This is an automated email from the ASF dual-hosted git repository.
tqchen pushed a change to branch unity-staging
in repository https://gitbox.apache.org/repos/asf/tvm.git
omit 0f6c4674c1 [Unity][BYOC] Add pattern-based partitioning pass (#14054)
omit 41684ee567 [Unity][VM] Add per-op profiling support (#14053)
omit e9a6e49f3a [Unity][TVMScript] Overload `__neg__` for relax expr
(#14045)
omit 73a45e3022 [Unity][Pass] FuseOps FuseTIR fixes (#14044)
omit b3cf43af34 [Unity] Statement rewriter for DataflowBlock (#14043)
omit 82f7a5e049 [Unity] Relax dataflow pattern language (matching) (#14041)
omit 3a909e2b95 [Unity] Update tests to adapt to latest TVMScript syntax
(#14039)
omit 59527567cc [Unity] Disallow inline prim_func in relax IR (#14040)
omit a3ecb2f726 [Unity][Pass] Block-level static memory planning (#14038)
omit 405ef9dbf6 [Unity] Initial PyTorch Frontend (#14037)
omit a952bdb26b [Unity][Op] Add ShapeExpr Tests for Reshape Op (#14035)
omit 77604b464b [Unity][Pass] Operator legalization (#14029)
omit 2b4e4d930a [Unity][TVMScript] Move tir/relax import in script out of
__init__.py (#14033)
omit 29fe7a6c28 [Unity][Pass] Wellformed Analysis (#14032)
omit eec6b0f2f8 [Unity][BlockBuilder] CallTE convert PrimValue args
(#14028)
omit cf51272588 [Unity][Pass] Normalize Pass (#14031)
omit 18ebd15791 [Unity] Relay -> Relax translator (#14026)
omit 8dd819fcf2 [Unity][Pass][TuningAPI] Introduce TuningAPI and
MetaSchedule pass (#14014)
omit 3928e9b63c [Unity][Pass] BindParams pass, FoldConstant pass (#14016)
omit 55810737be [Unity][VM] Supporting "compiled" exec mode. (#14015)
omit f0e22c9b1a [Unity][Pass] LambdaLift pass (#14012)
omit 6fa4490d8c [Unity][Pass] Operator Fusion Passes (#14001)
omit de32194595 [Unity] NestedMsg Support utility (#13995)
omit 4e2cb4d468 [Unity] Relax op: manipulation (#13989)
omit aaac5c44bc [Unity] Relax op: search (#13992)
omit c3dfa324f9 [Unity] Relax op: linear algebra (#13988)
omit ae81980963 [Unity] Relax op: creation (#13984)
omit 7dd27f0758 [Unity] Relax op: neural networks (#13993)
omit 0063f899db [Unity] Relax op: statistical (#13991)
omit 05116eec7a [Unity] Relax op: arithmetic, comparison (#13983)
omit 0be100160f [Unity] Relax op: image (#13994)
omit a3d25f0a16 [Unity] Relax op: set (#13990)
omit 3e4af0dc9b [Unity] Relax op: datatype (#13986)
omit 9f13b7bd6c [Unity] Relax op: index (#13987)
omit de46e9c368 [Unity][TVMScript] Use explicit `R.shape` in TVMScript
(#13979)
omit c21a7ddfcc [Unity] e2e Relax minimum build flow (#13961)
omit ff1a02c66a [Unity] Relax VM shape lowering pass (#13956)
omit 8fb1c9c577 [Unity] Relax VM codegen (#13954)
omit 45a9bfd1df [Unity] Relax TVMScript Printer (#13944)
omit ed3653bfb2 [Unity] Relax TVMScript Parser. (#13932)
omit 14c057ba9e [Unity] Relax BlockBuilder and ExprMutator (#13926)
omit 9f24dac1ac [Unity] Basic StructInfo Analysis and Expr construction
(#13916)
omit 10622618ab [Unity][CI] Unity specific jenkins setup (do not upstream
to main) (#13910)
omit 746e2b58a8 [Unity][IR] First-class StructInfo (#13907)
omit 666b839d42 [Unity] Relax expressions and types (#13901)
omit 6bb0c05943 [Unity] Relax VM (#13878)
add f80771f556 [Frontend][Paddle] Add norm and one_hot_v2 op (#14049)
add 5ec33bb558 [TOPI] Support non-batch cases for topi.nll_loss (#14060)
add 347d79cbcf [microTVM][CRT]Move Makefile to CMake to be cross-platform
compatible (#14013)
add 49e098971f [APP] Update android_rpc build tools version (#14052)
add b59da5904c [Target] Add target tags for Apple Silicon GPU (#14068)
add fbd641c92b [TOPI] Bugfix arm_cpu schedule_conv2d_spatial_pack_nhwc
schedule (#14003)
add 19d970cab1 [MetaSchedule] Fix anchor-block flow with empty design
space generator (#14047)
add a06c36c5b6 [CI] Fix test skipping pytest attribute (#14064)
add 0a269defe1 [Codegen][CUDA] Add error message for missing fragment info
(#14073)
add bd26c24c69 [MetaSchedule] MutateTileSize skip single-candidate
SampleCategorical (#14072)
add 6191711a16 [COMMUNITY] mkatanbaf -> Reviewer (#14085)
add 330fe8d17c [CRT] USE CMake for CRT standalone libraries (#14025)
add 5f610f45a0 [CI] Update Compute Library to v22.11 (#14084)
add 3cdb079407 [TORCH] scatter_reduce implementation (#14018)
add 1e988a4a6f [WEB] Reduce memleak in web runtime (#14086)
new 2f9650f648 [Unity] Relax VM (#13878)
new 70dfbb3783 [Unity] Relax expressions and types (#13901)
new a485a1dcad [Unity][IR] First-class StructInfo (#13907)
new 071c9de0ea [Unity][CI] Unity specific jenkins setup (do not upstream
to main) (#13910)
new 9449f7fc24 [Unity] Basic StructInfo Analysis and Expr construction
(#13916)
new b758112b5b [Unity] Relax BlockBuilder and ExprMutator (#13926)
new 79de7f4995 [Unity] Relax TVMScript Parser. (#13932)
new 6b8b04243a [Unity] Relax TVMScript Printer (#13944)
new 27ac65fd84 [Unity] Relax VM codegen (#13954)
new f4f0057500 [Unity] Relax VM shape lowering pass (#13956)
new 2ac2d04f38 [Unity] e2e Relax minimum build flow (#13961)
new 36e41e1211 [Unity][TVMScript] Use explicit `R.shape` in TVMScript
(#13979)
new e0f59c7f75 [Unity] Relax op: index (#13987)
new 4d12d68943 [Unity] Relax op: datatype (#13986)
new 20b3436e8a [Unity] Relax op: set (#13990)
new 18d0dc23de [Unity] Relax op: image (#13994)
new 1cab6af7f4 [Unity] Relax op: arithmetic, comparison (#13983)
new cf5468bc24 [Unity] Relax op: statistical (#13991)
new 95f609df68 [Unity] Relax op: neural networks (#13993)
new 070e1ae6c4 [Unity] Relax op: creation (#13984)
new 6c64a29c61 [Unity] Relax op: linear algebra (#13988)
new eb7b829230 [Unity] Relax op: search (#13992)
new 37bb48acca [Unity] Relax op: manipulation (#13989)
new a7c5975be8 [Unity] NestedMsg Support utility (#13995)
new 59dfc38765 [Unity][Pass] Operator Fusion Passes (#14001)
new 5a63ff7e85 [Unity][Pass] LambdaLift pass (#14012)
new 122521b68c [Unity][VM] Supporting "compiled" exec mode. (#14015)
new 0e948d84bd [Unity][Pass] BindParams pass, FoldConstant pass (#14016)
new e5b13d8982 [Unity][Pass][TuningAPI] Introduce TuningAPI and
MetaSchedule pass (#14014)
new 40fa1d39fe [Unity] Relay -> Relax translator (#14026)
new e997c676ce [Unity][Pass] Normalize Pass (#14031)
new 731292b6da [Unity][BlockBuilder] CallTE convert PrimValue args
(#14028)
new c251c4ba7c [Unity][Pass] Wellformed Analysis (#14032)
new 8c6437e92f [Unity][TVMScript] Move tir/relax import in script out of
__init__.py (#14033)
new 17bebf65ea [Unity][Pass] Operator legalization (#14029)
new add20d325f [Unity][Op] Add ShapeExpr Tests for Reshape Op (#14035)
new b78ae86324 [Unity] Initial PyTorch Frontend (#14037)
new c2b4ad1c8c [Unity][Pass] Block-level static memory planning (#14038)
new e9a57c562c [Unity] Disallow inline prim_func in relax IR (#14040)
new 970d2148ba [Unity] Update tests to adapt to latest TVMScript syntax
(#14039)
new 3ae748f3a9 [Unity] Relax dataflow pattern language (matching) (#14041)
new fe2110febc [Unity] Statement rewriter for DataflowBlock (#14043)
new 7f8d0e9bfb [Unity][Pass] FuseOps FuseTIR fixes (#14044)
new a32cde1fa0 [Unity][TVMScript] Overload `__neg__` for relax expr
(#14045)
new ee98cd4511 [Unity][VM] Add per-op profiling support (#14053)
new 511112b173 [Unity][BYOC] Add pattern-based partitioning pass (#14054)
new f93a50c2d0 [Unity] Relax op: collapse sum (#14059)
new dd6431e863 [Unity][Fix][Pass] Fix FuseOps for lack graph edges (#14058)
new bf2f8118df [Unity][Pass] Remove Unused Function (#14061)
new 46ebe80819 [Unity][BYOC] Add pass to merge composite functions to
offload large subgraphs (#14062)
new 77c763f251 [Unity][Frontend] Annotate number of non-static input of FX
function (#14067)
new 8f8c3f0474 [Unity][Transform] Add LiftTransformParams pass (#14069)
new 825bb8125d [Unity][BYOC][Pass] RunCodegen and TensorRT (#14078)
new ae4ed43d5a [Unity][Pass] Canonicalize Bindings (#14079)
new 068e52096a [Unity] Add testcases for `expr_args_converter` (#14080)
new 561bd88684 [Unity][BYOC] Add CUTLASS backend (#14081)
new 9607b3f974 [Unity][BYOC] Add DNNL backend (#14082)
new 7544cf6869 [Unity][Op] `log_softmax` and `cross_entropy_with_logits`
(#14083)
new dc4c75bb28 [Unity][Analysis] TIR pattern kind analysis for
multi-buffer write block (#14075)
new d0caea5986 [Unity][Fix][Pass] FoldConstant with DCE in dataflow block
(#14087)
new 6ea9afc408 [Unity] Refactor Relax Build JIT UX (#14088)
This update added new revisions after undoing existing revisions.
That is to say, some revisions that were in the old version of the
branch are not in the new version. This situation occurs
when a user --force pushes a change and generates a repository
containing something like this:
* -- * -- B -- O -- O -- O (0f6c4674c1)
\
N -- N -- N refs/heads/unity-staging (6ea9afc408)
You should already have received notification emails for all of the O
revisions, and so the following emails describe only the N revisions
from the common base, B.
Any revisions marked "omit" are not gone; other references still
refer to them. Any revisions marked "discard" are gone forever.
The 61 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:
CONTRIBUTORS.md | 1 +
apps/android_camera/README.md | 5 +-
apps/android_camera/app/build.gradle | 11 +-
.../app/src/main/AndroidManifest.xml | 3 +-
apps/android_camera/build.gradle | 4 +-
apps/android_deploy/README.md | 14 +-
apps/android_deploy/app/build.gradle | 25 +-
apps/android_deploy/app/download-models.gradle | 18 +
.../app/src/main/AndroidManifest.xml | 3 +-
apps/android_deploy/build.gradle | 4 +-
apps/android_rpc/README.md | 10 +-
apps/android_rpc/app/build.gradle | 24 +-
apps/android_rpc/app/src/main/AndroidManifest.xml | 3 +-
apps/android_rpc/build.gradle | 4 +-
apps/bundle_deploy/Makefile | 28 +-
apps/bundle_deploy/build_model.py | 6 -
.../template_project/CMakeLists.txt.template | 9 +-
.../zephyr/template_project/microtvm_api_server.py | 2 +-
apps/relax_examples/e2e_auto_tir.py | 2 +-
apps/relax_examples/mlp.py | 2 +-
apps/relax_examples/nn_module.py | 2 +-
apps/relax_examples/resnet.py | 2 +-
cmake/modules/Micro.cmake | 2 +-
cmake/modules/StandaloneCrt.cmake | 2 +-
cmake/modules/contrib/CUTLASS.cmake | 4 +-
cmake/modules/contrib/DNNL.cmake | 8 +-
cmake/modules/contrib/TensorRT.cmake | 2 +-
.../ubuntu_download_arm_compute_lib_binaries.sh | 2 +-
include/tvm/ir/module.h | 6 +
include/tvm/meta_schedule/database.h | 7 +-
include/tvm/relax/transform.h | 127 ++-
include/tvm/relax/utils.h | 11 +-
include/tvm/topi/nn.h | 28 +
python/tvm/contrib/cutlass/build.py | 193 +++-
python/tvm/contrib/cutlass/gemm_profiler.py | 4 +-
python/tvm/contrib/cutlass/gen_gemm.py | 26 +-
python/tvm/contrib/cutlass/gen_tensor_op.py | 4 +
python/tvm/ir/function.py | 26 +-
python/tvm/ir/module.py | 35 +
python/tvm/meta_schedule/database/database.py | 2 +-
python/tvm/meta_schedule/relax_integration.py | 6 +-
python/tvm/relax/__init__.py | 8 +-
python/tvm/relax/exec_builder.py | 2 +-
python/tvm/relax/frontend/torch/dynamo.py | 4 +-
python/tvm/relax/frontend/torch/fx_translator.py | 30 +-
python/tvm/relax/op/manipulate.py | 53 +
python/tvm/relax/op/nn/nn.py | 55 +
.../tvm/relax/transform/legalize_ops/manipulate.py | 5 +
python/tvm/relax/transform/legalize_ops/nn.py | 20 +
python/tvm/relax/transform/transform.py | 74 ++
.../transform/tuning_api/default_functions.py | 6 +-
python/tvm/relax/utils.py | 12 +-
python/tvm/relax/vm_build.py | 317 ++++++
python/tvm/relay/frontend/paddlepaddle.py | 28 +
python/tvm/relay/frontend/pytorch.py | 54 +-
python/tvm/{relax/vm.py => runtime/relax_vm.py} | 225 +----
python/tvm/script/ir_builder/relax/ir.py | 4 +
python/tvm/topi/arm_cpu/conv2d_spatial_pack.py | 76 +-
python/tvm/topi/reduction.py | 31 +
python/tvm/topi/scatter_elements.py | 4 +-
src/ir/function.cc | 18 +
src/ir/module.cc | 12 +
src/meta_schedule/database/database.cc | 15 +
src/meta_schedule/database/json_database.cc | 12 +-
src/meta_schedule/database/memory_database.cc | 13 +-
src/meta_schedule/mutator/mutate_tile_size.cc | 7 +
src/meta_schedule/trace_apply.cc | 4 +-
src/relax/analysis/tir_op_pattern_kind.cc | 14 +-
src/relax/analysis/udchain.cc | 5 +-
.../backend/contrib/codegen_json/codegen_json.h | 222 +++--
src/relax/backend/contrib/cutlass/codegen.cc | 274 +++++
src/relax/backend/contrib/dnnl/codegen.cc | 105 ++
src/relax/backend/contrib/tensorrt/codegen.cc | 267 +++++
src/relax/backend/contrib/utils.h | 127 +++
src/relax/ir/expr.cc | 24 +
src/relax/op/nn/nn.cc | 74 ++
src/relax/op/nn/nn.h | 6 +
src/relax/op/tensor/manipulate.cc | 130 +++
src/relax/op/tensor/manipulate.h | 21 +
src/relax/transform/canonicalize_bindings.cc | 135 +++
src/relax/transform/fold_constant.cc | 14 +-
src/relax/transform/fuse_ops.cc | 9 +-
src/relax/transform/lift_transform_params.cc | 297 ++++++
src/relax/transform/merge_composite_functions.cc | 355 +++++++
src/relax/transform/remove_unused_funcs.cc | 120 +++
src/relax/transform/run_codegen.cc | 190 ++++
src/relax/transform/utils.h | 122 +++
src/relax/utils.cc | 29 +
src/relay/backend/contrib/codegen_c/codegen_c.h | 35 +-
src/relay/backend/contrib/cutlass/codegen.h | 21 +
src/runtime/crt/CMakeLists.txt | 51 +
src/runtime/crt/Makefile | 85 --
src/runtime/crt/host/CMakeLists.txt.template | 59 ++
src/runtime/crt/host/Makefile.template | 87 --
src/runtime/crt/host/microtvm_api_server.py | 60 +-
src/target/source/codegen_cuda.cc | 4 +
src/target/tag.cc | 15 +
src/tir/schedule/utils.h | 8 +
src/topi/reduction.cc | 4 +
tests/lint/check_file_type.py | 2 +-
tests/micro/arduino/test_arduino_workflow.py | 2 +-
tests/python/frontend/paddlepaddle/test_forward.py | 26 +
tests/python/frontend/pytorch/test_forward.py | 26 +
tests/python/integration/test_auto_tensorize.py | 2 +-
tests/python/relax/test_codegen_cutlass.py | 427 ++++++++
tests/python/relax/test_codegen_dnnl.py | 120 +++
tests/python/relax/test_codegen_tensorrt.py | 124 +++
tests/python/relax/test_expr_args_converter.py | 146 +++
tests/python/relax/test_frontend_from_fx.py | 51 +-
tests/python/relax/test_op_manipulate.py | 326 ++++++
tests/python/relax/test_op_nn.py | 173 +++-
tests/python/relax/test_pipeline.py | 2 +-
tests/python/relax/test_relay_translator.py | 2 +-
.../test_transform_annotate_tir_op_pattern.py | 27 +
tests/python/relax/test_transform_bind_params.py | 4 +-
.../relax/test_transform_canonicalize_bindings.py | 224 +++++
tests/python/relax/test_transform_codegen_pass.py | 260 +++++
tests/python/relax/test_transform_fold_constant.py | 3 -
tests/python/relax/test_transform_fuse_ops.py | 124 ++-
.../test_transform_legalize_ops_manipulate.py | 103 ++
.../python/relax/test_transform_legalize_ops_nn.py | 268 +++++
.../relax/test_transform_lift_transform_params.py | 295 ++++++
.../test_transform_merge_composite_functions.py | 1051 ++++++++++++++++++++
.../relax/test_transform_remove_unused_funcs.py | 211 ++++
.../relax/test_tvmscript_parser_op_manipulate.py | 33 +
tests/python/relax/test_tvmscript_parser_op_nn.py | 33 +
tests/python/relax/test_utils.py | 107 ++
tests/python/relax/test_vm_build.py | 62 +-
tests/python/relax/test_vm_codegen_only.py | 10 +-
tests/python/relax/test_vm_codegen_tir.py | 2 +-
tests/python/relax/test_vm_profiler.py | 8 +-
tests/python/relay/test_vm.py | 10 +-
tests/python/topi/python/test_topi_loss.py | 11 +-
tests/python/topi/python/test_topi_reduce.py | 39 +
.../test_meta_schedule_mutator_mutate_tile_size.py | 13 +
.../test_meta_schedule_relay_integration.py | 14 +-
web/.eslintignore | 1 +
web/README.md | 2 +-
web/apps/node/example.js | 2 +
web/emcc/wasm_runtime.cc | 2 +-
web/src/ctypes.ts | 22 +
web/src/index.ts | 5 +-
web/src/rpc_server.ts | 19 +-
web/src/runtime.ts | 642 +++++++++---
web/tests/node/test_module_load.js | 15 +-
web/tests/node/test_ndarray.js | 16 +-
web/tests/node/{test_ndarray.js => test_object.js} | 35 +-
web/tests/node/test_packed_func.js | 59 +-
web/tests/python/websock_rpc_test.py | 1 -
149 files changed, 8913 insertions(+), 984 deletions(-)
create mode 100644 python/tvm/relax/vm_build.py
mode change 100644 => 100755 python/tvm/relay/frontend/paddlepaddle.py
rename python/tvm/{relax/vm.py => runtime/relax_vm.py} (73%)
copy src/{relay => relax}/backend/contrib/codegen_json/codegen_json.h (61%)
create mode 100644 src/relax/backend/contrib/cutlass/codegen.cc
create mode 100644 src/relax/backend/contrib/dnnl/codegen.cc
create mode 100644 src/relax/backend/contrib/tensorrt/codegen.cc
create mode 100644 src/relax/backend/contrib/utils.h
create mode 100644 src/relax/transform/canonicalize_bindings.cc
create mode 100644 src/relax/transform/lift_transform_params.cc
create mode 100644 src/relax/transform/merge_composite_functions.cc
create mode 100644 src/relax/transform/remove_unused_funcs.cc
create mode 100644 src/relax/transform/run_codegen.cc
create mode 100644 src/relax/transform/utils.h
create mode 100644 src/runtime/crt/CMakeLists.txt
delete mode 100644 src/runtime/crt/Makefile
create mode 100644 src/runtime/crt/host/CMakeLists.txt.template
delete mode 100644 src/runtime/crt/host/Makefile.template
mode change 100644 => 100755 tests/python/frontend/paddlepaddle/test_forward.py
create mode 100644 tests/python/relax/test_codegen_cutlass.py
create mode 100644 tests/python/relax/test_codegen_dnnl.py
create mode 100644 tests/python/relax/test_codegen_tensorrt.py
create mode 100644 tests/python/relax/test_expr_args_converter.py
create mode 100644 tests/python/relax/test_transform_canonicalize_bindings.py
create mode 100644 tests/python/relax/test_transform_codegen_pass.py
create mode 100644 tests/python/relax/test_transform_lift_transform_params.py
create mode 100644
tests/python/relax/test_transform_merge_composite_functions.py
create mode 100644 tests/python/relax/test_transform_remove_unused_funcs.py
create mode 100644 tests/python/relax/test_utils.py
copy web/tests/node/{test_ndarray.js => test_object.js} (64%)
