From 9e150657c3c4d100f534f95ee52fa791d45e10da Mon Sep 17 00:00:00 2001 From: angelayi Date: Thu, 3 Jul 2025 11:42:04 -0700 Subject: [PATCH] export release notes --- 2.8.0/done/result_export.md | 116 ++++++++++++++++++++++++ 2.8.0/todo/result_composability.md | 2 + 2.8.0/todo/result_export.md | 128 --------------------------- 2.8.0/todo/result_inductor (aoti).md | 2 +- 4 files changed, 119 insertions(+), 129 deletions(-) create mode 100644 2.8.0/done/result_export.md delete mode 100644 2.8.0/todo/result_export.md diff --git a/2.8.0/done/result_export.md b/2.8.0/done/result_export.md new file mode 100644 index 0000000..72b9bf0 --- /dev/null +++ b/2.8.0/done/result_export.md @@ -0,0 +1,116 @@ + +# Release Notes worksheet export + +The main goal of this process is to rephrase all the commit messages below to make them **clear and easy to read** by the end user. You should follow the following instructions to do so: + +* **Please clean up and format commit titles to be readable by the general PyTorch user.** Make sure you're [following the guidance here](https://docs.google.com/document/d/14OmgGBr1w6gl1VO47GGGdwrIaUNr92DFhQbY_NEk8mQ/edit)! Your resulting notes must be consistent and easy to read. +* Please sort commits into the following categories (you should not rename the categories!), I tried to pre-sort these to ease your work, feel free to move commits around if the current categorization is not good. +* Anything that is not public facing needs to be removed. +* If anything is miscategorized/belongs to another domain, move it to `miscategorized.md`. +* Please scan through `miscategorized.md` and handle any commits that belong within your domain according to these instructions. +* We place a lot of emphasis on the “BC-breaking” and “deprecation” sections. Those should be where the most effort goes in. The “improvements” and “bug fixes” for Python API should be nice as well. +* Once you are finished, move this very file from `todo/` to `done/` and submit a pull request. + +The categories below are as follows: + +* BC breaking: All commits that are BC-breaking. These are the most important commits. If any pre-sorted commit is actually BC-breaking, do move it to this section. Each commit should contain a paragraph explaining the rational behind the change as well as an example for how to update user code [BC-Guidelines](https://docs.google.com/document/d/14OmgGBr1w6gl1VO47GGGdwrIaUNr92DFhQbY_NEk8mQ/edit#heading=h.a9htwgvvec1m). +* Deprecations: All commits introducing deprecation. Each commit should include a small example explaining what should be done to update user code. +* new_features: All commits introducing a new feature (new functions, new submodule, new supported platform etc) +* improvements: All commits providing improvements to existing feature should be here (new backend for a function, new argument, better numerical stability) +* bug fixes: All commits that fix bugs and behaviors that do not match the documentation +* performance: All commits that are added mainly for performance (we separate this from improvements above to make it easier for users to look for it) +* documentation: All commits that add/update documentation +* Developers: All commits that are not end-user facing but still impact people that compile from source, develop into pytorch, extend pytorch, etc +* not user facing: All commits that are not public end-user facing and hence should be dropped from the release notes + +## export +### bc breaking +- `strict=False` is set as the default in `torch.export.export` and `export_for_training`. ([#148790](https://github.com/pytorch/pytorch/pull/148790), [#150941](https://github.com/pytorch/pytorch/pull/150941)) +- Remove `torch.export.export_for_inference` in favor of doing `torch.export.export_for_training().run_decompositions()`. ([#149078](https://github.com/pytorch/pytorch/pull/149078)) +### deprecation +### new features +- A new version of export `draft-export` -- https://docs.pytorch.org/docs/main/draft_export.html ([#152637](https://github.com/pytorch/pytorch/pull/152637), [#153219](https://github.com/pytorch/pytorch/pull/153219), [#149465](https://github.com/pytorch/pytorch/pull/149465), [#153627](https://github.com/pytorch/pytorch/pull/153627), [#154190](https://github.com/pytorch/pytorch/pull/154190), [#155744](https://github.com/pytorch/pytorch/pull/155744), [#150876](https://github.com/pytorch/pytorch/pull/150876), [#150948](https://github.com/pytorch/pytorch/pull/150948), [#151051](https://github.com/pytorch/pytorch/pull/151051), [#151065](https://github.com/pytorch/pytorch/pull/151065), [#150809](https://github.com/pytorch/pytorch/pull/150809), [#151797](https://github.com/pytorch/pytorch/pull/151797)) +- Introduce `AdditionalInputs` to specify dynamic shapes -- https://docs.pytorch.org/docs/main/export.html#torch.export.dynamic_shapes.AdditionalInputs ([#150144](https://github.com/pytorch/pytorch/pull/150144), [#151970](https://github.com/pytorch/pytorch/pull/151970)) +- Add min/max ranges for dim hints ([#149590](https://github.com/pytorch/pytorch/pull/149590)) +- Allow registering normal classes to `pytree.register_dataclass` ([#147752](https://github.com/pytorch/pytorch/pull/147752)) +- Allow specifying integer inputs as dynamic ([#151842](https://github.com/pytorch/pytorch/pull/151842)) + +### improvements +- Improve error message on constraint violation error ([#155738](https://github.com/pytorch/pytorch/pull/155738), [#152924](https://github.com/pytorch/pytorch/pull/152924), [#155603](https://github.com/pytorch/pytorch/pull/155603), [#151407](https://github.com/pytorch/pytorch/pull/151407)) +- Support python assertion with symints. ([#149444](https://github.com/pytorch/pytorch/pull/149444)) +- Fix tensor_constant and buffer naming conflicts in TS converter ([#148803](https://github.com/pytorch/pytorch/pull/148803)) +- Update remove runtime asserts pass ([#149198](https://github.com/pytorch/pytorch/pull/149198)) +- Add meta[val] to getattr nodes ([#154934](https://github.com/pytorch/pytorch/pull/154934)) +- Preserve custom metadata for tensor constants ([#152241](https://github.com/pytorch/pytorch/pull/152241)) +- Preserve custom meta in placeholders ([#149661](https://github.com/pytorch/pytorch/pull/149661)) +- Handle non OpNamespace type during decomposition. ([#149431](https://github.com/pytorch/pytorch/pull/149431)) +- Add mark_compiled_region support ([#149296](https://github.com/pytorch/pytorch/pull/149296)) +- Raise error when Dim.DYNAMIC 0/1 specializes ([#150716](https://github.com/pytorch/pytorch/pull/150716)) +- Warn when Dim.AUTO 0/1 specializes ([#151827](https://github.com/pytorch/pytorch/pull/151827)) +- Check tuple length mismatch for dynamic_shapes spec ([#150976](https://github.com/pytorch/pytorch/pull/150976)) +- Suggest dynamic re-export in input constraints hook ([#151624](https://github.com/pytorch/pytorch/pull/151624)) +- Improve handling of builtin ops (min, max, math.pow) ([#151348](https://github.com/pytorch/pytorch/pull/151348)) +- Add `from_node` metadata for nodes in gm.module() ([#155053](https://github.com/pytorch/pytorch/pull/155053)) + +### bug fixes +- Fix aten.is_nonzero for more than one elem tensors ([#149637](https://github.com/pytorch/pytorch/pull/149637)) +- Fix dynamic_shapes spec for moco ([#148772](https://github.com/pytorch/pytorch/pull/148772)) +- Fix ival swap in unflattener ([#149206](https://github.com/pytorch/pytorch/pull/149206)) +- Fix dynamic shapes repordering bug ([#149528](https://github.com/pytorch/pytorch/pull/149528)) +- Fix subclass access custom op bug ([#149698](https://github.com/pytorch/pytorch/pull/149698)) +- Patch dynamo configs when nonstrict tracing ([#149295](https://github.com/pytorch/pytorch/pull/149295)) +- Fix range constraints for expr ([#150103](https://github.com/pytorch/pytorch/pull/150103)) +- Fix multidimensional slicing ([#150104](https://github.com/pytorch/pytorch/pull/150104)) +- Fix deserialization of None inuts ([#150515](https://github.com/pytorch/pytorch/pull/150515)) +- Fix propagating unbacked symint in AOTI lowering ([#150570](https://github.com/pytorch/pytorch/pull/150570)) +- Expand `allowed_getattr_types` to include torch.Tensor ([#150867](https://github.com/pytorch/pytorch/pull/150867)) +- Fix aten.div type promotion for FakeTensor ([#150874](https://github.com/pytorch/pytorch/pull/150874)) +- Fix implicit state dict modification ([#151436](https://github.com/pytorch/pytorch/pull/151436)) +- Support SymInt minlength for torch.bincount() ([#152497](https://github.com/pytorch/pytorch/pull/152497)) +- Ignore None buffers ([#152571](https://github.com/pytorch/pytorch/pull/152571)) +- Fix None outputs in unflattener ([#153000](https://github.com/pytorch/pytorch/pull/153000)) +- Support functools.partial forward in non-strict ([#153408](https://github.com/pytorch/pytorch/pull/153408)) +- Support no inputs to unflattened module ([#153474](https://github.com/pytorch/pytorch/pull/153474)) +- Remove unused constants instead of lifting them ([#153800](https://github.com/pytorch/pytorch/pull/153800)) +- Avoid float/bool specialization for scalar tensor construction ([#154661](https://github.com/pytorch/pytorch/pull/154661)) +- Add math module for deserialization ([#154643](https://github.com/pytorch/pytorch/pull/154643)) +- Fix serialization for call_torchbind hop with as_none argument ([#155647](https://github.com/pytorch/pytorch/pull/155647)) +- Remove broken check for multiple cpp files in PT2 package ([#155149](https://github.com/pytorch/pytorch/pull/155149)) +- Handle aten.to at submodule boundaries ([#153972](https://github.com/pytorch/pytorch/pull/153972)) +- Preserve Enum types during torch.export serialization and deserialization ([#154821](https://github.com/pytorch/pytorch/pull/154821)) + +### performance +- Cache unflattened gm ([#150030](https://github.com/pytorch/pytorch/pull/150030)) + +### docs +- Add Mini tutorial for provenance tracking ([#152211](https://github.com/pytorch/pytorch/pull/152211)) +- Fix outdated docstring of torch.export.export regarding strict flag ([#149077](https://github.com/pytorch/pytorch/pull/149077)) +- Pretty print graph signature ([#149710](https://github.com/pytorch/pytorch/pull/149710)) +- Fix spelling mistake ([#155495](https://github.com/pytorch/pytorch/pull/155495)) +- Fix typos in docstring ([#155485](https://github.com/pytorch/pytorch/pull/155485)) +- pyfmt lint more export files ([#155783](https://github.com/pytorch/pytorch/pull/155783), [#154485](https://github.com/pytorch/pytorch/pull/154485), [#154487](https://github.com/pytorch/pytorch/pull/154487), [#154488](https://github.com/pytorch/pytorch/pull/154488)) +- Better error message for schema check in torch.export.load ([#156361](https://github.com/pytorch/pytorch/pull/156361)) +- Update docs for Dims ([#156262](https://github.com/pytorch/pytorch/pull/156262)) +- Update docs for ExportGraphSiganture ([#156244](https://github.com/pytorch/pytorch/pull/156244)) + +### devs +- Add TracingContext ([#149294](https://github.com/pytorch/pytorch/pull/149294)) +- Monkeypatch fake mode so it errors on invalid custom ops ([#149410](https://github.com/pytorch/pytorch/pull/149410)) +- Fix torch export docs for preserve_module_call_signature ([#151140](https://github.com/pytorch/pytorch/pull/151140)) +- Improve error message for deserializing custom triton op ([#152029](https://github.com/pytorch/pytorch/pull/152029)) +- Better type annotation for lift_constants_pass ([#152072](https://github.com/pytorch/pytorch/pull/152072)) +- Refactor `InputAdapter` (#152459) ([#152575](https://github.com/pytorch/pytorch/pull/152575)) +- Swap functorch --> torch._higher_order_ops ([#152620](https://github.com/pytorch/pytorch/pull/152620)) + +### Untopiced + + +### not user facing +- [export] set is_exporting() for strict ([#151833](https://github.com/pytorch/pytorch/pull/151833)) +- [export][cond] support merging constant ints as unbacked symint ([#152742](https://github.com/pytorch/pytorch/pull/152742)) +- Support exporting hop with function schema argument ([#152073](https://github.com/pytorch/pytorch/pull/152073)) +- Minor refactor to trace.py ([#149240](https://github.com/pytorch/pytorch/pull/149240)) +- [export] Make aoti_call_delegate hop traceable ([#148804](https://github.com/pytorch/pytorch/pull/148804)) +- [DRAFT] Initial version of sticky export ([#151047](https://github.com/pytorch/pytorch/pull/151047)) +- GPU lowering uses aoti_call_delegate ([#153282](https://github.com/pytorch/pytorch/pull/153282)) +### security diff --git a/2.8.0/todo/result_composability.md b/2.8.0/todo/result_composability.md index d944ddb..c788b05 100644 --- a/2.8.0/todo/result_composability.md +++ b/2.8.0/todo/result_composability.md @@ -124,5 +124,7 @@ The categories below are as follows: - Remove guard_size_oblivious from is_nonzero proxy call check ([#154164](https://github.com/pytorch/pytorch/pull/154164)) - Fix evaluate_expr to include suppress_guards_tls in cache key ([#152661](https://github.com/pytorch/pytorch/pull/152661)) - [aotd] Support mutations of the same input in fw and bw ([#155354](https://github.com/pytorch/pytorch/pull/155354)) +- [multigraph] use specializations in compile_and_call_fx_graph ([#153449](https://github.com/pytorch/pytorch/pull/153449)) +- Add guard_or_false for computeStorageNbytes ([#150483](https://github.com/pytorch/pytorch/pull/150483)) ### security diff --git a/2.8.0/todo/result_export.md b/2.8.0/todo/result_export.md deleted file mode 100644 index bf876fc..0000000 --- a/2.8.0/todo/result_export.md +++ /dev/null @@ -1,128 +0,0 @@ - -# Release Notes worksheet export - -The main goal of this process is to rephrase all the commit messages below to make them **clear and easy to read** by the end user. You should follow the following instructions to do so: - -* **Please clean up and format commit titles to be readable by the general PyTorch user.** Make sure you're [following the guidance here](https://docs.google.com/document/d/14OmgGBr1w6gl1VO47GGGdwrIaUNr92DFhQbY_NEk8mQ/edit)! Your resulting notes must be consistent and easy to read. -* Please sort commits into the following categories (you should not rename the categories!), I tried to pre-sort these to ease your work, feel free to move commits around if the current categorization is not good. -* Anything that is not public facing needs to be removed. -* If anything is miscategorized/belongs to another domain, move it to `miscategorized.md`. -* Please scan through `miscategorized.md` and handle any commits that belong within your domain according to these instructions. -* We place a lot of emphasis on the “BC-breaking” and “deprecation” sections. Those should be where the most effort goes in. The “improvements” and “bug fixes” for Python API should be nice as well. -* Once you are finished, move this very file from `todo/` to `done/` and submit a pull request. - -The categories below are as follows: - -* BC breaking: All commits that are BC-breaking. These are the most important commits. If any pre-sorted commit is actually BC-breaking, do move it to this section. Each commit should contain a paragraph explaining the rational behind the change as well as an example for how to update user code [BC-Guidelines](https://docs.google.com/document/d/14OmgGBr1w6gl1VO47GGGdwrIaUNr92DFhQbY_NEk8mQ/edit#heading=h.a9htwgvvec1m). -* Deprecations: All commits introducing deprecation. Each commit should include a small example explaining what should be done to update user code. -* new_features: All commits introducing a new feature (new functions, new submodule, new supported platform etc) -* improvements: All commits providing improvements to existing feature should be here (new backend for a function, new argument, better numerical stability) -* bug fixes: All commits that fix bugs and behaviors that do not match the documentation -* performance: All commits that are added mainly for performance (we separate this from improvements above to make it easier for users to look for it) -* documentation: All commits that add/update documentation -* Developers: All commits that are not end-user facing but still impact people that compile from source, develop into pytorch, extend pytorch, etc -* not user facing: All commits that are not public end-user facing and hence should be dropped from the release notes - -## export -### bc breaking -- Set non-strict export as default mode ([#148790](https://github.com/pytorch/pytorch/pull/148790)) -- Remove torch.export.export_for_inference ([#149078](https://github.com/pytorch/pytorch/pull/149078)) -- [BC-breaking] Set NonStrict as default for export_for_training ([#150941](https://github.com/pytorch/pytorch/pull/150941)) -### deprecation -### new features -- Draft-export ([#149465](https://github.com/pytorch/pytorch/pull/149465), [#153627](https://github.com/pytorch/pytorch/pull/153627), [#154190](https://github.com/pytorch/pytorch/pull/154190), [#155744](https://github.com/pytorch/pytorch/pull/155744)) - -### improvements -- Improve error message on constraint violation error ([#155738](https://github.com/pytorch/pytorch/pull/155738), [#152924](https://github.com/pytorch/pytorch/pull/152924), [#155603](https://github.com/pytorch/pytorch/pull/155603), [#151407](https://github.com/pytorch/pytorch/pull/151407)) -- Add meta[val] to getattr nodes ([#154934](https://github.com/pytorch/pytorch/pull/154934)) -- Support python assertion with symints. ([#149444](https://github.com/pytorch/pytorch/pull/149444)) - -### bug fixes -- Fix aten.is_nonzero for more than one elem tensors ([#149637](https://github.com/pytorch/pytorch/pull/149637)) - -### performance -### docs -### devs -### Untopiced -- [export] Fix tensor_constant and buffer naming conflicts in TS converter ([#148803](https://github.com/pytorch/pytorch/pull/148803)) -- Fix outdated docstring of torch.export.export regarding strict flag ([#149077](https://github.com/pytorch/pytorch/pull/149077)) -- [torchbench] fix dynamic_shapes spec for moco ([#148772](https://github.com/pytorch/pytorch/pull/148772)) -- [export] Minor refactor to trace.py ([#149240](https://github.com/pytorch/pytorch/pull/149240)) -- [export] Update remove runtime asserts pass ([#149198](https://github.com/pytorch/pytorch/pull/149198)) -- [export] Add TracingContext ([#149294](https://github.com/pytorch/pytorch/pull/149294)) -- debug ival swap ([#149206](https://github.com/pytorch/pytorch/pull/149206)) -- Fix dynamic shapes repordering bug ([#149528](https://github.com/pytorch/pytorch/pull/149528)) -- Monkeypatch fake mode so it errors on invalid custom ops ([#149410](https://github.com/pytorch/pytorch/pull/149410)) -- [export] Handle non OpNamespace type during decomposition. ([#149431](https://github.com/pytorch/pytorch/pull/149431)) -- preserve custom meta in placeholders ([#149661](https://github.com/pytorch/pytorch/pull/149661)) -- Fix subclass access custom op bug ([#149698](https://github.com/pytorch/pytorch/pull/149698)) -- pretty print graph signature ([#149710](https://github.com/pytorch/pytorch/pull/149710)) -- [export] Patch dynamo configs when nonstrict tracing ([#149295](https://github.com/pytorch/pytorch/pull/149295)) -- [export] Add mark_compiled_region support ([#149296](https://github.com/pytorch/pytorch/pull/149296)) -- [export] Save unflattened gm ([#150030](https://github.com/pytorch/pytorch/pull/150030)) -- fix range constraints for expr ([#150103](https://github.com/pytorch/pytorch/pull/150103)) -- [reland] Support torchbind in OSS proxy executor ([#150196](https://github.com/pytorch/pytorch/pull/150196)) -- [export] min/max ranges for dim hints ([#149590](https://github.com/pytorch/pytorch/pull/149590)) -- infer dynamic shapes through additional inputs ([#150144](https://github.com/pytorch/pytorch/pull/150144)) -- [pytree] Register normal class to register_dataclass ([#147752](https://github.com/pytorch/pytorch/pull/147752)) -- multidimensional slicing ([#150104](https://github.com/pytorch/pytorch/pull/150104)) -- [export] Fix deserialization issue ([#150515](https://github.com/pytorch/pytorch/pull/150515)) -- [aoti] Fix cannot determine truth value of Relation error when propagating unbacked symint in lowering ([#150570](https://github.com/pytorch/pytorch/pull/150570)) -- [export] Make aoti_call_delegate hop traceable ([#148804](https://github.com/pytorch/pytorch/pull/148804)) -- [export] raise when Dim.DYNAMIC 0/1 specializes ([#150716](https://github.com/pytorch/pytorch/pull/150716)) -- Expand allowed_getattr_types_for_subgm to torch.Tensor ([#150867](https://github.com/pytorch/pytorch/pull/150867)) -- Fix aten.div type promotion for FakeTensor ([#150874](https://github.com/pytorch/pytorch/pull/150874)) -- [export] Refine draft-export CVE with Dim.AUTO ([#150876](https://github.com/pytorch/pytorch/pull/150876)) -- Add real_tensor to the FakeTensor in node.meta["val"] ([#150948](https://github.com/pytorch/pytorch/pull/150948)) -- [export] check tuple length mismatch for dynamic_shapes spec ([#150976](https://github.com/pytorch/pytorch/pull/150976)) -- [export] Make draft-export predispatch=True by default ([#151051](https://github.com/pytorch/pytorch/pull/151051)) -- [export] Add draft-export to error msg ([#151065](https://github.com/pytorch/pytorch/pull/151065)) -- [export] Integrate meta kernel generation with draft-export ([#150809](https://github.com/pytorch/pytorch/pull/150809)) -- Fix implicit state dict modification ([#151436](https://github.com/pytorch/pytorch/pull/151436)) -- [doc fix] fix torch export docs for preserve_module_call_signature ([#151140](https://github.com/pytorch/pytorch/pull/151140)) -- [export] suggest dynamic re-export in input constraints hook ([#151624](https://github.com/pytorch/pytorch/pull/151624)) -- Rename register_fake_profile to unsafe_generate_fake_kernels ([#151797](https://github.com/pytorch/pytorch/pull/151797)) -- Infra for handling builtin ops (min, max, math.pow) ([#151348](https://github.com/pytorch/pytorch/pull/151348)) -- [export] Enable symint inputs for AdditionalInputs and ShapesCollection ([#151842](https://github.com/pytorch/pytorch/pull/151842)) -- [DRAFT] INitial version of sticky export ([#151047](https://github.com/pytorch/pytorch/pull/151047)) -- [export] improve error message for deserializing custom triton op ([#152029](https://github.com/pytorch/pytorch/pull/152029)) -- [export] Preserve custom metadata for tensor constants ([#152241](https://github.com/pytorch/pytorch/pull/152241)) -- Fix additional inputs to error on inconsistent constants ([#151970](https://github.com/pytorch/pytorch/pull/151970)) -- [export] support SymInt minlength for torch.bincount() ([#152497](https://github.com/pytorch/pytorch/pull/152497)) -- [export][be] better type annotation for lift_constants_pass ([#152072](https://github.com/pytorch/pytorch/pull/152072)) -- [export] warn when Dim.AUTO 0/1 specializes ([#151827](https://github.com/pytorch/pytorch/pull/151827)) -- [export] Ignore None buffers ([#152571](https://github.com/pytorch/pytorch/pull/152571)) -- BE: Swap functorch --> torch._higher_order_ops ([#152620](https://github.com/pytorch/pytorch/pull/152620)) -- [IR] Input Adapter refactor prototype (#152459) ([#152575](https://github.com/pytorch/pytorch/pull/152575)) -- [export] Add draft-export docs ([#152637](https://github.com/pytorch/pytorch/pull/152637)) -- [export] Unflatten None ([#153000](https://github.com/pytorch/pytorch/pull/153000)) -- Mini tutorial for provenance tracking ([#152211](https://github.com/pytorch/pytorch/pull/152211)) -- [dynamic shapes] guard_or_false for computeStorageNbytes ([#150483](https://github.com/pytorch/pytorch/pull/150483)) -- GPU lowering uses aoti_call_delegate ([#153282](https://github.com/pytorch/pytorch/pull/153282)) -- [export] support functools.partial forward (non-strict) ([#153408](https://github.com/pytorch/pytorch/pull/153408)) -- [export] Make draft_export public ([#153219](https://github.com/pytorch/pytorch/pull/153219)) -- [export] Support no inputs in unflattened module ([#153474](https://github.com/pytorch/pytorch/pull/153474)) -- [export] Remove unused constants ([#153800](https://github.com/pytorch/pytorch/pull/153800)) -- [export] avoid float/bool specialization for scalar tensor construction ([#154661](https://github.com/pytorch/pytorch/pull/154661)) -- [Export] Add math module for deserialization ([#154643](https://github.com/pytorch/pytorch/pull/154643)) -- Adding from_node for nodes in gm.module() ([#155053](https://github.com/pytorch/pytorch/pull/155053)) -- Preserve Enum types during torch.export serialization and deserialization ([#154821](https://github.com/pytorch/pytorch/pull/154821)) -- Fix spelling mistake ([#155495](https://github.com/pytorch/pytorch/pull/155495)) -- [Export] Fix some typos in docstring ([#155485](https://github.com/pytorch/pytorch/pull/155485)) -- [export] Fix serialization for call_torchbind hop with as_none argument ([#155647](https://github.com/pytorch/pytorch/pull/155647)) -- pyfmt lint more export files ([#155783](https://github.com/pytorch/pytorch/pull/155783)) -- [export] Remove broken check for multiple cpp files in PT2 package ([#155149](https://github.com/pytorch/pytorch/pull/155149)) -- Handle aten.to at submodule boundaries ([#153972](https://github.com/pytorch/pytorch/pull/153972)) -- [export] Update documents for ExportGraphSiganture ([#156244](https://github.com/pytorch/pytorch/pull/156244)) -- [ez][export] Better error message for schema check in torch.export.load ([#156361](https://github.com/pytorch/pytorch/pull/156361)) -- [export] update docs for Dims ([#156262](https://github.com/pytorch/pytorch/pull/156262)) -### not user facing -- [export] set is_exporting() for strict ([#151833](https://github.com/pytorch/pytorch/pull/151833)) -- [export][cond] support merging constant ints as unbacked symint ([#152742](https://github.com/pytorch/pytorch/pull/152742)) -- [export][cond] support merging constant ints as unbacked symint ([#152742](https://github.com/pytorch/pytorch/pull/152742)) -- pyfmt lint torch/_export/serde/serialize.py ([#154485](https://github.com/pytorch/pytorch/pull/154485)) -- pyfmt lint torch/_export/serde/schema.py ([#154487](https://github.com/pytorch/pytorch/pull/154487)) -- pyfmt lint torch/_export/passes/replace_view_ops_with_view_copy_ops_pass.py ([#154488](https://github.com/pytorch/pytorch/pull/154488)) -- [multigraph] use specializations in compile_and_call_fx_graph ([#153449](https://github.com/pytorch/pytorch/pull/153449)) -- Support exporting hop with function schema argument ([#152073](https://github.com/pytorch/pytorch/pull/152073)) -### security diff --git a/2.8.0/todo/result_inductor (aoti).md b/2.8.0/todo/result_inductor (aoti).md index d7967bf..577cecb 100644 --- a/2.8.0/todo/result_inductor (aoti).md +++ b/2.8.0/todo/result_inductor (aoti).md @@ -27,6 +27,7 @@ The categories below are as follows: ### bc breaking ### deprecation ### new features +- Torchbind objects supported in AOTInductor ([#150196](https://github.com/pytorch/pytorch/pull/150196), [#154265](https://github.com/pytorch/pytorch/pull/154265)) ### improvements - [MPS] Implement backward pass for interpolate_trilinear ([#156373](https://github.com/pytorch/pytorch/pull/156373)) ### bug fixes @@ -42,7 +43,6 @@ The categories below are as follows: - [AOTI][XPU] Refactor AOTInductor runtime API for Intel GPU. ([#153929](https://github.com/pytorch/pytorch/pull/153929)) - [AOTI][refactor] Fix an anonymous namespace issue ([#154033](https://github.com/pytorch/pytorch/pull/154033)) - [3/n][Optimus][Auto-AC][reland] Support any fp8 quantization type and set scaling as the default" ([#154057](https://github.com/pytorch/pytorch/pull/154057)) -- Fix torchbind path in AOTI package loader ([#154265](https://github.com/pytorch/pytorch/pull/154265)) - Use get_device_context in aoti runtime for XPU directly ([#154360](https://github.com/pytorch/pytorch/pull/154360)) - Add new ops in fallback ops ([#154251](https://github.com/pytorch/pytorch/pull/154251)) - [4/n][Optimus][Auto-AC] Expose the config to skip the dynamo gaurds to avoid recompile ([#154152](https://github.com/pytorch/pytorch/pull/154152))