[Stablehlo] Add AtenIndexTensor StableHlo support
#2107
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
AmosLewis
requested changes
May 24, 2023
There was a problem hiding this comment.
@zhekunz2 This patch looks good to me. I have tested it in OPT model. Could you clean this patch and merge all 8 commits into one? I will need this patch in nod-ai/SHARK-Studio#1441.
Hi @AmosLewis , usually those commits will be squashed during merging. If you don't have further opinion, I can merge it |
Yes. Just merged it. Thanks. |
AmosLewis
approved these changes
May 24, 2023
mgehre-amd
added a commit
to Xilinx/torch-mlir
that referenced
this pull request
May 30, 2023
* update PyTorch version to 2.1.0.dev20230523 (llvm#2148) - torch version: 2.1.0.dev20230523 - torch commit hash: 981d4c2578d10d8a96d173471802fc2812541fb1 - torchvision version: 0.16.0.dev20230523 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Torch Dialect] Add split.tensor support + recompose rules (llvm#2102) * add split.tensor support + recompose rules * add e2e test * address comments * address comments * erase op in recomposeOp --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * feat: add version differentiation for some tests * feat: activate more configs * [Stablehlo] Add `AtenIndexTensor` StableHlo support (llvm#2107) * Add AtenIndexTensor StableHlo support * clean up * Empty commit, trigger test * try to debug hanging test * fix segfulat * fix bad include --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [arm64] Fix release builds for ARM64 (llvm#2157) Tested on Ubuntu 23.04 on Ampere Altra instance. * [Stablehlo] Add aten.uniform lowering (llvm#2101) * add uniform stablehlo lowering * add unit test * new line * rm redundant file * Empty commit, trigger test * fix include * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * update PyTorch version to 2.1.0.dev20230525 (llvm#2167) - torch version: 2.1.0.dev20230525 - torch commit hash: eb2ef134b4e834a9b8a8b6de86ddd7d2780ce0ac - torchvision version: 0.16.0.dev20230525 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * CI: disable caching for release builds (llvm#2168) This patch adds a (default-true) input called `cache-enabled` to the setup-build action, so that when the input is false, ccache is not setup on the host machine. This patch also sets the input to be false for the release builds. * Add alias analysis for cast-like ops to maximize-value-semantics (llvm#2160) When `use_tracing=True` is used to import a model into Torch-MLIR, several casts get inserted in the IR to bridge the untyped inputs and outputs with the typed body of the computation. These casts create extra aliases of tensors that cause the current analysis in `maximize-value-semantics` to fail. In particular, the `maximize-value-semantics` analysis assumes that the only valid alias right after an overwrite is the overwritten alias. So, if there is a use of a casted version of the overwritten alias after the overwrite, the analysis fails. This commit improves the analysis by identifying all cast-like aliases of the overwritten alias and allowing such aliases to be used after an overwrite. Because this issue only arises when using tracing, it cannot be currently tested e2e, so only lit test is added. * only setup python for non-docker platforms (llvm#2171) Original PR was accidentally merged to a branch. Re-landing same PR to main now * Remove spurious pip in Release builds (llvm#2172) (left over from a previous commit that was approved and landed in a branch on accident) * [Torch Op] Add AtenChunkOp support (llvm#2152) * add chunkOp support * update LTC xfail list * address comments * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * Add ARM64 release builds (llvm#2159) Creates a build_linux_arm64 job that builds the release on an arm64 self-hosted runner. Drop Python 3.10 support Pass TM_TORCH_VERSION to choose the Stable PyTorch version (since arm64 doesn't have nightly builds) Borrows nightly / stable Pytorch switch from the WIP llvm#2038 * Delete another spurious pip (llvm#2173) * update PyTorch version to 2.1.0.dev20230526 (llvm#2175) - torch version: 2.1.0.dev20230526 - torch commit hash: 10b46f7c7f69f9bf705d2b6ea53efb9c59145685 - torchvision version: 0.16.0.dev20230526 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Stablehlo] Enable Stablehlo backend with arith dialect (llvm#2139) * Add correct type checking for tm_tensor.attention * [TM_TENSOR] Add `aten.scatter.[src|value]` op This commit adds support of `aten.scatter.src` and `aten.scatter.value` ops. Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com> * refactor: change implementation to use less requirement files * refactor: remove contraints used for testing * fix: revert some requirement file names * refactor: remove unnecessary ninja install * fix: fix version parsing * Keep installing ccache, because its not installed on the github default runners --------- Signed-off-by: Gaurav Shukla <gaurav@nod-labs.com> Co-authored-by: Maximilian Bartel <maximilian.bartel@amd.com> Co-authored-by: Sean Silva <silvasean@google.com> Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> Co-authored-by: Zhekun Zhang <32320144+zhekunz2@users.noreply.github.com> Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> Co-authored-by: powderluv <powderluv@users.noreply.github.com> Co-authored-by: Ashay Rane <ashay@users.noreply.github.com> Co-authored-by: Ramiro Leal-Cavazos <ramiroleal050@gmail.com> Co-authored-by: Yuanqiang Liu <liuyuanqiang.yqliu@bytedance.com> Co-authored-by: George Petterson <gpetters@protonmail.com> Co-authored-by: Gaurav Shukla <gaurav@nod-labs.com>
mgehre-amd
added a commit
to Xilinx/torch-mlir
that referenced
this pull request
May 30, 2023
* update PyTorch version to 2.1.0.dev20230523 (llvm#2148) - torch version: 2.1.0.dev20230523 - torch commit hash: 981d4c2578d10d8a96d173471802fc2812541fb1 - torchvision version: 0.16.0.dev20230523 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Torch Dialect] Add split.tensor support + recompose rules (llvm#2102) * add split.tensor support + recompose rules * add e2e test * address comments * address comments * erase op in recomposeOp --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * feat: add version differentiation for some tests * feat: activate more configs * [Stablehlo] Add `AtenIndexTensor` StableHlo support (llvm#2107) * Add AtenIndexTensor StableHlo support * clean up * Empty commit, trigger test * try to debug hanging test * fix segfulat * fix bad include --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [arm64] Fix release builds for ARM64 (llvm#2157) Tested on Ubuntu 23.04 on Ampere Altra instance. * [Stablehlo] Add aten.uniform lowering (llvm#2101) * add uniform stablehlo lowering * add unit test * new line * rm redundant file * Empty commit, trigger test * fix include * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * update PyTorch version to 2.1.0.dev20230525 (llvm#2167) - torch version: 2.1.0.dev20230525 - torch commit hash: eb2ef134b4e834a9b8a8b6de86ddd7d2780ce0ac - torchvision version: 0.16.0.dev20230525 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * CI: disable caching for release builds (llvm#2168) This patch adds a (default-true) input called `cache-enabled` to the setup-build action, so that when the input is false, ccache is not setup on the host machine. This patch also sets the input to be false for the release builds. * Add alias analysis for cast-like ops to maximize-value-semantics (llvm#2160) When `use_tracing=True` is used to import a model into Torch-MLIR, several casts get inserted in the IR to bridge the untyped inputs and outputs with the typed body of the computation. These casts create extra aliases of tensors that cause the current analysis in `maximize-value-semantics` to fail. In particular, the `maximize-value-semantics` analysis assumes that the only valid alias right after an overwrite is the overwritten alias. So, if there is a use of a casted version of the overwritten alias after the overwrite, the analysis fails. This commit improves the analysis by identifying all cast-like aliases of the overwritten alias and allowing such aliases to be used after an overwrite. Because this issue only arises when using tracing, it cannot be currently tested e2e, so only lit test is added. * [LINALG] Add dynamic support for `PrimMinIntOp` * Fix types + off-by-1 error, clamp `end` in slice+copy_ recomposition The `copy_` op being replaced by `RecomposeSliceCopy_` operates on a subset of the tensor being mutated, while the `index_put` op being used to replace the `copy_` op operates on the entire tensor being mutated. This means that the result type of the `index_put` should be the type of the input to `index_put` and we need to make sure that `copy_` does not have users before replacing to avoid type conflicts. This commit also fixes the result type used for the `AtenArangeStartStepOp`, and an off-by-1 error when creating the indices vector. Lastly, this commit also clamps the `end` value from the slice to the size of the dimension. * only setup python for non-docker platforms (llvm#2171) Original PR was accidentally merged to a branch. Re-landing same PR to main now * Remove spurious pip in Release builds (llvm#2172) (left over from a previous commit that was approved and landed in a branch on accident) * [Torch Op] Add AtenChunkOp support (llvm#2152) * add chunkOp support * update LTC xfail list * address comments * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * Add ARM64 release builds (llvm#2159) Creates a build_linux_arm64 job that builds the release on an arm64 self-hosted runner. Drop Python 3.10 support Pass TM_TORCH_VERSION to choose the Stable PyTorch version (since arm64 doesn't have nightly builds) Borrows nightly / stable Pytorch switch from the WIP llvm#2038 * Delete another spurious pip (llvm#2173) * update PyTorch version to 2.1.0.dev20230526 (llvm#2175) - torch version: 2.1.0.dev20230526 - torch commit hash: 10b46f7c7f69f9bf705d2b6ea53efb9c59145685 - torchvision version: 0.16.0.dev20230526 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Stablehlo] Enable Stablehlo backend with arith dialect (llvm#2139) * Add correct type checking for tm_tensor.attention * [TM_TENSOR] Add `aten.scatter.[src|value]` op This commit adds support of `aten.scatter.src` and `aten.scatter.value` ops. Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com> * refactor: change implementation to use less requirement files * refactor: remove contraints used for testing * fix: revert some requirement file names * refactor: remove unnecessary ninja install * fix: fix version parsing * Keep installing ccache, because its not installed on the github default runners --------- Signed-off-by: Gaurav Shukla <gaurav@nod-labs.com> Co-authored-by: Maximilian Bartel <maximilian.bartel@amd.com> Co-authored-by: Sean Silva <silvasean@google.com> Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> Co-authored-by: Zhekun Zhang <32320144+zhekunz2@users.noreply.github.com> Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> Co-authored-by: powderluv <powderluv@users.noreply.github.com> Co-authored-by: Ashay Rane <ashay@users.noreply.github.com> Co-authored-by: Ramiro Leal-Cavazos <ramiroleal050@gmail.com> Co-authored-by: Yuanqiang Liu <liuyuanqiang.yqliu@bytedance.com> Co-authored-by: George Petterson <gpetters@protonmail.com> Co-authored-by: Gaurav Shukla <gaurav@nod-labs.com>
mgehre-amd
added a commit
to Xilinx/torch-mlir
that referenced
this pull request
May 30, 2023
* update PyTorch version to 2.1.0.dev20230523 (llvm#2148) - torch version: 2.1.0.dev20230523 - torch commit hash: 981d4c2578d10d8a96d173471802fc2812541fb1 - torchvision version: 0.16.0.dev20230523 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Torch Dialect] Add split.tensor support + recompose rules (llvm#2102) * add split.tensor support + recompose rules * add e2e test * address comments * address comments * erase op in recomposeOp --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [Stablehlo] Add `AtenIndexTensor` StableHlo support (llvm#2107) * Add AtenIndexTensor StableHlo support * clean up * Empty commit, trigger test * try to debug hanging test * fix segfulat * fix bad include --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [arm64] Fix release builds for ARM64 (llvm#2157) Tested on Ubuntu 23.04 on Ampere Altra instance. * [Stablehlo] Add aten.uniform lowering (llvm#2101) * add uniform stablehlo lowering * add unit test * new line * rm redundant file * Empty commit, trigger test * fix include * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * update PyTorch version to 2.1.0.dev20230525 (llvm#2167) - torch version: 2.1.0.dev20230525 - torch commit hash: eb2ef134b4e834a9b8a8b6de86ddd7d2780ce0ac - torchvision version: 0.16.0.dev20230525 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * CI: disable caching for release builds (llvm#2168) This patch adds a (default-true) input called `cache-enabled` to the setup-build action, so that when the input is false, ccache is not setup on the host machine. This patch also sets the input to be false for the release builds. * Add alias analysis for cast-like ops to maximize-value-semantics (llvm#2160) When `use_tracing=True` is used to import a model into Torch-MLIR, several casts get inserted in the IR to bridge the untyped inputs and outputs with the typed body of the computation. These casts create extra aliases of tensors that cause the current analysis in `maximize-value-semantics` to fail. In particular, the `maximize-value-semantics` analysis assumes that the only valid alias right after an overwrite is the overwritten alias. So, if there is a use of a casted version of the overwritten alias after the overwrite, the analysis fails. This commit improves the analysis by identifying all cast-like aliases of the overwritten alias and allowing such aliases to be used after an overwrite. Because this issue only arises when using tracing, it cannot be currently tested e2e, so only lit test is added. * only setup python for non-docker platforms (llvm#2171) Original PR was accidentally merged to a branch. Re-landing same PR to main now * Remove spurious pip in Release builds (llvm#2172) (left over from a previous commit that was approved and landed in a branch on accident) * [Torch Op] Add AtenChunkOp support (llvm#2152) * add chunkOp support * update LTC xfail list * address comments * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * Add ARM64 release builds (llvm#2159) Creates a build_linux_arm64 job that builds the release on an arm64 self-hosted runner. Drop Python 3.10 support Pass TM_TORCH_VERSION to choose the Stable PyTorch version (since arm64 doesn't have nightly builds) Borrows nightly / stable Pytorch switch from the WIP llvm#2038 * Delete another spurious pip (llvm#2173) * update PyTorch version to 2.1.0.dev20230526 (llvm#2175) - torch version: 2.1.0.dev20230526 - torch commit hash: 10b46f7c7f69f9bf705d2b6ea53efb9c59145685 - torchvision version: 0.16.0.dev20230526 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Stablehlo] Enable Stablehlo backend with arith dialect (llvm#2139) * Add `ReadOnly` trait to `copy.to_vtensor` Before inlining a global slot, the users of the global slot are checked to see if they are `ReadOnly` or `MemoryEffectFree` to make sure that the global slot is not being mutated. Because the op `copy.to_vtensor` currently does not have the `ReadOnly` trait, if a global slot is passed to `copy.to_vtensor`, the pass `InlineGlobalSlots` will fail. The op `copy.to_vtensor` is `ReadOnly`, since it does not modify the contents of the input tensor; it simply makes a new copy. This commit adds the trait as well as an e2e test that generates the case of a global slot being passed to a `copy.to_vtensor`. * Add correct type checking for tm_tensor.attention * [TM_TENSOR] Add `aten.scatter.[src|value]` op This commit adds support of `aten.scatter.src` and `aten.scatter.value` ops. Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com> --------- Signed-off-by: Gaurav Shukla <gaurav@nod-labs.com> Co-authored-by: Sean Silva <silvasean@google.com> Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> Co-authored-by: Zhekun Zhang <32320144+zhekunz2@users.noreply.github.com> Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> Co-authored-by: powderluv <powderluv@users.noreply.github.com> Co-authored-by: Ashay Rane <ashay@users.noreply.github.com> Co-authored-by: Ramiro Leal-Cavazos <ramiroleal050@gmail.com> Co-authored-by: Yuanqiang Liu <liuyuanqiang.yqliu@bytedance.com> Co-authored-by: George Petterson <gpetters@protonmail.com> Co-authored-by: Gaurav Shukla <gaurav@nod-labs.com>
mgehre-amd
added a commit
to Xilinx/torch-mlir
that referenced
this pull request
May 30, 2023
* update PyTorch version to 2.1.0.dev20230523 (llvm#2148) - torch version: 2.1.0.dev20230523 - torch commit hash: 981d4c2578d10d8a96d173471802fc2812541fb1 - torchvision version: 0.16.0.dev20230523 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Torch Dialect] Add split.tensor support + recompose rules (llvm#2102) * add split.tensor support + recompose rules * add e2e test * address comments * address comments * erase op in recomposeOp --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * feat: add version differentiation for some tests * feat: activate more configs * [Stablehlo] Add `AtenIndexTensor` StableHlo support (llvm#2107) * Add AtenIndexTensor StableHlo support * clean up * Empty commit, trigger test * try to debug hanging test * fix segfulat * fix bad include --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [arm64] Fix release builds for ARM64 (llvm#2157) Tested on Ubuntu 23.04 on Ampere Altra instance. * [Stablehlo] Add aten.uniform lowering (llvm#2101) * add uniform stablehlo lowering * add unit test * new line * rm redundant file * Empty commit, trigger test * fix include * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * update PyTorch version to 2.1.0.dev20230525 (llvm#2167) - torch version: 2.1.0.dev20230525 - torch commit hash: eb2ef134b4e834a9b8a8b6de86ddd7d2780ce0ac - torchvision version: 0.16.0.dev20230525 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * CI: disable caching for release builds (llvm#2168) This patch adds a (default-true) input called `cache-enabled` to the setup-build action, so that when the input is false, ccache is not setup on the host machine. This patch also sets the input to be false for the release builds. * Add alias analysis for cast-like ops to maximize-value-semantics (llvm#2160) When `use_tracing=True` is used to import a model into Torch-MLIR, several casts get inserted in the IR to bridge the untyped inputs and outputs with the typed body of the computation. These casts create extra aliases of tensors that cause the current analysis in `maximize-value-semantics` to fail. In particular, the `maximize-value-semantics` analysis assumes that the only valid alias right after an overwrite is the overwritten alias. So, if there is a use of a casted version of the overwritten alias after the overwrite, the analysis fails. This commit improves the analysis by identifying all cast-like aliases of the overwritten alias and allowing such aliases to be used after an overwrite. Because this issue only arises when using tracing, it cannot be currently tested e2e, so only lit test is added. * only setup python for non-docker platforms (llvm#2171) Original PR was accidentally merged to a branch. Re-landing same PR to main now * Remove spurious pip in Release builds (llvm#2172) (left over from a previous commit that was approved and landed in a branch on accident) * [Torch Op] Add AtenChunkOp support (llvm#2152) * add chunkOp support * update LTC xfail list * address comments * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * Add ARM64 release builds (llvm#2159) Creates a build_linux_arm64 job that builds the release on an arm64 self-hosted runner. Drop Python 3.10 support Pass TM_TORCH_VERSION to choose the Stable PyTorch version (since arm64 doesn't have nightly builds) Borrows nightly / stable Pytorch switch from the WIP llvm#2038 * Delete another spurious pip (llvm#2173) * update PyTorch version to 2.1.0.dev20230526 (llvm#2175) - torch version: 2.1.0.dev20230526 - torch commit hash: 10b46f7c7f69f9bf705d2b6ea53efb9c59145685 - torchvision version: 0.16.0.dev20230526 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Stablehlo] Enable Stablehlo backend with arith dialect (llvm#2139) * Add correct type checking for tm_tensor.attention * [TM_TENSOR] Add `aten.scatter.[src|value]` op This commit adds support of `aten.scatter.src` and `aten.scatter.value` ops. Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com> * refactor: change implementation to use less requirement files * refactor: remove contraints used for testing * fix: revert some requirement file names * refactor: remove unnecessary ninja install * fix: fix version parsing * refactor: remove dependency on torchvision in main requirements file * refactor: remove index url * style: remove unnecesary line switch * fix: readd index url * build_tools/python_deploy/build_linux_packages.sh: wheel: Install cpu version of torch * Remove fetch-depth * setup.py: Revert ccache back to upstream --------- Signed-off-by: Gaurav Shukla <gaurav@nod-labs.com> Co-authored-by: Maximilian Bartel <maximilian.bartel@amd.com> Co-authored-by: Sean Silva <silvasean@google.com> Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> Co-authored-by: Zhekun Zhang <32320144+zhekunz2@users.noreply.github.com> Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> Co-authored-by: powderluv <powderluv@users.noreply.github.com> Co-authored-by: Ashay Rane <ashay@users.noreply.github.com> Co-authored-by: Ramiro Leal-Cavazos <ramiroleal050@gmail.com> Co-authored-by: Yuanqiang Liu <liuyuanqiang.yqliu@bytedance.com> Co-authored-by: George Petterson <gpetters@protonmail.com> Co-authored-by: Gaurav Shukla <gaurav@nod-labs.com>
mgehre-amd
added a commit
to Xilinx/torch-mlir
that referenced
this pull request
Jun 1, 2023
…am (#53) * update PyTorch version to 2.1.0.dev20230523 (llvm#2148) - torch version: 2.1.0.dev20230523 - torch commit hash: 981d4c2578d10d8a96d173471802fc2812541fb1 - torchvision version: 0.16.0.dev20230523 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Torch Dialect] Add split.tensor support + recompose rules (llvm#2102) * add split.tensor support + recompose rules * add e2e test * address comments * address comments * erase op in recomposeOp --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [Stablehlo] Add `AtenIndexTensor` StableHlo support (llvm#2107) * Add AtenIndexTensor StableHlo support * clean up * Empty commit, trigger test * try to debug hanging test * fix segfulat * fix bad include --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [arm64] Fix release builds for ARM64 (llvm#2157) Tested on Ubuntu 23.04 on Ampere Altra instance. * [Stablehlo] Add aten.uniform lowering (llvm#2101) * add uniform stablehlo lowering * add unit test * new line * rm redundant file * Empty commit, trigger test * fix include * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * update PyTorch version to 2.1.0.dev20230525 (llvm#2167) - torch version: 2.1.0.dev20230525 - torch commit hash: eb2ef134b4e834a9b8a8b6de86ddd7d2780ce0ac - torchvision version: 0.16.0.dev20230525 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * CI: disable caching for release builds (llvm#2168) This patch adds a (default-true) input called `cache-enabled` to the setup-build action, so that when the input is false, ccache is not setup on the host machine. This patch also sets the input to be false for the release builds. * Add alias analysis for cast-like ops to maximize-value-semantics (llvm#2160) When `use_tracing=True` is used to import a model into Torch-MLIR, several casts get inserted in the IR to bridge the untyped inputs and outputs with the typed body of the computation. These casts create extra aliases of tensors that cause the current analysis in `maximize-value-semantics` to fail. In particular, the `maximize-value-semantics` analysis assumes that the only valid alias right after an overwrite is the overwritten alias. So, if there is a use of a casted version of the overwritten alias after the overwrite, the analysis fails. This commit improves the analysis by identifying all cast-like aliases of the overwritten alias and allowing such aliases to be used after an overwrite. Because this issue only arises when using tracing, it cannot be currently tested e2e, so only lit test is added. * only setup python for non-docker platforms (llvm#2171) Original PR was accidentally merged to a branch. Re-landing same PR to main now * Remove spurious pip in Release builds (llvm#2172) (left over from a previous commit that was approved and landed in a branch on accident) * [Torch Op] Add AtenChunkOp support (llvm#2152) * add chunkOp support * update LTC xfail list * address comments * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * Add ARM64 release builds (llvm#2159) Creates a build_linux_arm64 job that builds the release on an arm64 self-hosted runner. Drop Python 3.10 support Pass TM_TORCH_VERSION to choose the Stable PyTorch version (since arm64 doesn't have nightly builds) Borrows nightly / stable Pytorch switch from the WIP llvm#2038 * Delete another spurious pip (llvm#2173) * update PyTorch version to 2.1.0.dev20230526 (llvm#2175) - torch version: 2.1.0.dev20230526 - torch commit hash: 10b46f7c7f69f9bf705d2b6ea53efb9c59145685 - torchvision version: 0.16.0.dev20230526 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Stablehlo] Enable Stablehlo backend with arith dialect (llvm#2139) * Add correct type checking for tm_tensor.attention * [TM_TENSOR] Add `aten.scatter.[src|value]` op This commit adds support of `aten.scatter.src` and `aten.scatter.value` ops. Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com> * [MLIR][TORCH] Add support for the total_weight for aten.nll_loss_forward op Signed-Off By: Vivek Khandelwal <vivek@nod-labs.com> * Add Stable PyTorch CI Pipeline (llvm#2038) * feat: split pytorch requirements into stable and nightly * fix: add true to tests to see full output * refactor: add comments to explain true statement * feat: move some tests to experimental mode * refactor: refactor pipeline into more fine grained difference * feat: add version differentiation for some tests * feat: activate more configs * refactor: change implementation to use less requirement files * refactor: remove contraints used for testing * fix: revert some requirement file names * refactor: remove unnecessary ninja install * fix: fix version parsing * refactor: remove dependency on torchvision in main requirements file * refactor: remove index url * style: remove unnecesary line switch * fix: readd index url * Add `ReadOnly` trait to `copy.to_vtensor` (llvm#2179) Before inlining a global slot, the users of the global slot are checked to see if they are `ReadOnly` or `MemoryEffectFree` to make sure that the global slot is not being mutated. Because the op `copy.to_vtensor` currently does not have the `ReadOnly` trait, if a global slot is passed to `copy.to_vtensor`, the pass `InlineGlobalSlots` will fail. The op `copy.to_vtensor` is `ReadOnly`, since it does not modify the contents of the input tensor; it simply makes a new copy. This commit adds the trait as well as an e2e test that generates the case of a global slot being passed to a `copy.to_vtensor`. * [Importer] import constant tuple (llvm#2132) * [Importer] import constant tuple * update * update * update * e2e_testing/xfail_sets.py: LTC: xfail PrimsSumFloatModule_basic --------- Signed-off-by: Gaurav Shukla <gaurav@nod-labs.com> Co-authored-by: Sean Silva <silvasean@google.com> Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> Co-authored-by: Zhekun Zhang <32320144+zhekunz2@users.noreply.github.com> Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> Co-authored-by: powderluv <powderluv@users.noreply.github.com> Co-authored-by: Ashay Rane <ashay@users.noreply.github.com> Co-authored-by: Ramiro Leal-Cavazos <ramiroleal050@gmail.com> Co-authored-by: Yuanqiang Liu <liuyuanqiang.yqliu@bytedance.com> Co-authored-by: George Petterson <gpetters@protonmail.com> Co-authored-by: Gaurav Shukla <gaurav@nod-labs.com> Co-authored-by: Vivek Khandelwal <vivekkhandelwal1424@gmail.com> Co-authored-by: maxbartel <maximilian.bartel@amd.com>
mgehre-amd
added a commit
to Xilinx/torch-mlir
that referenced
this pull request
Jun 2, 2023
* update PyTorch version to 2.1.0.dev20230523 (llvm#2148) - torch version: 2.1.0.dev20230523 - torch commit hash: 981d4c2578d10d8a96d173471802fc2812541fb1 - torchvision version: 0.16.0.dev20230523 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Torch Dialect] Add split.tensor support + recompose rules (llvm#2102) * add split.tensor support + recompose rules * add e2e test * address comments * address comments * erase op in recomposeOp --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [Stablehlo] Add `AtenIndexTensor` StableHlo support (llvm#2107) * Add AtenIndexTensor StableHlo support * clean up * Empty commit, trigger test * try to debug hanging test * fix segfulat * fix bad include --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [arm64] Fix release builds for ARM64 (llvm#2157) Tested on Ubuntu 23.04 on Ampere Altra instance. * [Stablehlo] Add aten.uniform lowering (llvm#2101) * add uniform stablehlo lowering * add unit test * new line * rm redundant file * Empty commit, trigger test * fix include * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * update PyTorch version to 2.1.0.dev20230525 (llvm#2167) - torch version: 2.1.0.dev20230525 - torch commit hash: eb2ef134b4e834a9b8a8b6de86ddd7d2780ce0ac - torchvision version: 0.16.0.dev20230525 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * CI: disable caching for release builds (llvm#2168) This patch adds a (default-true) input called `cache-enabled` to the setup-build action, so that when the input is false, ccache is not setup on the host machine. This patch also sets the input to be false for the release builds. * Add alias analysis for cast-like ops to maximize-value-semantics (llvm#2160) When `use_tracing=True` is used to import a model into Torch-MLIR, several casts get inserted in the IR to bridge the untyped inputs and outputs with the typed body of the computation. These casts create extra aliases of tensors that cause the current analysis in `maximize-value-semantics` to fail. In particular, the `maximize-value-semantics` analysis assumes that the only valid alias right after an overwrite is the overwritten alias. So, if there is a use of a casted version of the overwritten alias after the overwrite, the analysis fails. This commit improves the analysis by identifying all cast-like aliases of the overwritten alias and allowing such aliases to be used after an overwrite. Because this issue only arises when using tracing, it cannot be currently tested e2e, so only lit test is added. * only setup python for non-docker platforms (llvm#2171) Original PR was accidentally merged to a branch. Re-landing same PR to main now * Remove spurious pip in Release builds (llvm#2172) (left over from a previous commit that was approved and landed in a branch on accident) * [Torch Op] Add AtenChunkOp support (llvm#2152) * add chunkOp support * update LTC xfail list * address comments * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * Add ARM64 release builds (llvm#2159) Creates a build_linux_arm64 job that builds the release on an arm64 self-hosted runner. Drop Python 3.10 support Pass TM_TORCH_VERSION to choose the Stable PyTorch version (since arm64 doesn't have nightly builds) Borrows nightly / stable Pytorch switch from the WIP llvm#2038 * Delete another spurious pip (llvm#2173) * update PyTorch version to 2.1.0.dev20230526 (llvm#2175) - torch version: 2.1.0.dev20230526 - torch commit hash: 10b46f7c7f69f9bf705d2b6ea53efb9c59145685 - torchvision version: 0.16.0.dev20230526 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Stablehlo] Enable Stablehlo backend with arith dialect (llvm#2139) * Add correct type checking for tm_tensor.attention * [TM_TENSOR] Add `aten.scatter.[src|value]` op This commit adds support of `aten.scatter.src` and `aten.scatter.value` ops. Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com> * [MLIR][TORCH] Add support for the total_weight for aten.nll_loss_forward op Signed-Off By: Vivek Khandelwal <vivek@nod-labs.com> * Add Stable PyTorch CI Pipeline (llvm#2038) * feat: split pytorch requirements into stable and nightly * fix: add true to tests to see full output * refactor: add comments to explain true statement * feat: move some tests to experimental mode * refactor: refactor pipeline into more fine grained difference * feat: add version differentiation for some tests * feat: activate more configs * refactor: change implementation to use less requirement files * refactor: remove contraints used for testing * fix: revert some requirement file names * refactor: remove unnecessary ninja install * fix: fix version parsing * refactor: remove dependency on torchvision in main requirements file * refactor: remove index url * style: remove unnecesary line switch * fix: readd index url * Add `ReadOnly` trait to `copy.to_vtensor` (llvm#2179) Before inlining a global slot, the users of the global slot are checked to see if they are `ReadOnly` or `MemoryEffectFree` to make sure that the global slot is not being mutated. Because the op `copy.to_vtensor` currently does not have the `ReadOnly` trait, if a global slot is passed to `copy.to_vtensor`, the pass `InlineGlobalSlots` will fail. The op `copy.to_vtensor` is `ReadOnly`, since it does not modify the contents of the input tensor; it simply makes a new copy. This commit adds the trait as well as an e2e test that generates the case of a global slot being passed to a `copy.to_vtensor`. * [Importer] import constant tuple (llvm#2132) * [Importer] import constant tuple * update * update * update * update PyTorch version to 2.1.0.dev20230531 (llvm#2188) - torch version: 2.1.0.dev20230531 - torch commit hash: 48552338649ccc467060f5f93dbe19e2acbc4d1a - torchvision version: 0.16.0.dev20230531 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Torch Dialect] Add support for AtenScalarTensorOp (llvm#2085) * add scalar_tensor op * add dynamo pass test; needs PR2062 * try to fix * Empty commit, trigger test * Empty commit, trigger test * address comments * use dtype function * fix decompose rule * remove unused include * Empty commit, trigger test * fix test * disable ltc * fix dtype --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> --------- Signed-off-by: Gaurav Shukla <gaurav@nod-labs.com> Co-authored-by: Sean Silva <silvasean@google.com> Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> Co-authored-by: Zhekun Zhang <32320144+zhekunz2@users.noreply.github.com> Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> Co-authored-by: powderluv <powderluv@users.noreply.github.com> Co-authored-by: Ashay Rane <ashay@users.noreply.github.com> Co-authored-by: Ramiro Leal-Cavazos <ramiroleal050@gmail.com> Co-authored-by: Yuanqiang Liu <liuyuanqiang.yqliu@bytedance.com> Co-authored-by: George Petterson <gpetters@protonmail.com> Co-authored-by: Gaurav Shukla <gaurav@nod-labs.com> Co-authored-by: Vivek Khandelwal <vivekkhandelwal1424@gmail.com> Co-authored-by: maxbartel <maximilian.bartel@amd.com>
mgehre-amd
added a commit
to Xilinx/torch-mlir
that referenced
this pull request
Jun 5, 2023
* update PyTorch version to 2.1.0.dev20230523 (llvm#2148) - torch version: 2.1.0.dev20230523 - torch commit hash: 981d4c2578d10d8a96d173471802fc2812541fb1 - torchvision version: 0.16.0.dev20230523 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Torch Dialect] Add split.tensor support + recompose rules (llvm#2102) * add split.tensor support + recompose rules * add e2e test * address comments * address comments * erase op in recomposeOp --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [Stablehlo] Add `AtenIndexTensor` StableHlo support (llvm#2107) * Add AtenIndexTensor StableHlo support * clean up * Empty commit, trigger test * try to debug hanging test * fix segfulat * fix bad include --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * [arm64] Fix release builds for ARM64 (llvm#2157) Tested on Ubuntu 23.04 on Ampere Altra instance. * [Stablehlo] Add aten.uniform lowering (llvm#2101) * add uniform stablehlo lowering * add unit test * new line * rm redundant file * Empty commit, trigger test * fix include * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * update PyTorch version to 2.1.0.dev20230525 (llvm#2167) - torch version: 2.1.0.dev20230525 - torch commit hash: eb2ef134b4e834a9b8a8b6de86ddd7d2780ce0ac - torchvision version: 0.16.0.dev20230525 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * CI: disable caching for release builds (llvm#2168) This patch adds a (default-true) input called `cache-enabled` to the setup-build action, so that when the input is false, ccache is not setup on the host machine. This patch also sets the input to be false for the release builds. * Add alias analysis for cast-like ops to maximize-value-semantics (llvm#2160) When `use_tracing=True` is used to import a model into Torch-MLIR, several casts get inserted in the IR to bridge the untyped inputs and outputs with the typed body of the computation. These casts create extra aliases of tensors that cause the current analysis in `maximize-value-semantics` to fail. In particular, the `maximize-value-semantics` analysis assumes that the only valid alias right after an overwrite is the overwritten alias. So, if there is a use of a casted version of the overwritten alias after the overwrite, the analysis fails. This commit improves the analysis by identifying all cast-like aliases of the overwritten alias and allowing such aliases to be used after an overwrite. Because this issue only arises when using tracing, it cannot be currently tested e2e, so only lit test is added. * only setup python for non-docker platforms (llvm#2171) Original PR was accidentally merged to a branch. Re-landing same PR to main now * Remove spurious pip in Release builds (llvm#2172) (left over from a previous commit that was approved and landed in a branch on accident) * [Torch Op] Add AtenChunkOp support (llvm#2152) * add chunkOp support * update LTC xfail list * address comments * address comments --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * Add ARM64 release builds (llvm#2159) Creates a build_linux_arm64 job that builds the release on an arm64 self-hosted runner. Drop Python 3.10 support Pass TM_TORCH_VERSION to choose the Stable PyTorch version (since arm64 doesn't have nightly builds) Borrows nightly / stable Pytorch switch from the WIP llvm#2038 * Delete another spurious pip (llvm#2173) * update PyTorch version to 2.1.0.dev20230526 (llvm#2175) - torch version: 2.1.0.dev20230526 - torch commit hash: 10b46f7c7f69f9bf705d2b6ea53efb9c59145685 - torchvision version: 0.16.0.dev20230526 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Stablehlo] Enable Stablehlo backend with arith dialect (llvm#2139) * Add correct type checking for tm_tensor.attention * [TM_TENSOR] Add `aten.scatter.[src|value]` op This commit adds support of `aten.scatter.src` and `aten.scatter.value` ops. Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com> * [MLIR][TORCH] Add support for the total_weight for aten.nll_loss_forward op Signed-Off By: Vivek Khandelwal <vivek@nod-labs.com> * Add Stable PyTorch CI Pipeline (llvm#2038) * feat: split pytorch requirements into stable and nightly * fix: add true to tests to see full output * refactor: add comments to explain true statement * feat: move some tests to experimental mode * refactor: refactor pipeline into more fine grained difference * feat: add version differentiation for some tests * feat: activate more configs * refactor: change implementation to use less requirement files * refactor: remove contraints used for testing * fix: revert some requirement file names * refactor: remove unnecessary ninja install * fix: fix version parsing * refactor: remove dependency on torchvision in main requirements file * refactor: remove index url * style: remove unnecesary line switch * fix: readd index url * Add `ReadOnly` trait to `copy.to_vtensor` (llvm#2179) Before inlining a global slot, the users of the global slot are checked to see if they are `ReadOnly` or `MemoryEffectFree` to make sure that the global slot is not being mutated. Because the op `copy.to_vtensor` currently does not have the `ReadOnly` trait, if a global slot is passed to `copy.to_vtensor`, the pass `InlineGlobalSlots` will fail. The op `copy.to_vtensor` is `ReadOnly`, since it does not modify the contents of the input tensor; it simply makes a new copy. This commit adds the trait as well as an e2e test that generates the case of a global slot being passed to a `copy.to_vtensor`. * [Importer] import constant tuple (llvm#2132) * [Importer] import constant tuple * update * update * update * update PyTorch version to 2.1.0.dev20230531 (llvm#2188) - torch version: 2.1.0.dev20230531 - torch commit hash: 48552338649ccc467060f5f93dbe19e2acbc4d1a - torchvision version: 0.16.0.dev20230531 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * [Torch Dialect] Add support for AtenScalarTensorOp (llvm#2085) * add scalar_tensor op * add dynamo pass test; needs PR2062 * try to fix * Empty commit, trigger test * Empty commit, trigger test * address comments * use dtype function * fix decompose rule * remove unused include * Empty commit, trigger test * fix test * disable ltc * fix dtype --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> * update PyTorch version to 2.1.0.dev20230601 (llvm#2189) * [LINALG] Add dynamic support for `PrimMinIntOp` * Fix types + off-by-1 error, clamp `end` in slice+copy_ recomposition The `copy_` op being replaced by `RecomposeSliceCopy_` operates on a subset of the tensor being mutated, while the `index_put` op being used to replace the `copy_` op operates on the entire tensor being mutated. This means that the result type of the `index_put` should be the type of the input to `index_put` and we need to make sure that `copy_` does not have users before replacing to avoid type conflicts. This commit also fixes the result type used for the `AtenArangeStartStepOp`, and an off-by-1 error when creating the indices vector. Lastly, this commit also clamps the `end` value from the slice to the size of the dimension. * CI: Spot fixes related to nightly and stable PyTorch builds (llvm#2190) * CI: Skip (redundant) libtorch build when using stable PyTorch version When we use PyTorch stable builds, there is no need to build libtorch from source, making the stable-pytorch-with-torch-binary-OFF configuration redundant with stable-pytorch-with-torch-binary-ON. This patch drops the redundant configuration from CI. * CI: Simplify guard conditions for creating and using libtorch cache Whether libtorch is enabled or not is predicated on a host of conditions such as the platform, in-tree versus out-of-tree build, and stable versus nightly PyTorch builds. Instead of repeating these conditions to guard whether to create or use the libtorch cache artifacts (and getting them almost incorrect), this patch predicates the relevant pipeline steps to whether libtorch is enabled, thus making the conditions far simpler. * update PyTorch version to 2.1.0.dev20230602 (llvm#2191) - torch version: 2.1.0.dev20230602 - torch commit hash: 52c7a761c5cb6ae94acf2298827309fba3dbc0f4 - torchvision version: 0.16.0.dev20230602 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * update PyTorch version to 2.1.0.dev20230603 (llvm#2193) - torch version: 2.1.0.dev20230603 - torch commit hash: 7726721661ea114acb81a860519d0a1501d88fca - torchvision version: 0.16.0.dev20230603 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> * update PyTorch version to 2.1.0.dev20230604 (llvm#2195) - torch version: 2.1.0.dev20230604 - torch commit hash: 810edae5137bdc0cd25ac2f133d6633d6146b1e9 - torchvision version: 0.16.0.dev20230604 Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> --------- Signed-off-by: Gaurav Shukla <gaurav@nod-labs.com> Co-authored-by: Sean Silva <silvasean@google.com> Co-authored-by: Roll PyTorch Action <torch-mlir@users.noreply.github.com> Co-authored-by: Zhekun Zhang <32320144+zhekunz2@users.noreply.github.com> Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com> Co-authored-by: powderluv <powderluv@users.noreply.github.com> Co-authored-by: Ashay Rane <ashay@users.noreply.github.com> Co-authored-by: Ramiro Leal-Cavazos <ramiroleal050@gmail.com> Co-authored-by: Yuanqiang Liu <liuyuanqiang.yqliu@bytedance.com> Co-authored-by: George Petterson <gpetters@protonmail.com> Co-authored-by: Gaurav Shukla <gaurav@nod-labs.com> Co-authored-by: Vivek Khandelwal <vivekkhandelwal1424@gmail.com> Co-authored-by: maxbartel <maximilian.bartel@amd.com>
gpetters94
pushed a commit
to gpetters94/mlir-npcomp
that referenced
this pull request
Jul 7, 2023
* Add AtenIndexTensor StableHlo support * clean up * Empty commit, trigger test * try to debug hanging test * fix segfulat * fix bad include --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com>
gpetters94
pushed a commit
to gpetters94/mlir-npcomp
that referenced
this pull request
Jul 7, 2023
* Add AtenIndexTensor StableHlo support * clean up * Empty commit, trigger test * try to debug hanging test * fix segfulat * fix bad include --------- Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.