 100644 tests/cpp/target/se_scope_test.cc
create mode 100644 tests/python/contrib/test_ethosu/test_replace_pooling.py
copy tests/python/contrib/test_hexagon/{README.md => test_conv2d_blocked.md}
(95%)
create mode 100644 tests/python/contrib/test_hexagon/test_conv2d_conv2d.md
create mode 100644 tests/python/contrib/test_hexagon/test_conv2d_conv2d.py
create mode 100644 tests/python/target/test_se_scope.py
create mode 100644 tests/python/unittest/test_meta_schedule_integration.py
create mode 100644
tests/python/unittest/test_tir_transform_split_host_device.py
create mode 100644 tests/python/unittest/test_tvmscript_type.py
