llvm
/
torch-mlir
mirror of https://github.com/llvm/torch-mlir
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
torch-mlir/python/CMakeLists.txt

160 lines
6.2 KiB
CMake
Raw Normal View History

[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
include(AddMLIRPython)
Set some wheel building optimization options. * Also adds a requirements.txt and updates docs to reference it versus stringy pip install. * Adds doc with instructions on creating a wheel. Fixes #370
2021-10-22 12:18:08 +08:00
# Disables generation of "version soname" (i.e. libFoo.so.<version>), which
# causes pure duplication as part of Python wheels.
set(CMAKE_PLATFORM_NO_VERSIONED_SONAME ON)
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
# The directory at which the Python import tree begins.
# See documentation for `declare_mlir_python_sources`'s ROOT_DIR
# argument.
set(TORCH_MLIR_PYTHON_ROOT_DIR "${CMAKE_CURRENT_SOURCE_DIR}/torch_mlir")
Port the bulk of the remaining code to torch-mlir This leaves no real code outside torch-mlir. This also renames the "npcomp backend contract" to "linalg on tensors backend contract" as the name of the abstraction layer that RefBackend (IREE too) accepts.
2021-09-28 02:36:44 +08:00
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
# We vendor our own MLIR instance in the `torch_mlir` namespace.
add_compile_definitions("MLIR_PYTHON_PACKAGE_PREFIX=torch_mlir.")
[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
################################################################################
# Sources
################################################################################
declare_mlir_python_sources(TorchMLIRPythonSources)
declare_mlir_python_sources(TorchMLIRPythonExtensions)
Add a new `torch_mlir.compile` method. This makes it much easier to convert models and hides all the ClassAnnotator complexity. This also adds a new example `torchscript_resnet18_all_output_types.py` which shows the ResNet18 IR for all output types. Also, - This moves `run_pipeline_with_repro_report` to `torch_mlir.compiler_utils`.
2022-04-20 08:30:09 +08:00
declare_mlir_python_sources(TorchMLIRPythonSources.TopLevel
ROOT_DIR "${TORCH_MLIR_PYTHON_ROOT_DIR}"
ADD_TO_PARENT TorchMLIRPythonSources
SOURCES
__init__.py
compiler_utils.py
)
[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
declare_mlir_python_sources(TorchMLIRPythonSources.Dialects
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
ROOT_DIR "${TORCH_MLIR_PYTHON_ROOT_DIR}"
[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
ADD_TO_PARENT TorchMLIRPythonSources
)
declare_mlir_dialect_python_bindings(
ADD_TO_PARENT TorchMLIRPythonSources.Dialects
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
ROOT_DIR "${TORCH_MLIR_PYTHON_ROOT_DIR}"
[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
TD_FILE dialects/TorchBinding.td
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
SOURCES dialects/torch/__init__.py
[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
DIALECT_NAME torch
)
################################################################################
# Extensions
################################################################################
declare_mlir_python_extension(TorchMLIRPythonExtensions.Main
MODULE_NAME _torchMlir
ADD_TO_PARENT TorchMLIRPythonExtensions
SOURCES
Enable python extensions when building out of tree (#363)
2021-10-28 08:04:12 +08:00
TorchMLIRModule.cpp
[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
EMBED_CAPI_LINK_LIBS
TorchMLIRCAPI
PRIVATE_LINK_LIBS
LLVMSupport
)
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
################################################################################
# Optionally handle JIT IR importer.
################################################################################
Enable libtorch caching and source builds (#1004) Add an option to cache libtorch/ releases if you don't want to download the latest. Add an option to enable source builds. TESTS: macOS: verify with / without cache downloads verify source builds -- shared and static Linux: Build Tests and Release builds
2022-07-06 01:25:43 +08:00
option(LIBTORCH_SRC_BUILD "Build libtorch from source" OFF)
option(LIBTORCH_CACHE "Cache libtorch downloads" OFF)
option(LIBTORCH_VARIANT "libtorch variant [shared|static]-[cxx11]" "libtorch-shared")
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
if(TORCH_MLIR_ENABLE_JIT_IR_IMPORTER)
Enable libtorch caching and source builds (#1004) Add an option to cache libtorch/ releases if you don't want to download the latest. Add an option to enable source builds. TESTS: macOS: verify with / without cache downloads verify source builds -- shared and static Linux: Build Tests and Release builds
2022-07-06 01:25:43 +08:00
# if LIBTORCH_SRC_BUILD=OFF (default) we link against a binary libtorch :
# if LIBTORCH_CACHE=OFF (default) the build downloads the latest released version of libtorch
# if LIBTORCH_CACHE=ON the build uses an available libtorch/ in the root directory.
# In either case if there is no libtorch/ the latest released version for the platform is downloaded
# A developer could unpack their own libtorch/ and it would be respected
# if LIBTORCH_SRC_BUILD=ON we build libtorch/pytorch and link against it
# if LIBTORCH_VARIANT=*cxx11abi* we build with the new CXX11 ABI similar to upstream pytorch/builder
# if LIBTORCH_VARIANT=*shared* (default) we build a shared library (and pytorch .whl)
# else if LIBTORCH_VARIANT=*static* we build a static libtorch
Speed up libtorch build. (#1031)
2022-07-12 09:46:49 +08:00
set(ENV{LIBTORCH_SRC_BUILD} ${LIBTORCH_SRC_BUILD})
set(ENV{LIBTORCH_CACHE} ${LIBTORCH_CACHE})
set(ENV{LIBTORCH_VARIANT} ${LIBTORCH_VARIANT})
set(ENV{CMAKE_OSX_ARCHITECTURES} ${CMAKE_OSX_ARCHITECTURES})
set(ENV{CMAKE_C_COMPILER_LAUNCHER} ${CMAKE_C_COMPILER_LAUNCHER})
set(ENV{CMAKE_CXX_COMPILER_LAUNCHER} ${CMAKE_CXX_COMPILER_LAUNCHER})
Link against libtorch (#955) This moves torch-mlir to link against libtorch on macOS and linux TESTS: Tests pass. Tested release builds on linux and macOS
2022-07-01 03:40:17 +08:00
execute_process(
Speed up libtorch build. (#1031)
2022-07-12 09:46:49 +08:00
COMMAND ${CMAKE_CURRENT_SOURCE_DIR}/../build_tools/build_libtorch.sh
RESULT_VARIABLE _result
Link against libtorch (#955) This moves torch-mlir to link against libtorch on macOS and linux TESTS: Tests pass. Tested release builds on linux and macOS
2022-07-01 03:40:17 +08:00
)
build: improve robustness of cmake and shell scripts (#1018) On my local machine, `unzip` didn't exist (producing a "command not found" error), but CMake ignored the error. Although the build did succeed (because it found a previously-built version of libtorch), it seems better to abort builds on such failures, so this patch checks the return code of all external process invocations. Along similar lines, this patch also updates the shell scripts in `build_tools` to extensively use double-quoting to prevent unintentional word splitting or globbing. Since some of the scripts execute `rm` while using shell variables, this patch also adds the preamble `set -u` to abort execution if an undefined variable is referenced, so that we reduce the chances of executing `rm -rf /` if the path expression happens to refer to an undefined variable.
2022-07-07 05:39:30 +08:00
if(_result)
message(FATAL_ERROR "Failed to run `build_libtorch.sh`")
endif()
Link against libtorch (#955) This moves torch-mlir to link against libtorch on macOS and linux TESTS: Tests pass. Tested release builds on linux and macOS
2022-07-01 03:40:17 +08:00
set(TORCH_INSTALL_PREFIX "libtorch")
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
add_subdirectory(torch_mlir/dialects/torch/importer/jit_ir)
Port the bulk of the remaining code to torch-mlir This leaves no real code outside torch-mlir. This also renames the "npcomp backend contract" to "linalg on tensors backend contract" as the name of the abstraction layer that RefBackend (IREE too) accepts.
2021-09-28 02:36:44 +08:00
add_subdirectory(torch_mlir_e2e_test)
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
endif()
This PR implements an eager mode backend for PyTorch through the torch-mlir framework. This is accomplished by overriding the `__torch_dispatch__` class method on wrapper subclass `TorchMLIRTensor(torch.Tensor)`. Effectively, this mode works by compiling op by op as the NN is eagerly executed by PyTorch. Entailed in that compilation is building a representation of the op that can be `torch.jit.script`ed, importing using `ModuleBuilder`, and then executing (e.g., with `RefBackendLinalgOnTensorsBackend`). This mode includes a fallback to conventional PyTorch if anything in the torch-mlir compilation process fails (e.g., unsupported op). Currently, all e2e tests pass execpt for two that involve an upstream PyTorch bug (https://github.com/pytorch/pytorch/issues/74400). High priority next steps: 1. A compile cache in order to speed up reruns of the same NN. 2. Integration with IREE (though not in this repo). 3. Integration with `torch.distributed`.
2022-03-03 07:02:18 +08:00
################################################################################
# Eager mode
################################################################################
add_subdirectory(torch_mlir/eager_mode)
Allow torch-mlir to support PyTorch extensions. (#895) PyTorch allows new operators to be registered dynamically in modules. Torch-mlir already makes it fairly straightforward to add support for new operators, and this commit just extends that support to allow new PyTorch ops to come from a external module. This does *not* allow ops to be dynamically loaded into torch-mlir. Torch-mlir must still be compiled with support built-in. Add a `_torch_mlir_custom_op_example` subpackage to `torch_mlir` which registers an demonstration op. It will not be imported by default when importing torch_mlir. It's strictly for testing and documentation. Adds an end-to-end test for the `torch_mlir_custom_op_example::identity` op. With all these changes, we should now be actively testing PyTorch extension support with all future patches.
2022-06-14 05:51:30 +08:00
################################################################################
# Custom op example
# Required for running the update_torch_ods.sh and update_shape_lib.sh scripts.
################################################################################
Update CMakeLists.txt Emergency fix to unblock the nightly Release builder
2022-06-15 05:38:35 +08:00
# TODO: renable once it build on macOS Intel / M1
#add_subdirectory(torch_mlir/_torch_mlir_custom_op_example)
Allow torch-mlir to support PyTorch extensions. (#895) PyTorch allows new operators to be registered dynamically in modules. Torch-mlir already makes it fairly straightforward to add support for new operators, and this commit just extends that support to allow new PyTorch ops to come from a external module. This does *not* allow ops to be dynamically loaded into torch-mlir. Torch-mlir must still be compiled with support built-in. Add a `_torch_mlir_custom_op_example` subpackage to `torch_mlir` which registers an demonstration op. It will not be imported by default when importing torch_mlir. It's strictly for testing and documentation. Adds an end-to-end test for the `torch_mlir_custom_op_example::identity` op. With all these changes, we should now be actively testing PyTorch extension support with all future patches.
2022-06-14 05:51:30 +08:00
[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
################################################################################
# Generate packages and shared library
# Downstreams typically will not use these, but they are useful for local
# testing.
################################################################################
set(_source_components
# TODO: Core is now implicitly building/registering all dialects, increasing
# build burden by ~5x. Make it stop.
Bring up new RefBackend. `tools/torchscript_e2e_test.sh` is all green. This needs a few passes I put into torch-mlir/lib/RefBackend (not to be confused with `npcomp/lib/RefBackend`, which will soon be deleted). For the sake of review, since this brings together a lot of things, I split this into its own commit. I temporarily commented out some "list" stuff that we are going to remove as part of the torch-mlir refocus.
2021-09-23 00:55:09 +08:00
# TODO: Reduce dependencies. We need ExecutionEngine and a bunch of passes
# for the reference backend, but logically they can be separate. But seemingly
# the only way to handle that is to create a separate mlir python package
# tree, which seems excessive.
MLIRPythonSources
MLIRPythonExtension.Core
MLIRPythonExtension.AllPassesRegistration
MLIRPythonExtension.ExecutionEngine
[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
TorchMLIRPythonSources
TorchMLIRPythonExtensions
)
add_mlir_python_common_capi_library(TorchMLIRAggregateCAPI
INSTALL_COMPONENT TorchMLIRPythonModules
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
INSTALL_DESTINATION python_packages/torch_mlir/torch_mlir/_mlir_libs
OUTPUT_DIRECTORY "${TORCH_MLIR_PYTHON_PACKAGES_DIR}/torch_mlir/torch_mlir/_mlir_libs"
[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
RELATIVE_INSTALL_ROOT "../../../.."
DECLARED_SOURCES ${_source_components}
)
add_mlir_python_modules(TorchMLIRPythonModules
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
ROOT_PREFIX "${TORCH_MLIR_PYTHON_PACKAGES_DIR}/torch_mlir/torch_mlir"
INSTALL_PREFIX "python_packages/torch_mlir/torch_mlir"
[torch-mlir earthmoving (2/N)] Python code movement. This moves the bulk of the Python code (including the Torch interop) from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also required reconciling a bunch of other Python-related stuff, like the `torch` dialects. As I did this, it was simpler to just remove all the old numpy/basicpy stuff because we were going to delete it anyway and it was faster than debugging an intermediate state that would only last O(days) anyway. torch-mlir has two top-level python packages (built into the `python_packages` directory): - `torch_mlir_dialects`: `torch` dialect Python bindings (does not depend on PyTorch). This also involves building the aggregate CAPI for `torch-mlir`. - `torch_mlir`: bindings to the part of the code that links against PyTorch (or C++ code that transitively does). Additionally, there remain two more Python packages in npcomp (but outside `torch-mlir`): - `npcomp_torch`: Contains the e2e test framework and testing configs that plug into RefBackend and IREE. - `npcomp_core`: Contains the low-level interfaces to RefBackend and IREE that `npcomp_torch` uses, along with its own `MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR python bindings. (all other functionality has been stripped out) After all the basicpy/numpy deletions, the `npcomp` C++ code is now very tiny. It basically just contains RefBackend and the `TorchConversion` dialect/passes (e.g. `TorchToLinalg.cpp`). Correspondingly, there are now 4 main testing targets paralleling the Python layering (which is reflective of the deeper underlying dependency structure) - `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code. - `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g. TorchScript import) - `check-frontends-pytorch`: Checks the little code we have in `frontends/pytorch` -- mainly things related to the e2e framework itself. - `check-npcomp`: Checks the pure MLIR C++ code inside npcomp. There is a target `check-npcomp-all` that runs all of them. The `torch-mlir/build_standalone.sh` script does a standalone build of `torch-mlir`. The e2e tests (`tools/torchscript_e2e_test.sh`) are working too. The update_torch_ods script now lives in `torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone build. This change also required a fix upstream related to cross-shlib Python dependencies, so we also update llvm-project to 8dca953dd39c0cd8c80decbeb38753f58a4de580 to get https://reviews.llvm.org/D109776 (no other fixes were needed for the integrate, thankfully). This completes most of the large source code changes. Next will be bringing the CI/packaging/examples back to life.
2021-09-11 02:44:38 +08:00
DECLARED_SOURCES ${_source_components}
COMMON_CAPI_LINK_LIBS
TorchMLIRAggregateCAPI
)
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
# TODO: Find a cleaner way to do this.
# Can we build the JIT IR importer with `declare_mlir_python_extension`?
# Then it would "just work".
if(TORCH_MLIR_ENABLE_JIT_IR_IMPORTER)
add_dependencies(TorchMLIRPythonModules TorchMLIRJITIRImporter)
Port the bulk of the remaining code to torch-mlir This leaves no real code outside torch-mlir. This also renames the "npcomp backend contract" to "linalg on tensors backend contract" as the name of the abstraction layer that RefBackend (IREE too) accepts.
2021-09-28 02:36:44 +08:00
# Build the E2E Tests (which depend on the JIT IR importer now).
add_dependencies(TorchMLIRPythonModules TorchMLIRE2ETestPythonModules)
Implement Python relayout from #311 Fixes https://github.com/llvm/mlir-npcomp/issues/311 The key change is that TorchPlugin is folded into `torch_mlir.dialects.torch.importer.jit_ir` (it imports the PyTorch JIT's IR, so that's a good, scoped name for it). The CMake option `-DTORCH_MLIR_ENABLE_JIT_IR_IMPORTER=OFF` disables it, which allows building without a PyTorch native dependency.
2021-09-21 04:55:36 +08:00
endif()
Port the bulk of the remaining code to torch-mlir This leaves no real code outside torch-mlir. This also renames the "npcomp backend contract" to "linalg on tensors backend contract" as the name of the abstraction layer that RefBackend (IREE too) accepts.
2021-09-28 02:36:44 +08:00
Update CMakeLists.txt
2022-06-15 05:46:52 +08:00
# TODO: Add after macOS builds are fixed
#add_dependencies(TorchMLIRPythonModules torch_mlir_custom_op_example)
Allow torch-mlir to support PyTorch extensions. (#895) PyTorch allows new operators to be registered dynamically in modules. Torch-mlir already makes it fairly straightforward to add support for new operators, and this commit just extends that support to allow new PyTorch ops to come from a external module. This does *not* allow ops to be dynamically loaded into torch-mlir. Torch-mlir must still be compiled with support built-in. Add a `_torch_mlir_custom_op_example` subpackage to `torch_mlir` which registers an demonstration op. It will not be imported by default when importing torch_mlir. It's strictly for testing and documentation. Adds an end-to-end test for the `torch_mlir_custom_op_example::identity` op. With all these changes, we should now be actively testing PyTorch extension support with all future patches.
2022-06-14 05:51:30 +08:00
Port the bulk of the remaining code to torch-mlir This leaves no real code outside torch-mlir. This also renames the "npcomp backend contract" to "linalg on tensors backend contract" as the name of the abstraction layer that RefBackend (IREE too) accepts.
2021-09-28 02:36:44 +08:00
add_subdirectory(test)
Allow torch-mlir to support PyTorch extensions. (#895) PyTorch allows new operators to be registered dynamically in modules. Torch-mlir already makes it fairly straightforward to add support for new operators, and this commit just extends that support to allow new PyTorch ops to come from a external module. This does *not* allow ops to be dynamically loaded into torch-mlir. Torch-mlir must still be compiled with support built-in. Add a `_torch_mlir_custom_op_example` subpackage to `torch_mlir` which registers an demonstration op. It will not be imported by default when importing torch_mlir. It's strictly for testing and documentation. Adds an end-to-end test for the `torch_mlir_custom_op_example::identity` op. With all these changes, we should now be actively testing PyTorch extension support with all future patches.
2022-06-14 05:51:30 +08:00