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 169448b8a9 [Unity][Frontend] FX translator returning weights with
`keep_params_as_input` (#14197)
discard a76332a8f2 [Unity][Frontend] FX translator supporting more ops (#14196)
discard b858986f4b [Unity][Op] Legalize `round`, `floor`, `ceil`, `sign`
(#14198)
discard a1b9261f3e [Unity][Op] Argmax and argmin (#14195)
discard a29ec38a07 [Unity][Op] Group normalization (#14194)
discard c293969aaa [Unity][Transform] LiftTransformParams handling multiple
functions (#14192)
discard 2cef9d4377 [Unity][WEBGPU] Codegen improvements and WebRuntime (#14187)
discard c9ae7941d0 [Unity][OP] Add an operator for fused multi head attention
(#14150)
discard 5e58b7e457 [Unity][Analysis] Restore Python bindings for var analyses
(#14180)
discard a6411eecd3 [Unity][Op] Full support of Relax op `power` (#14171)
discard 321412beb1 [Unity][BYOC] Add batch matmul support to Relax CUTLASS
BYOC (#14166)
discard 3d7b55c66b [Unity][Analysis] Analysis for detecting recursion in Relax
(#14149)
discard d2d0bafbab [Unity] Add bind_constants option to FuseOpsByPattern
(#14151)
discard b098ae0e37 [Unity][BYOC] Use Relax legalize + CPU build for reference
in tests (#14162)
discard 5304964616 [Unity][Analysis] Checking function return struct info in
well-formed check (#14155)
discard de9c84127a [Unity][Pass] Support Symbolic Shape Deduction during
BindParam (#14154)
discard c008f7d084 [Unity][Debugging] AST printer (#14152)
discard 6b86111283 [Unity][Pass] Enhance constant folding to fold relax ops by
evaluating them. (#14146)
discard c1b8f8a7fb [Unity][Legalize] Fix Scalar Constant Legalization (#14127)
discard 620da6f8e9 [Unity] Add callback to FuseOpsByPattern to check match
result is accepted (#14109)
discard 6077887837 [Unity][BYOC] Assign group to unused bindings and ignroe
PrimFunc (#14139)
discard 08a0c63837 [Unity][TVMScript] emit_te sugar (#14123)
discard ec5e478c1f [Unity][BYOC] Add transposed matmul support to Relax
CUTLASS BYOC (#14128)
discard 89c2170228 [Unity] Add Global info (#14132)
discard 834d626cc6 [Unity][WEB] Relax vm on web runtime (#14131)
discard a110a23d62 [Unity][BlockBuilder] Add `name_hint` argument for `emit`
and `emit_output` (#14126)
discard d41a2be7f2 [Unity][Fix] Fix bug in MergeCompositeFunctions (#14117)
discard 8849dcb47d [Unity] Update tests again to adapt to latest TVMScript
syntax (#14115)
discard 7a6a6a2393 [Unity][BYOC]Add relax backend pattern registry (#14106)
discard 2af3ec9204 [Unity] Remove attributes of relax.print, assert and unique
(#14101)
discard 4e8ddc81e7 [Unity][Layout] Add layout transformation analysis for
PrimFunc (#14066)
discard ec50992045 [Unity] Relax Recursive function (#14092)
discard b482a149d4 [Unity] Lower `shape_of` to a builtin (#14093)
discard d3e844bb51 [Unity] Fix typo in the comment (#14096)
discard 846edd58e2 [Unity][Relax] Set Shape Function to Be Host Function
(#14090)
discard f355a29055 [Unity] Refactor Relax Build JIT UX (#14088)
discard df290df975 [Unity][Fix][Pass] FoldConstant with DCE in dataflow block
(#14087)
discard 15642fa9bb [Unity][Analysis] TIR pattern kind analysis for
multi-buffer write block (#14075)
discard 7f08122274 [Unity][Op] `log_softmax` and `cross_entropy_with_logits`
(#14083)
discard 2bc5cd495f [Unity][BYOC] Add DNNL backend (#14082)
discard 30349b73ff [Unity][BYOC] Add CUTLASS backend (#14081)
discard 989f645f4e [Unity] Add testcases for `expr_args_converter` (#14080)
discard 4575409c7d [Unity][Pass] Canonicalize Bindings (#14079)
discard 5e7005d967 [Unity][BYOC][Pass] RunCodegen and TensorRT (#14078)
discard b27326c552 [Unity][Transform] Add LiftTransformParams pass (#14069)
discard 93e3b96dd2 [Unity][Frontend] Annotate number of non-static input of FX
function (#14067)
discard bdae3d6b66 [Unity][BYOC] Add pass to merge composite functions to
offload large subgraphs (#14062)
discard 2a8333f932 [Unity][Pass] Remove Unused Function (#14061)
discard d353855d57 [Unity][Fix][Pass] Fix FuseOps for lack graph edges (#14058)
discard f804b8832a [Unity] Relax op: collapse sum (#14059)
discard 1660159905 [Unity][BYOC] Add pattern-based partitioning pass (#14054)
discard 53b5fab032 [Unity][VM] Add per-op profiling support (#14053)
discard 45afc08856 [Unity][TVMScript] Overload `__neg__` for relax expr
(#14045)
discard 341ef4239b [Unity][Pass] FuseOps FuseTIR fixes (#14044)
discard c0158714dc [Unity] Statement rewriter for DataflowBlock (#14043)
discard 167f09a8cf [Unity] Relax dataflow pattern language (matching) (#14041)
discard 582efa6b02 [Unity] Update tests to adapt to latest TVMScript syntax
(#14039)
discard fddc43d835 [Unity] Disallow inline prim_func in relax IR (#14040)
discard 111f71f928 [Unity][Pass] Block-level static memory planning (#14038)
discard 1d4118dec0 [Unity] Initial PyTorch Frontend (#14037)
discard 29fb528fb0 [Unity][Op] Add ShapeExpr Tests for Reshape Op (#14035)
discard 5e5e994ffc [Unity][Pass] Operator legalization (#14029)
discard 2a64d3ea69 [Unity][TVMScript] Move tir/relax import in script out of
__init__.py (#14033)
discard 28c12d2a6f [Unity][Pass] Wellformed Analysis (#14032)
discard 9a876efecc [Unity][BlockBuilder] CallTE convert PrimValue args
(#14028)
discard 9ddbc488f1 [Unity][Pass] Normalize Pass (#14031)
discard be61b78370 [Unity] Relay -> Relax translator (#14026)
discard 3200f9237a [Unity][Pass][TuningAPI] Introduce TuningAPI and
MetaSchedule pass (#14014)
discard a1e237358a [Unity][Pass] BindParams pass, FoldConstant pass (#14016)
discard 6419af94cc [Unity][VM] Supporting "compiled" exec mode. (#14015)
discard c1be350206 [Unity][Pass] LambdaLift pass (#14012)
discard bf7e8eaf3a [Unity][Pass] Operator Fusion Passes (#14001)
discard 900776db89 [Unity] NestedMsg Support utility (#13995)
discard 47b43f1a92 [Unity] Relax op: manipulation (#13989)
discard f9d1dccf7a [Unity] Relax op: search (#13992)
discard 202a6e6cf5 [Unity] Relax op: linear algebra (#13988)
discard c11338702d [Unity] Relax op: creation (#13984)
discard f8867390a6 [Unity] Relax op: neural networks (#13993)
discard b52113a0a9 [Unity] Relax op: statistical (#13991)
discard 379c8d28a7 [Unity] Relax op: arithmetic, comparison (#13983)
discard 4d0485cd20 [Unity] Relax op: image (#13994)
discard a0ecbe955a [Unity] Relax op: set (#13990)
discard 0ed6b9aebf [Unity] Relax op: datatype (#13986)
discard a3e275df76 [Unity] Relax op: index (#13987)
discard b9807a62d4 [Unity][TVMScript] Use explicit `R.shape` in TVMScript
(#13979)
discard c50df572c5 [Unity] e2e Relax minimum build flow (#13961)
discard 4ba6f1d175 [Unity] Relax VM shape lowering pass (#13956)
discard 5a13e0d4ae [Unity] Relax VM codegen (#13954)
discard 6425cd1b41 [Unity] Relax TVMScript Printer (#13944)
discard 23b31ddcab [Unity] Relax TVMScript Parser. (#13932)
discard 12f38f5d46 [Unity] Relax BlockBuilder and ExprMutator (#13926)
discard 39e1571752 [Unity] Basic StructInfo Analysis and Expr construction
(#13916)
discard 2b2a8bfede [Unity][CI] Unity specific jenkins setup (do not upstream
to main) (#13910)
discard 1e5110d557 [Unity][IR] First-class StructInfo (#13907)
discard 543503d2a7 [Unity] Relax expressions and types (#13901)
discard f79850db8b [Unity] Relax VM (#13878)
add befdc4e631 [Fix][TIR] LowerCrossThreadReduction with write-back
predicate (#14199)
new e7b02f29c1 [Unity] Relax VM (#13878)
new 9508a18d9e [Unity] Relax expressions and types (#13901)
new 4d46290e7b [Unity][IR] First-class StructInfo (#13907)
new b59ad4839c [Unity][CI] Unity specific jenkins setup (do not upstream
to main) (#13910)
new ff8bfa2732 [Unity] Basic StructInfo Analysis and Expr construction
(#13916)
new bb0c129b7c [Unity] Relax BlockBuilder and ExprMutator (#13926)
new 1807e6f015 [Unity] Relax TVMScript Parser. (#13932)
new 846a2c59aa [Unity] Relax TVMScript Printer (#13944)
new 850d6a4c66 [Unity] Relax VM codegen (#13954)
new a966cf115b [Unity] Relax VM shape lowering pass (#13956)
new f735d932f6 [Unity] e2e Relax minimum build flow (#13961)
new ad4185cf07 [Unity][TVMScript] Use explicit `R.shape` in TVMScript
(#13979)
new e8227b91f8 [Unity] Relax op: index (#13987)
new 2e08c8c8bc [Unity] Relax op: datatype (#13986)
new 68a04a8144 [Unity] Relax op: set (#13990)
new 5723ebb2fd [Unity] Relax op: image (#13994)
new 72bca0f3ea [Unity] Relax op: arithmetic, comparison (#13983)
new f491b96273 [Unity] Relax op: statistical (#13991)
new 71437f7488 [Unity] Relax op: neural networks (#13993)
new 9113fc992d [Unity] Relax op: creation (#13984)
new c788135337 [Unity] Relax op: linear algebra (#13988)
new b6818bba6b [Unity] Relax op: search (#13992)
new 17cf44657d [Unity] Relax op: manipulation (#13989)
new 9317ec8db7 [Unity] NestedMsg Support utility (#13995)
new 07e0dfb06e [Unity][Pass] Operator Fusion Passes (#14001)
new 33c4aabff9 [Unity][Pass] LambdaLift pass (#14012)
new fd5c73d62c [Unity][VM] Supporting "compiled" exec mode. (#14015)
new 733fc00145 [Unity][Pass] BindParams pass, FoldConstant pass (#14016)
new 88852c1e16 [Unity][Pass][TuningAPI] Introduce TuningAPI and
MetaSchedule pass (#14014)
new 24470c95d7 [Unity] Relay -> Relax translator (#14026)
new fc0540cf11 [Unity][Pass] Normalize Pass (#14031)
new 91adf7b0a4 [Unity][BlockBuilder] CallTE convert PrimValue args
(#14028)
new 53f800d32f [Unity][Pass] Wellformed Analysis (#14032)
new 596d472a4f [Unity][TVMScript] Move tir/relax import in script out of
__init__.py (#14033)
new b150b1a6ac [Unity][Pass] Operator legalization (#14029)
new 449e094e80 [Unity][Op] Add ShapeExpr Tests for Reshape Op (#14035)
new 9b5f2148fb [Unity] Initial PyTorch Frontend (#14037)
new fd35d1ef5e [Unity][Pass] Block-level static memory planning (#14038)
new d8fdd5cdb6 [Unity] Disallow inline prim_func in relax IR (#14040)
new 8039f6a75e [Unity] Update tests to adapt to latest TVMScript syntax
(#14039)
new 782c632604 [Unity] Relax dataflow pattern language (matching) (#14041)
new 1b85765d5c [Unity] Statement rewriter for DataflowBlock (#14043)
new c1439b3adb [Unity][Pass] FuseOps FuseTIR fixes (#14044)
new 180bead06b [Unity][TVMScript] Overload `__neg__` for relax expr
(#14045)
new fe528f647c [Unity][VM] Add per-op profiling support (#14053)
new aa55c05560 [Unity][BYOC] Add pattern-based partitioning pass (#14054)
new fada709394 [Unity] Relax op: collapse sum (#14059)
new 06de35ec74 [Unity][Fix][Pass] Fix FuseOps for lack graph edges (#14058)
new a9032d98e6 [Unity][Pass] Remove Unused Function (#14061)
new ed2696a595 [Unity][BYOC] Add pass to merge composite functions to
offload large subgraphs (#14062)
new 0d5883539d [Unity][Frontend] Annotate number of non-static input of FX
function (#14067)
new 246c4c13b8 [Unity][Transform] Add LiftTransformParams pass (#14069)
new e8a0c4d01f [Unity][BYOC][Pass] RunCodegen and TensorRT (#14078)
new 4ad8d64155 [Unity][Pass] Canonicalize Bindings (#14079)
new 3a64963901 [Unity] Add testcases for `expr_args_converter` (#14080)
new b8460ebabc [Unity][BYOC] Add CUTLASS backend (#14081)
new defc15b6f2 [Unity][BYOC] Add DNNL backend (#14082)
new 7645aa7688 [Unity][Op] `log_softmax` and `cross_entropy_with_logits`
(#14083)
new 6e2d7bbfd4 [Unity][Analysis] TIR pattern kind analysis for
multi-buffer write block (#14075)
new acd0e0b999 [Unity][Fix][Pass] FoldConstant with DCE in dataflow block
(#14087)
new 19509408b4 [Unity] Refactor Relax Build JIT UX (#14088)
new cf36b7bdac [Unity][Relax] Set Shape Function to Be Host Function
(#14090)
new b1f2d530fc [Unity] Fix typo in the comment (#14096)
new a8338e6b69 [Unity] Lower `shape_of` to a builtin (#14093)
new 74f3007c81 [Unity] Relax Recursive function (#14092)
new 98d0a01bc9 [Unity][Layout] Add layout transformation analysis for
PrimFunc (#14066)
new b755a6f004 [Unity] Remove attributes of relax.print, assert and unique
(#14101)
new eaaa1fbdcb [Unity][BYOC]Add relax backend pattern registry (#14106)
new 89bb68b0a0 [Unity] Update tests again to adapt to latest TVMScript
syntax (#14115)
new 02b3a1f2ff [Unity][Fix] Fix bug in MergeCompositeFunctions (#14117)
new c7d2c38dd9 [Unity][BlockBuilder] Add `name_hint` argument for `emit`
and `emit_output` (#14126)
new 57c86eb774 [Unity][WEB] Relax vm on web runtime (#14131)
new 61c27610c9 [Unity] Add Global info (#14132)
new 368d9f6193 [Unity][BYOC] Add transposed matmul support to Relax
CUTLASS BYOC (#14128)
new 4713b527cb [Unity][TVMScript] emit_te sugar (#14123)
new 28c6825655 [Unity][BYOC] Assign group to unused bindings and ignroe
PrimFunc (#14139)
new 63ce37d30b [Unity] Add callback to FuseOpsByPattern to check match
result is accepted (#14109)
new 7a5d313068 [Unity][Legalize] Fix Scalar Constant Legalization (#14127)
new c29ac7eef0 [Unity][Pass] Enhance constant folding to fold relax ops by
evaluating them. (#14146)
new 4c085d2212 [Unity][Debugging] AST printer (#14152)
new 8a1e623fc1 [Unity][Pass] Support Symbolic Shape Deduction during
BindParam (#14154)
new 67659acf34 [Unity][Analysis] Checking function return struct info in
well-formed check (#14155)
new e57f59178f [Unity][BYOC] Use Relax legalize + CPU build for reference
in tests (#14162)
new 3fa880aeb5 [Unity] Add bind_constants option to FuseOpsByPattern
(#14151)
new 6ee79e1e08 [Unity][Analysis] Analysis for detecting recursion in Relax
(#14149)
new 8423811429 [Unity][BYOC] Add batch matmul support to Relax CUTLASS
BYOC (#14166)
new 781bfe0a21 [Unity][Op] Full support of Relax op `power` (#14171)
new 475f3c24fa [Unity][Analysis] Restore Python bindings for var analyses
(#14180)
new 38315afbf2 [Unity][OP] Add an operator for fused multi head attention
(#14150)
new ed5367d161 [Unity][WEBGPU] Codegen improvements and WebRuntime (#14187)
new 1f042218d3 [Unity][Transform] LiftTransformParams handling multiple
functions (#14192)
new 22b65bc6eb [Unity][Op] Group normalization (#14194)
new 88ab7302ba [Unity][Op] Argmax and argmin (#14195)
new 0e98e6e478 [Unity][Op] Legalize `round`, `floor`, `ceil`, `sign`
(#14198)
new 284b278706 [Unity][Frontend] FX translator supporting more ops (#14196)
new f7f24b7d16 [Unity][Frontend] FX translator returning weights with
`keep_params_as_input` (#14197)
new d103ee2070 [Unity][Fix] FX translating dtype (#14201)
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 (169448b8a9)
\
N -- N -- N refs/heads/unity-staging (d103ee2070)
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 97 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/relax/frontend/torch/fx_translator.py | 15 +++++++++------
src/tir/transforms/lower_cross_thread_reduction.cc | 8 +++++++-
tests/python/relax/test_frontend_from_fx.py | 14 ++++++++++----
.../test_tir_transform_lower_cross_thread_reduction.py | 14 ++++++++++++++
4 files changed, 40 insertions(+), 11 deletions(-)
