This is an automated email from the ASF dual-hosted git repository.
tqchen pushed a change to branch unity-staging
in repository https://gitbox.apache.org/repos/asf/tvm.git
discard abc474887b [TVMScript][Unity] Improve PyLint Compatibility (#14276)
discard 536498b17a [Unity][ci] Use CPU-SMALL instances (#14256)
discard aaf00080e5 [Unity] Introduce call_dps_packed (#14183)
discard 0e56872c20 [Unity] Consider target context for Relay to Relax
conversion (#14269)
discard e27508de6e [Unity][Frontend] Import `tanh` and fix `layer_norm`
(#14247)
discard 2c205e6ea0 [Unity][BYOC] Add conv2d and residual block patterns for
Relax cutlass BYOC (#14252)
discard 229bfaca1b [Unity] Allow user defined func attrs in emit_te (#14255)
discard 73ab563208 [Unity][Op] Add repeat, tile, conv2d_transpose, avg_pool2d
(#14238)
discard 48eb2582e1 [Unity][Op][Tweak] Improve `StructInfo` inference for
`shape_of` (#14243)
discard af1792ebb9 [Unity][WEB] Improve ndarray cache (#14236)
discard e0ec19adfa [Unity][WEB] Update text prompts for syntactical
correctness (#14237)
discard c77caeb65a [Unity][TVMScript] Fix prim_func lost issue in
relax.emit_te (#14189)
discard 70ecb435be [Unity][TVMScript] Enable Context-Aware Parsing (#14234)
discard 40ca12b6e1 [Unity][Bugfix] Do not include `PrimFunc`s in the
dependency graph when checking for recursion (#14228)
discard 62bb5b7630 [Unity][Transform] SimplifyNormInference (#14221)
discard 2d6e528e1c [Unity] Improve implementation of FuseOps (#14229)
discard f055ab90db [Unity] ensure memory.alloc_tensor/storage roundtrippable
(#14226)
discard 8fbd14d7b1 [Unity][WEB] Simplify WebGPU Codegen per spec (#14225)
discard eb0faae8d3 [Unity][Transform] Memory plan across the IRModule (#14220)
discard 8379ed97ef [Unity][BYOC] Add dynamic shape support to CUTLASS matmul
(#14216)
discard 2d03ca5735 [Unity][Frontend] from_fx keeps parameters in order (#14214)
discard 02a824a56a [Unity][WEB] Improve webgpu codegen options to skip
readonly (#14213)
discard 60ca6b0aca [Unity][Frontend] FX translator supports unwrapping unit
return tuple (#14212)
discard 7f33052905 [Unity][Frontend] Attach imported model weights, deprecate
ImporterOutput (#14211)
discard 4ae250cb26 [Unity] Introduce Default GPU Schedule Pass (#14182)
discard f7bc435ad3 [Unity][Frontend] FX translator support torch.baddbmm
(#14202)
discard 433e126bda [Unity][TIR][Pass] ForceNarrowIndexToInt32 (#14203)
discard 4b89fa518e [Unity][Fix] FX translating dtype (#14201)
discard bf4359f174 [Unity][Frontend] FX translator returning weights with
`keep_params_as_input` (#14197)
discard f8a3784815 [Unity][Frontend] FX translator supporting more ops (#14196)
discard 7205c0c255 [Unity][Op] Legalize `round`, `floor`, `ceil`, `sign`
(#14198)
discard 0f40563fed [Unity][Op] Argmax and argmin (#14195)
discard 823bedc8ff [Unity][Op] Group normalization (#14194)
discard 10df00c36d [Unity][Transform] LiftTransformParams handling multiple
functions (#14192)
discard d72f6fc919 [Unity][WEBGPU] Codegen improvements and WebRuntime (#14187)
discard cc9afb3bba [Unity][OP] Add an operator for fused multi head attention
(#14150)
discard 0004320e82 [Unity][Analysis] Restore Python bindings for var analyses
(#14180)
discard 5d7adf90c9 [Unity][Op] Full support of Relax op `power` (#14171)
discard 80cba01e4a [Unity][BYOC] Add batch matmul support to Relax CUTLASS
BYOC (#14166)
discard 4b681c805e [Unity][Analysis] Analysis for detecting recursion in Relax
(#14149)
discard 1cbc233f35 [Unity] Add bind_constants option to FuseOpsByPattern
(#14151)
discard c8ec0ffd47 [Unity][BYOC] Use Relax legalize + CPU build for reference
in tests (#14162)
discard d3c9e3c72b [Unity][Analysis] Checking function return struct info in
well-formed check (#14155)
discard bab52c564f [Unity][Pass] Support Symbolic Shape Deduction during
BindParam (#14154)
discard c72af8043e [Unity][Debugging] AST printer (#14152)
discard 02ce8fa5ae [Unity][Pass] Enhance constant folding to fold relax ops by
evaluating them. (#14146)
discard 3103bc643e [Unity][Legalize] Fix Scalar Constant Legalization (#14127)
discard fbb6fba774 [Unity] Add callback to FuseOpsByPattern to check match
result is accepted (#14109)
discard 15d30ba845 [Unity][BYOC] Assign group to unused bindings and ignroe
PrimFunc (#14139)
discard 055e2adada [Unity][TVMScript] emit_te sugar (#14123)
discard 25b00b9b19 [Unity][BYOC] Add transposed matmul support to Relax
CUTLASS BYOC (#14128)
discard a54001759d [Unity] Add Global info (#14132)
discard 48c309e3e3 [Unity][WEB] Relax vm on web runtime (#14131)
discard 8ed0b4d8b8 [Unity][BlockBuilder] Add `name_hint` argument for `emit`
and `emit_output` (#14126)
discard d0a8113593 [Unity][Fix] Fix bug in MergeCompositeFunctions (#14117)
discard 04ee096572 [Unity] Update tests again to adapt to latest TVMScript
syntax (#14115)
discard 4e9223c18c [Unity][BYOC]Add relax backend pattern registry (#14106)
discard 08fb0046d1 [Unity] Remove attributes of relax.print, assert and unique
(#14101)
discard a15cbb31bc [Unity][Layout] Add layout transformation analysis for
PrimFunc (#14066)
discard c281556950 [Unity] Relax Recursive function (#14092)
discard b070703c44 [Unity] Lower `shape_of` to a builtin (#14093)
discard 8369c84a2b [Unity] Fix typo in the comment (#14096)
discard 58c51b2c50 [Unity][Relax] Set Shape Function to Be Host Function
(#14090)
discard 8627e158f3 [Unity] Refactor Relax Build JIT UX (#14088)
discard b7d09e394f [Unity][Fix][Pass] FoldConstant with DCE in dataflow block
(#14087)
discard 74343978e9 [Unity][Analysis] TIR pattern kind analysis for
multi-buffer write block (#14075)
discard 3fafeb2089 [Unity][Op] `log_softmax` and `cross_entropy_with_logits`
(#14083)
discard 6758d4222f [Unity][BYOC] Add DNNL backend (#14082)
discard 336fb434e5 [Unity][BYOC] Add CUTLASS backend (#14081)
discard c6158401ee [Unity] Add testcases for `expr_args_converter` (#14080)
discard fd2cfc5f2a [Unity][Pass] Canonicalize Bindings (#14079)
discard 2fa1246a6e [Unity][BYOC][Pass] RunCodegen and TensorRT (#14078)
discard 8700dfc00b [Unity][Transform] Add LiftTransformParams pass (#14069)
discard 680c811b86 [Unity][Frontend] Annotate number of non-static input of FX
function (#14067)
discard 8a46f67af1 [Unity][BYOC] Add pass to merge composite functions to
offload large subgraphs (#14062)
discard f04810b118 [Unity][Pass] Remove Unused Function (#14061)
discard 8efdb5bfe7 [Unity][Fix][Pass] Fix FuseOps for lack graph edges (#14058)
discard 3736c5fcdf [Unity] Relax op: collapse sum (#14059)
discard 96d21260bf [Unity][BYOC] Add pattern-based partitioning pass (#14054)
discard 8432c1ff21 [Unity][VM] Add per-op profiling support (#14053)
discard 5a33692494 [Unity][TVMScript] Overload `__neg__` for relax expr
(#14045)
discard c8455a23e0 [Unity][Pass] FuseOps FuseTIR fixes (#14044)
discard 604fd4ce94 [Unity] Statement rewriter for DataflowBlock (#14043)
discard 9166b5c7b6 [Unity] Relax dataflow pattern language (matching) (#14041)
discard bce1fd0514 [Unity] Update tests to adapt to latest TVMScript syntax
(#14039)
discard 444e1ee26e [Unity] Disallow inline prim_func in relax IR (#14040)
discard 889749014a [Unity][Pass] Block-level static memory planning (#14038)
discard a063fa2b6b [Unity] Initial PyTorch Frontend (#14037)
discard 080ac05f49 [Unity][Op] Add ShapeExpr Tests for Reshape Op (#14035)
discard 14f16d5994 [Unity][Pass] Operator legalization (#14029)
discard 286f737481 [Unity][TVMScript] Move tir/relax import in script out of
__init__.py (#14033)
discard 2f148e383b [Unity][Pass] Wellformed Analysis (#14032)
discard da584fadc5 [Unity][BlockBuilder] CallTE convert PrimValue args
(#14028)
discard ff44563d87 [Unity][Pass] Normalize Pass (#14031)
discard 4eb8887636 [Unity] Relay -> Relax translator (#14026)
discard 6107a8c018 [Unity][Pass][TuningAPI] Introduce TuningAPI and
MetaSchedule pass (#14014)
discard b96dbcec2d [Unity][Pass] BindParams pass, FoldConstant pass (#14016)
discard fb8260aa0d [Unity][VM] Supporting "compiled" exec mode. (#14015)
discard bfd2ca073c [Unity][Pass] LambdaLift pass (#14012)
discard 2b6cc7b35b [Unity][Pass] Operator Fusion Passes (#14001)
discard 17b4090846 [Unity] NestedMsg Support utility (#13995)
discard 6ec3ca0422 [Unity] Relax op: manipulation (#13989)
discard f2ae9fc2ce [Unity] Relax op: search (#13992)
discard 8a8674f242 [Unity] Relax op: linear algebra (#13988)
discard 25d66c0b62 [Unity] Relax op: creation (#13984)
discard 1525c64151 [Unity] Relax op: neural networks (#13993)
discard b94c6e4f11 [Unity] Relax op: statistical (#13991)
discard 73e2cb1237 [Unity] Relax op: arithmetic, comparison (#13983)
discard ec01577489 [Unity] Relax op: image (#13994)
discard 20995b420d [Unity] Relax op: set (#13990)
discard 47106a3250 [Unity] Relax op: datatype (#13986)
discard 57800b3103 [Unity] Relax op: index (#13987)
discard 3e8560e4bc [Unity][TVMScript] Use explicit `R.shape` in TVMScript
(#13979)
discard dc52afb470 [Unity] e2e Relax minimum build flow (#13961)
discard 84b9b1dd3d [Unity] Relax VM shape lowering pass (#13956)
discard 70c466d1fe [Unity] Relax VM codegen (#13954)
discard 974d00c535 [Unity] Relax TVMScript Printer (#13944)
discard 60ae9aca7f [Unity] Relax TVMScript Parser. (#13932)
discard 187d5f87f2 [Unity] Relax BlockBuilder and ExprMutator (#13926)
discard c1a0907f60 [Unity] Basic StructInfo Analysis and Expr construction
(#13916)
discard fff7f9201a [Unity][CI] Unity specific jenkins setup (do not upstream
to main) (#13910)
discard efd9a8a0ad [Unity][IR] First-class StructInfo (#13907)
discard e9b9218cc4 [Unity] Relax expressions and types (#13901)
discard 99b5cdd22a [Unity] Relax VM (#13878)
add 06fabe4c5a [PaddlePaddle Hackathon 4][Frontend][Paddle]add
grid-sample/gaussian_random/flip/fill_zeros_like/unique for paddle frontend
(#14277)
add 9a99fc89a2 [Utils] Allow classmethod and staticmethod in
TVMDerivedObject (#14249)
new 5292e3b314 [Unity] Relax VM (#13878)
new 46ec39c8ad [Unity] Relax expressions and types (#13901)
new e9014d90ad [Unity][IR] First-class StructInfo (#13907)
new c35aedbd2d [Unity][CI] Unity specific jenkins setup (do not upstream
to main) (#13910)
new 59d7ff5931 [Unity] Basic StructInfo Analysis and Expr construction
(#13916)
new b8631adebc [Unity] Relax BlockBuilder and ExprMutator (#13926)
new 9fcc29ebfe [Unity] Relax TVMScript Parser. (#13932)
new 897cc206d9 [Unity] Relax TVMScript Printer (#13944)
new 26c0653220 [Unity] Relax VM codegen (#13954)
new 66bd67d432 [Unity] Relax VM shape lowering pass (#13956)
new 166d60c281 [Unity] e2e Relax minimum build flow (#13961)
new 6a62f21fba [Unity][TVMScript] Use explicit `R.shape` in TVMScript
(#13979)
new 6e6f222b44 [Unity] Relax op: index (#13987)
new 4c1b5b95a0 [Unity] Relax op: datatype (#13986)
new 41b811549d [Unity] Relax op: set (#13990)
new e28fe86403 [Unity] Relax op: image (#13994)
new 35daa96508 [Unity] Relax op: arithmetic, comparison (#13983)
new ba81ffa577 [Unity] Relax op: statistical (#13991)
new 1fa77f2f29 [Unity] Relax op: neural networks (#13993)
new 19c7925b57 [Unity] Relax op: creation (#13984)
new 1c51d92e96 [Unity] Relax op: linear algebra (#13988)
new 2affc5bb2b [Unity] Relax op: search (#13992)
new 388f9347fc [Unity] Relax op: manipulation (#13989)
new e6d1d2e2c0 [Unity] NestedMsg Support utility (#13995)
new 479a21cd31 [Unity][Pass] Operator Fusion Passes (#14001)
new 89bf330f34 [Unity][Pass] LambdaLift pass (#14012)
new 01a172e7ec [Unity][VM] Supporting "compiled" exec mode. (#14015)
new 4dd06e427e [Unity][Pass] BindParams pass, FoldConstant pass (#14016)
new f420250abf [Unity][Pass][TuningAPI] Introduce TuningAPI and
MetaSchedule pass (#14014)
new e3466b5478 [Unity] Relay -> Relax translator (#14026)
new 1e354c75c2 [Unity][Pass] Normalize Pass (#14031)
new 7ee0fe4fb6 [Unity][BlockBuilder] CallTE convert PrimValue args
(#14028)
new b6b1d272a9 [Unity][Pass] Wellformed Analysis (#14032)
new 966b874c84 [Unity][TVMScript] Move tir/relax import in script out of
__init__.py (#14033)
new 4c3c347a08 [Unity][Pass] Operator legalization (#14029)
new e5b0c6d879 [Unity][Op] Add ShapeExpr Tests for Reshape Op (#14035)
new f825714f5c [Unity] Initial PyTorch Frontend (#14037)
new bcb84f77eb [Unity][Pass] Block-level static memory planning (#14038)
new f2c3180d73 [Unity] Disallow inline prim_func in relax IR (#14040)
new 7fef9e3399 [Unity] Update tests to adapt to latest TVMScript syntax
(#14039)
new e1b5400e1b [Unity] Relax dataflow pattern language (matching) (#14041)
new ea105909ae [Unity] Statement rewriter for DataflowBlock (#14043)
new 50785f45b5 [Unity][Pass] FuseOps FuseTIR fixes (#14044)
new 9336baac97 [Unity][TVMScript] Overload `__neg__` for relax expr
(#14045)
new 34c0248626 [Unity][VM] Add per-op profiling support (#14053)
new 29f706f43c [Unity][BYOC] Add pattern-based partitioning pass (#14054)
new ec61975ba6 [Unity] Relax op: collapse sum (#14059)
new 7801803ee6 [Unity][Fix][Pass] Fix FuseOps for lack graph edges (#14058)
new 2b8a9cccd9 [Unity][Pass] Remove Unused Function (#14061)
new e5caf4085b [Unity][BYOC] Add pass to merge composite functions to
offload large subgraphs (#14062)
new a4b69ed1e2 [Unity][Frontend] Annotate number of non-static input of FX
function (#14067)
new 78ec7f1dea [Unity][Transform] Add LiftTransformParams pass (#14069)
new 9642340e8d [Unity][BYOC][Pass] RunCodegen and TensorRT (#14078)
new fd7c3cc197 [Unity][Pass] Canonicalize Bindings (#14079)
new 86490994c8 [Unity] Add testcases for `expr_args_converter` (#14080)
new db54a99138 [Unity][BYOC] Add CUTLASS backend (#14081)
new 255a2358fb [Unity][BYOC] Add DNNL backend (#14082)
new a115e9ecf1 [Unity][Op] `log_softmax` and `cross_entropy_with_logits`
(#14083)
new 30f004fe58 [Unity][Analysis] TIR pattern kind analysis for
multi-buffer write block (#14075)
new d12123a6cf [Unity][Fix][Pass] FoldConstant with DCE in dataflow block
(#14087)
new 9cb42187b8 [Unity] Refactor Relax Build JIT UX (#14088)
new ca18e9eacd [Unity][Relax] Set Shape Function to Be Host Function
(#14090)
new ff366ae4b8 [Unity] Fix typo in the comment (#14096)
new 0d9f3a3970 [Unity] Lower `shape_of` to a builtin (#14093)
new aad7502ceb [Unity] Relax Recursive function (#14092)
new 093bf6d521 [Unity][Layout] Add layout transformation analysis for
PrimFunc (#14066)
new 561ac3801d [Unity] Remove attributes of relax.print, assert and unique
(#14101)
new 0eabde5d89 [Unity][BYOC]Add relax backend pattern registry (#14106)
new 47f98f2da5 [Unity] Update tests again to adapt to latest TVMScript
syntax (#14115)
new b6224e3e99 [Unity][Fix] Fix bug in MergeCompositeFunctions (#14117)
new 3fa889dc2d [Unity][BlockBuilder] Add `name_hint` argument for `emit`
and `emit_output` (#14126)
new 5efb0b9311 [Unity][WEB] Relax vm on web runtime (#14131)
new ec0ec24653 [Unity] Add Global info (#14132)
new b1b76a573a [Unity][BYOC] Add transposed matmul support to Relax
CUTLASS BYOC (#14128)
new 6c15b4be3e [Unity][TVMScript] emit_te sugar (#14123)
new c5be69e08b [Unity][BYOC] Assign group to unused bindings and ignroe
PrimFunc (#14139)
new 267a92817c [Unity] Add callback to FuseOpsByPattern to check match
result is accepted (#14109)
new e9b62b1af8 [Unity][Legalize] Fix Scalar Constant Legalization (#14127)
new bb6eae0dfe [Unity][Pass] Enhance constant folding to fold relax ops by
evaluating them. (#14146)
new 8743e4032b [Unity][Debugging] AST printer (#14152)
new 34ba550eb4 [Unity][Pass] Support Symbolic Shape Deduction during
BindParam (#14154)
new 946cb26588 [Unity][Analysis] Checking function return struct info in
well-formed check (#14155)
new 17dff73ad1 [Unity][BYOC] Use Relax legalize + CPU build for reference
in tests (#14162)
new 08e8cf88e3 [Unity] Add bind_constants option to FuseOpsByPattern
(#14151)
new fbc8721c31 [Unity][Analysis] Analysis for detecting recursion in Relax
(#14149)
new 7cf4933a16 [Unity][BYOC] Add batch matmul support to Relax CUTLASS
BYOC (#14166)
new 2862549222 [Unity][Op] Full support of Relax op `power` (#14171)
new e4eb1f6ee3 [Unity][Analysis] Restore Python bindings for var analyses
(#14180)
new 240621d736 [Unity][OP] Add an operator for fused multi head attention
(#14150)
new ae321c51e7 [Unity][WEBGPU] Codegen improvements and WebRuntime (#14187)
new 27322e59a8 [Unity][Transform] LiftTransformParams handling multiple
functions (#14192)
new c9f64c6415 [Unity][Op] Group normalization (#14194)
new 45487af1da [Unity][Op] Argmax and argmin (#14195)
new 111495db36 [Unity][Op] Legalize `round`, `floor`, `ceil`, `sign`
(#14198)
new bbe8edd758 [Unity][Frontend] FX translator supporting more ops (#14196)
new 798bb66e15 [Unity][Frontend] FX translator returning weights with
`keep_params_as_input` (#14197)
new 497e9f6c7c [Unity][Fix] FX translating dtype (#14201)
new a62f0635c0 [Unity][TIR][Pass] ForceNarrowIndexToInt32 (#14203)
new 92486a8a0d [Unity][Frontend] FX translator support torch.baddbmm
(#14202)
new 3ea14402aa [Unity] Introduce Default GPU Schedule Pass (#14182)
new 193bb9d789 [Unity][Frontend] Attach imported model weights, deprecate
ImporterOutput (#14211)
new 2dd319d767 [Unity][Frontend] FX translator supports unwrapping unit
return tuple (#14212)
new 001ce2d580 [Unity][WEB] Improve webgpu codegen options to skip
readonly (#14213)
new 479529e763 [Unity][Frontend] from_fx keeps parameters in order (#14214)
new 6861939059 [Unity][BYOC] Add dynamic shape support to CUTLASS matmul
(#14216)
new 9b87830963 [Unity][Transform] Memory plan across the IRModule (#14220)
new b9c7efb6ff [Unity][WEB] Simplify WebGPU Codegen per spec (#14225)
new e6f335cdd3 [Unity] ensure memory.alloc_tensor/storage roundtrippable
(#14226)
new 3492f4b1cc [Unity] Improve implementation of FuseOps (#14229)
new 6ae4f25b3a [Unity][Transform] SimplifyNormInference (#14221)
new 29e41bc603 [Unity][Bugfix] Do not include `PrimFunc`s in the
dependency graph when checking for recursion (#14228)
new 6b73d38c45 [Unity][TVMScript] Enable Context-Aware Parsing (#14234)
new 0af0d79356 [Unity][TVMScript] Fix prim_func lost issue in
relax.emit_te (#14189)
new 252abf4ae1 [Unity][WEB] Update text prompts for syntactical
correctness (#14237)
new db42e205a5 [Unity][WEB] Improve ndarray cache (#14236)
new db153b17ef [Unity][Op][Tweak] Improve `StructInfo` inference for
`shape_of` (#14243)
new 8306c9c48b [Unity][Op] Add repeat, tile, conv2d_transpose, avg_pool2d
(#14238)
new 9bf6b9e410 [Unity] Allow user defined func attrs in emit_te (#14255)
new 5d1d033f4f [Unity][BYOC] Add conv2d and residual block patterns for
Relax cutlass BYOC (#14252)
new 6c2cfa8ef5 [Unity][Frontend] Import `tanh` and fix `layer_norm`
(#14247)
new 57732cc432 [Unity] Consider target context for Relay to Relax
conversion (#14269)
new 20247dd8c0 [Unity] Introduce call_dps_packed (#14183)
new d6a695157f [Unity][ci] Use CPU-SMALL instances (#14256)
new 4570472596 [TVMScript][Unity] Improve PyLint Compatibility (#14276)
This update added new revisions after undoing existing revisions.
That is to say, some revisions that were in the old version of the
branch are not in the new version. This situation occurs
when a user --force pushes a change and generates a repository
containing something like this:
* -- * -- B -- O -- O -- O (abc474887b)
\
N -- N -- N refs/heads/unity-staging (4570472596)
You should already have received notification emails for all of the O
revisions, and so the following emails describe only the N revisions
from the common base, B.
Any revisions marked "omit" are not gone; other references still
refer to them. Any revisions marked "discard" are gone forever.
The 124 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "add" were already present in the repository and have only
been added to this reference.
Summary of changes:
python/tvm/meta_schedule/utils.py | 3 +
python/tvm/relay/frontend/paddlepaddle.py | 102 ++++++++++++++++++++
tests/python/frontend/paddlepaddle/test_forward.py | 106 +++++++++++++++++++++
.../test_meta_schedule_post_order_apply.py | 21 ++++
4 files changed, 232 insertions(+)
