[tvm] branch main updated (b4d595c -> 1f5f3c9)

This is an automated email from the ASF dual-hosted git repository. masahi pushed a change to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git. from b4d595c Add CMake summary (#9696) add 1f5f3c9 [CUDA] Do not emit vector load on unaligned base offset (#9731) No

[GitHub] [tvm] masahi merged pull request #9731: [CUDA] Do not emit vector load on unaligned base offset

masahi merged pull request #9731: URL: https://github.com/apache/tvm/pull/9731 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[GitHub] [tvm] masahi closed issue #8540: [BUG] CUDA misaligned address when vectorizing float16

masahi closed issue #8540: URL: https://github.com/apache/tvm/issues/8540 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[GitHub] [tvm] daniperfer commented on issue #7990: LLVM ERROR: out of memory, when tracing maskrcnn model in deploy_object_detection_pytorch.py tutorial

daniperfer commented on issue #7990: URL: https://github.com/apache/tvm/issues/7990#issuecomment-993219644 I posted a question about this in the TVM discussion forums, in which I described how to change LLVM version. Unfortunately, it doesn't work for me :(, and I couldn't find a

[GitHub] [tvm] junrushao1994 commented on pull request #9729: [CORE] Make external mods first class field in the IR module

junrushao1994 commented on pull request #9729: URL: https://github.com/apache/tvm/pull/9729#issuecomment-993208836 Thank you @mbs-octoml and @comaniac for providing important context on this PR! Appreciated you guys doing so when discussing architectural changes! A fundamental data

[GitHub] [tvm] masahi closed issue #9653: [Bug] test_tensorrt.py::test_conv2d_transpose failed

masahi closed issue #9653: URL: https://github.com/apache/tvm/issues/9653 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[tvm] branch main updated (69cae0b -> b4d595c)

This is an automated email from the ASF dual-hosted git repository. masahi pushed a change to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git. from 69cae0b [Relay, BYOC] Make constant binding in PartitionGraph optional (#9721) add b4d595c Add CMake summary

[GitHub] [tvm] masahi merged pull request #9696: Add CMake summary

masahi merged pull request #9696: URL: https://github.com/apache/tvm/pull/9696 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[GitHub] [tvm] masahi opened a new pull request #9737: [CUTLASS] Add conv2d profiler

masahi opened a new pull request #9737: URL: https://github.com/apache/tvm/pull/9737 Thanks for contributing to TVM! Please refer to guideline https://tvm.apache.org/docs/contribute/ for useful information and tips. After the pull request is submitted, please request code reviews from

[GitHub] [tvm] zhiics commented on pull request #9729: [CORE] Make external mods first class field in the IR module

zhiics commented on pull request #9729: URL: https://github.com/apache/tvm/pull/9729#issuecomment-993172582 @mikepapadim Thanks for the PR. Could you elaborate a bit the motivation of moving the external runtime module into the IRModule? IMOH, the IRModule, particularly the relay IR here,

[tvm] branch main updated (d1dafbd -> 69cae0b)

This is an automated email from the ASF dual-hosted git repository. masahi pushed a change to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git. from d1dafbd [CUTLASS] More robust support for pattern matching and alignment (#9698) add 69cae0b [Relay, BYOC] Make

[GitHub] [tvm] masahi merged pull request #9721: [Relay, BYOC] Make constant binding in PartitionGraph optional

masahi merged pull request #9721: URL: https://github.com/apache/tvm/pull/9721 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[tvm] branch main updated (4e70931 -> d1dafbd)

This is an automated email from the ASF dual-hosted git repository. masahi pushed a change to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git. from 4e70931 [BugFix] Fix a wrong use of `std::move()` in cross-thread reduction lowering (#9728) add d1dafbd

[GitHub] [tvm] masahi merged pull request #9698: [CUTLASS] More robust support for pattern matching and alignment

masahi merged pull request #9698: URL: https://github.com/apache/tvm/pull/9698 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[GitHub] [tvm] CircleSpin commented on a change in pull request #9633: Tvmc python tutorial

CircleSpin commented on a change in pull request #9633: URL: https://github.com/apache/tvm/pull/9633#discussion_r768268601 ## File path: gallery/how_to/use_tvms_python_api/tvmc_python.py ## @@ -0,0 +1,252 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or

[GitHub] [tvm] CircleSpin commented on pull request #9633: Tvmc python tutorial

CircleSpin commented on pull request #9633: URL: https://github.com/apache/tvm/pull/9633#issuecomment-993095492 Hi @hogepodge, By all means if there's another ordering of the tutorials you prefer let's go with that. This tutorial seems to do the same thing as this one

[GitHub] [tvm] ganler commented on a change in pull request #9582: [BugFix] resolve integer 32. ~ 64. mismatch by casting

ganler commented on a change in pull request #9582: URL: https://github.com/apache/tvm/pull/9582#discussion_r757714983 ## File path: tests/python/relay/test_type_solver.py ## @@ -338,6 +341,34 @@ def test_incompatible_quantified_func_unification(): solver.Unify(ft1, ft2)

[GitHub] [tvm] junrushao1994 merged pull request #9728: [BugFix] Fix a wrong use of `std::move()` in cross-thread reduction lowering

junrushao1994 merged pull request #9728: URL: https://github.com/apache/tvm/pull/9728 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[tvm] branch main updated: [BugFix] Fix a wrong use of `std::move()` in cross-thread reduction lowering (#9728)

This is an automated email from the ASF dual-hosted git repository. junrushao pushed a commit to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git The following commit(s) were added to refs/heads/main by this push: new 4e70931 [BugFix] Fix a wrong use of `std::move()`

[GitHub] [tvm] adstraw opened a new pull request #9736: Add unit tests for HexagonBuffer

adstraw opened a new pull request #9736: URL: https://github.com/apache/tvm/pull/9736 Unit tests for HexagonBuffer class which run on CPU (not Hexagon) and test out full capabilities of HexagonBuffer e.g. discontiguous allocation that are not currently exposed through the Hexagon Device

[GitHub] [tvm] ccjoechou commented on pull request #9730: [pre-RFC][BYOC] Marvell ML/AI Accelerator Integration

ccjoechou commented on pull request #9730: URL: https://github.com/apache/tvm/pull/9730#issuecomment-993059437 Hello @tqchen, @jroesch, or any other reviewer: We have also posted a pre-RFC

[GitHub] [tvm] AndrewZhaoLuo opened a new pull request #9735: [WIP] Add faster type inference

AndrewZhaoLuo opened a new pull request #9735: URL: https://github.com/apache/tvm/pull/9735 TODO -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[GitHub] [tvm] jroesch commented on pull request #9730: [pre-RFC][BYOC] Marvell ML/AI Accelerator Integration

jroesch commented on pull request #9730: URL: https://github.com/apache/tvm/pull/9730#issuecomment-993048649 Thanks for your contribution. When you have a chance could you update the description of the PR with more details, and potentially link to the pre-RFC discussion? Thanks! --

[GitHub] [tvm] jroesch commented on pull request #9696: Add CMake summary

jroesch commented on pull request #9696: URL: https://github.com/apache/tvm/pull/9696#issuecomment-993045923 @driazati Great I feel like this will come in handy when debugging CI problems. -- This is an automated message from the Apache Git Service. To respond to the message, please

[tvm] branch main updated (f0ae9d3 -> 1728782)

This is an automated email from the ASF dual-hosted git repository. jroesch pushed a change to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git. from f0ae9d3 Add `make docs` and doc building instructions (#9534) add 1728782 Fix typos in runtime comments (#9726)

[GitHub] [tvm] jroesch merged pull request #9726: Fix typos in runtime comments

jroesch merged pull request #9726: URL: https://github.com/apache/tvm/pull/9726 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[GitHub] [tvm] areusch merged pull request #9534: Add `make docs` and doc building instructions

areusch merged pull request #9534: URL: https://github.com/apache/tvm/pull/9534 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[tvm] branch main updated (a674121 -> f0ae9d3)

This is an automated email from the ASF dual-hosted git repository. areusch pushed a change to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git. from a674121 [Relay] Non-recursive dependency graph (#9528) add f0ae9d3 Add `make docs` and doc building instructions

[GitHub] [tvm] mbs-octoml edited a comment on pull request #9729: [CORE] Make external mods first class field in the IR module

mbs-octoml edited a comment on pull request #9729: URL: https://github.com/apache/tvm/pull/9729#issuecomment-992994783 Thanks @junrushao1994. I'd be up for anything that replaces runtime::Module in IRModule with something more structured that nevertheless can still carry artifacts at the

[GitHub] [tvm] mbs-octoml commented on pull request #9729: [CORE] Make external mods first class field in the IR module

mbs-octoml commented on pull request #9729: URL: https://github.com/apache/tvm/pull/9729#issuecomment-992994783 Thanks @junrushao1994. I'd be up for anything that replaces runtime::Module in IRModule with something more structured that nevertheless can still carry artifacts at the various

[GitHub] [tvm] mbs-octoml opened a new pull request #9734: [Relay] Support large constants saved/loaded outside of VM executable

mbs-octoml opened a new pull request #9734: URL: https://github.com/apache/tvm/pull/9734 This allows constant tensors at or above a given byte limit to be marked as 'late bound' and saved/reloaded to a bytestream independent of the overall executable. Since the executable is often

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9613: [Relay] Re-run PlanDevices after LowerTE to flow new memory scope constraints.

mbs-octoml commented on a change in pull request #9613: URL: https://github.com/apache/tvm/pull/9613#discussion_r768182812 ## File path: include/tvm/relay/expr.h ## @@ -234,6 +234,15 @@ class Var : public Expr { */ TVM_DLL Var(Id vid, Type type_annotation, Span span =

[GitHub] [tvm] junrushao1994 edited a comment on pull request #9729: [CORE] Make external mods first class field in the IR module

junrushao1994 edited a comment on pull request #9729: URL: https://github.com/apache/tvm/pull/9729#issuecomment-992910847 I'm writing some discussion on the technical side. (bear with me i'm a slow typer) BTW, It would be great if any IR changes could go through the RFC process. --

[GitHub] [tvm] driazati opened a new pull request #9733: Combine unit and integration test steps into one stage

driazati opened a new pull request #9733: URL: https://github.com/apache/tvm/pull/9733 This removes the barrier wait between test and integration tests in CI. This will increase capacity requirements and usage but, assuming we can meet that with autoscaler, should reduce CI times by an

[GitHub] [tvm] driazati closed pull request #9709: Combine unit and integration test steps into one stage

driazati closed pull request #9709: URL: https://github.com/apache/tvm/pull/9709 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[GitHub] [tvm] driazati commented on a change in pull request #9710: Clarify error message for missing libraries

driazati commented on a change in pull request #9710: URL: https://github.com/apache/tvm/pull/9710#discussion_r768159328 ## File path: python/tvm/_ffi/libinfo.py ## @@ -114,18 +114,18 @@ def find_lib_path(name=None, search_path=None, optional=False): runtime_dll_path

[GitHub] [tvm] driazati opened a new pull request #9732: Merge Java unittests into GPU unittests

driazati opened a new pull request #9732: URL: https://github.com/apache/tvm/pull/9732 These run on the same node but the Java unit tests take < 1 min (whereas the node setup alone takes 3-4 min), so it seems like it would be good to just run the Java tests as part of the general GPU unit

[GitHub] [tvm] masahi opened a new pull request #9731: [CUDA] Do not emit vector load on unaligned base offset

masahi opened a new pull request #9731: URL: https://github.com/apache/tvm/pull/9731 Closes https://github.com/apache/tvm/issues/8540 cc @vinx13 @comaniac @Laurawly -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and

[GitHub] [tvm] junrushao1994 commented on pull request #9729: [CORE] Make external mods first class field in the IR module

junrushao1994 commented on pull request #9729: URL: https://github.com/apache/tvm/pull/9729#issuecomment-992910847 I'm writing some discussion on the technical side. (bear with me i'm a slow typer) BTW, It would be great if significant IR changes could go through the RFC process. --

[GitHub] [tvm] kparzysz-quic commented on a change in pull request #9631: [Hexagon] Add RPC Mechanism for Hexagon

kparzysz-quic commented on a change in pull request #9631: URL: https://github.com/apache/tvm/pull/9631#discussion_r768120523 ## File path: cmake/modules/Hexagon.cmake ## @@ -241,10 +340,39 @@ if (USE_HEXAGON_DEVICE STREQUAL "${PICK_NONE}") elseif(BUILD_FOR_ANDROID AND

[GitHub] [tvm] masahi commented on a change in pull request #9721: [Relay, BYOC] Make constant binding in PartitionGraph optional

masahi commented on a change in pull request #9721: URL: https://github.com/apache/tvm/pull/9721#discussion_r768112083 ## File path: src/relay/transforms/partition_graph.cc ## @@ -601,9 +607,10 @@ Pass PartitionGraph(String mod_name) { {flatten_tuples_pass,

[GitHub] [tvm] masahi commented on a change in pull request #9721: [Relay, BYOC] Make constant binding in PartitionGraph optional

masahi commented on a change in pull request #9721: URL: https://github.com/apache/tvm/pull/9721#discussion_r768112083 ## File path: src/relay/transforms/partition_graph.cc ## @@ -601,9 +607,10 @@ Pass PartitionGraph(String mod_name) { {flatten_tuples_pass,

[GitHub] [tvm] masahi commented on a change in pull request #9721: [Relay, BYOC] Make constant binding in PartitionGraph optional

masahi commented on a change in pull request #9721: URL: https://github.com/apache/tvm/pull/9721#discussion_r768109935 ## File path: python/tvm/relay/transform/transform.py ## @@ -695,17 +695,23 @@ def LambdaLift(): return _ffi_api.LambdaLift() -def

[GitHub] [tvm] ccjoechou opened a new pull request #9730: [pre-RFC][BYOC] Marvell ML/AI Accelerator Integration

ccjoechou opened a new pull request #9730: URL: https://github.com/apache/tvm/pull/9730 Thanks for contributing to TVM! Please refer to guideline https://tvm.apache.org/docs/contribute/ for useful information and tips. After the pull request is submitted, please request code reviews

[tvm] branch ci-docker-staging updated (6305c7d -> 0710915)

This is an automated email from the ASF dual-hosted git repository. areusch pushed a change to branch ci-docker-staging in repository https://gitbox.apache.org/repos/asf/tvm.git. discard 6305c7d [CI] Update ci_cpu docker image omit e8889ae [TVMScript] Add syntax sugar for T.handle and

[GitHub] [tvm] hzfan closed pull request #9720: [Relay] Non-recursive dependency graph

hzfan closed pull request #9720: URL: https://github.com/apache/tvm/pull/9720 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[tvm] branch main updated (6e9e4e6 -> a674121)

This is an automated email from the ASF dual-hosted git repository. comaniac pushed a change to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git. from 6e9e4e6 [TVMC] Add --opt-level to compile mode (#9722) add a674121 [Relay] Non-recursive dependency graph

[GitHub] [tvm] comaniac merged pull request #9528: [Relay] Non-recursive dependency graph

comaniac merged pull request #9528: URL: https://github.com/apache/tvm/pull/9528 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[GitHub] [tvm] comaniac commented on pull request #9528: [Relay] Non-recursive dependency graph

comaniac commented on pull request #9528: URL: https://github.com/apache/tvm/pull/9528#issuecomment-992817707 Thanks @hzfan @mbrookhart -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the

[GitHub] [tvm] driazati commented on pull request #9696: Add CMake summary

driazati commented on pull request #9696: URL: https://github.com/apache/tvm/pull/9696#issuecomment-992816914 It is now with https://github.com/apache/tvm/pull/9696/commits/e680e69d77ea9ed05d46fce701518578e21b45b0 ([test

[GitHub] [tvm] jroesch commented on a change in pull request #9613: [Relay] Re-run PlanDevices after LowerTE to flow new memory scope constraints.

jroesch commented on a change in pull request #9613: URL: https://github.com/apache/tvm/pull/9613#discussion_r768066039 ## File path: include/tvm/relay/expr.h ## @@ -234,6 +234,15 @@ class Var : public Expr { */ TVM_DLL Var(Id vid, Type type_annotation, Span span =

[GitHub] [tvm] Lunderberg commented on pull request #9727: [TE][TIR] Implement layout transformations, non-flat memory buffers

Lunderberg commented on pull request #9727: URL: https://github.com/apache/tvm/pull/9727#issuecomment-992799674 Fixing lint issues. The large number of commits listed is due to rebasing the fixes into the commits that introduced them. -- This is an automated message from the Apache Git

[GitHub] [tvm] driazati commented on pull request #9554: Implement [skip ci] for Jenkins

driazati commented on pull request #9554: URL: https://github.com/apache/tvm/pull/9554#issuecomment-992791257 Confirming sounds good, it should be ready to push now -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the

[GitHub] [tvm] driazati commented on a change in pull request #9554: Implement [skip ci] for Jenkins

driazati commented on a change in pull request #9554: URL: https://github.com/apache/tvm/pull/9554#discussion_r768049498 ## File path: tests/python/unittest/test_ci.py ## @@ -0,0 +1,116 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor

[GitHub] [tvm] areusch commented on a change in pull request #9554: Implement [skip ci] for Jenkins

areusch commented on a change in pull request #9554: URL: https://github.com/apache/tvm/pull/9554#discussion_r768041428 ## File path: tests/python/unittest/test_ci.py ## @@ -0,0 +1,116 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor

[GitHub] [tvm] junrushao1994 commented on a change in pull request #9729: [CORE] Make external mods first class field in the IR module

junrushao1994 commented on a change in pull request #9729: URL: https://github.com/apache/tvm/pull/9729#discussion_r768036732 ## File path: include/tvm/ir/module.h ## @@ -57,6 +57,8 @@ class IRModuleNode : public Object { Map functions; /*! \brief A map from global type

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9729: [CORE] Make external mods first class field in the IR module

mbs-octoml commented on a change in pull request #9729: URL: https://github.com/apache/tvm/pull/9729#discussion_r768029280 ## File path: include/tvm/ir/module.h ## @@ -57,6 +57,8 @@ class IRModuleNode : public Object { Map functions; /*! \brief A map from global type

[GitHub] [tvm-rfcs] areusch commented on pull request #46: Module Based Model Runtime for AOT

areusch commented on pull request #46: URL: https://github.com/apache/tvm-rfcs/pull/46#issuecomment-992762313 @manupa-arm thanks! I think ultimately the @ashutosh-arm pass could be a good way to implement ConstLoader, but for now I'm hoping we can simply rename. In the long run, I'd

[GitHub] [tvm] electriclilies commented on a change in pull request #9729: [CORE] Make external mods first class field in the IR module

electriclilies commented on a change in pull request #9729: URL: https://github.com/apache/tvm/pull/9729#discussion_r768016573 ## File path: include/tvm/ir/module.h ## @@ -57,6 +57,8 @@ class IRModuleNode : public Object { Map functions; /*! \brief A map from global type

[GitHub] [tvm] mbs-octoml commented on pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on pull request #9693: URL: https://github.com/apache/tvm/pull/9693#issuecomment-992747464 Thanks @electriclilies and @jroesch , merged your post-submit comments into the sequel. -- This is an automated message from the Apache Git Service. To respond to the

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r768010547 ## File path: src/relay/transforms/device_planner.cc ## @@ -844,7 +932,8 @@ class DeviceCapturer : public ExprMutator { // match. return

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r768009277 ## File path: src/relay/op/memory/on_device.h ## @@ -39,14 +39,50 @@ namespace relay { const Op& OnDeviceOp(); /*! - * \brief Wraps \p expr in an

[GitHub] [tvm] comaniac commented on a change in pull request #9729: [CORE] Make external mods first class field in the IR module

comaniac commented on a change in pull request #9729: URL: https://github.com/apache/tvm/pull/9729#discussion_r768008092 ## File path: include/tvm/ir/module.h ## @@ -57,6 +57,8 @@ class IRModuleNode : public Object { Map functions; /*! \brief A map from global type vars

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r768007923 ## File path: src/relay/op/memory/on_device.cc ## @@ -45,53 +43,74 @@ const Op& OnDeviceOp() { return op; } -Expr OnDevice(Expr expr, SEScope

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r768005529 ## File path: src/relay/op/memory/on_device.cc ## @@ -45,53 +43,74 @@ const Op& OnDeviceOp() { return op; } -Expr OnDevice(Expr expr, SEScope

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r768004662 ## File path: src/relay/op/memory/on_device.cc ## @@ -45,53 +43,74 @@ const Op& OnDeviceOp() { return op; } -Expr OnDevice(Expr expr, SEScope

[GitHub] [tvm] comaniac commented on a change in pull request #9721: [Relay, BYOC] Make constant binding in PartitionGraph optional

comaniac commented on a change in pull request #9721: URL: https://github.com/apache/tvm/pull/9721#discussion_r767983007 ## File path: python/tvm/relay/transform/transform.py ## @@ -695,17 +695,23 @@ def LambdaLift(): return _ffi_api.LambdaLift() -def

[GitHub] [tvm] junrushao1994 commented on a change in pull request #9729: [CORE] Make external mods first class field in the IR module

junrushao1994 commented on a change in pull request #9729: URL: https://github.com/apache/tvm/pull/9729#discussion_r767980273 ## File path: include/tvm/ir/module.h ## @@ -57,6 +57,8 @@ class IRModuleNode : public Object { Map functions; /*! \brief A map from global type

[GitHub] [tvm] junrushao1994 commented on a change in pull request #9729: [CORE] Make external mods first class field in the IR module

junrushao1994 commented on a change in pull request #9729: URL: https://github.com/apache/tvm/pull/9729#discussion_r767980273 ## File path: include/tvm/ir/module.h ## @@ -57,6 +57,8 @@ class IRModuleNode : public Object { Map functions; /*! \brief A map from global type

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r767974458 ## File path: src/relay/op/call/call.cc ## @@ -63,23 +63,23 @@ bool CallLoweredRel(const Array& types, int num_inputs, const Attrs& attrs const Op&

[GitHub] [tvm] mikepapadim opened a new pull request #9729: [CORE] Make external mods first class field in the IR module

mikepapadim opened a new pull request #9729: URL: https://github.com/apache/tvm/pull/9729 Co-authored-by: Lily Orth-Smith This PR extracts the external modules used by the TECOmpiler and various places in the AOT flow and makes them first class fields in the IRModule.

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r767972447 ## File path: src/printer/text_printer.cc ## @@ -36,49 +37,71 @@ static const char* kSemVer = "0.0.5"; Doc TextPrinter::PrintMod(const IRModule& mod) {

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r767970440 ## File path: src/printer/text_printer.cc ## @@ -36,49 +37,71 @@ static const char* kSemVer = "0.0.5"; Doc TextPrinter::PrintMod(const IRModule& mod) {

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r767969790 ## File path: include/tvm/relay/attrs/on_device.h ## @@ -54,44 +54,48 @@ namespace relay { * multiply(device_copy(add(%x, %y), src_se_scope=GPU,

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r767969023 ## File path: src/relay/transforms/device_planner.cc ## @@ -20,6 +20,10 @@ /*! * \file src/relay/transforms/device_planner.cc * \brief Determines a

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9693: [Relay] PlanDevices supports 'free' on_device annotations

mbs-octoml commented on a change in pull request #9693: URL: https://github.com/apache/tvm/pull/9693#discussion_r767968431 ## File path: include/tvm/target/se_scope.h ## @@ -170,7 +170,7 @@ class SEScopeNode : public AttrsNode { * * kInvalidDeviceType denotes

[GitHub] [tvm] abdulazizm commented on issue #7990: LLVM ERROR: out of memory, when tracing maskrcnn model in deploy_object_detection_pytorch.py tutorial

abdulazizm commented on issue #7990: URL: https://github.com/apache/tvm/issues/7990#issuecomment-992694906 @daniperfer sorry i am newbie, how can we change this llvm version? Is there any blog or documentation that you refer? -- This is an automated message from the Apache Git Service.

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9669: [BugFix] shapeOfAttrs should be registered before "vm.shape_of" used

mbs-octoml commented on a change in pull request #9669: URL: https://github.com/apache/tvm/pull/9669#discussion_r767960497 ## File path: tests/python/relay/test_vm.py ## @@ -1112,6 +1112,20 @@ def test_multi_targets(): tvm.testing.assert_allclose(actual_result.numpy(),

[GitHub] [tvm] manupa-arm commented on pull request #8472: [TIR] Introduce tir.allocate_const to TIR

manupa-arm commented on pull request #8472: URL: https://github.com/apache/tvm/pull/8472#issuecomment-992685727 @d-smirnov I suppose this is the first PR that needs to go in ? (I m waiting to review this though I've seen updates to the PR that depends on this). Let me know what

[tvm] branch main updated: [TVMC] Add --opt-level to compile mode (#9722)

This is an automated email from the ASF dual-hosted git repository. leandron pushed a commit to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git The following commit(s) were added to refs/heads/main by this push: new 6e9e4e6 [TVMC] Add --opt-level to compile mode

[GitHub] [tvm] leandron merged pull request #9722: [TVMC] Add --opt-level to compile mode

leandron merged pull request #9722: URL: https://github.com/apache/tvm/pull/9722 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[GitHub] [tvm] leandron commented on a change in pull request #9725: [microNPU] Update Arm(R) Ethos(TM)-U55 NPU demo README

leandron commented on a change in pull request #9725: URL: https://github.com/apache/tvm/pull/9725#discussion_r767949380 ## File path: apps/microtvm/ethosu/README.md ## @@ -43,6 +43,10 @@ If the demo is not run in the ci_cpu Docker container, then you will need the fo You

[GitHub] [tvm] MasterJH5574 commented on pull request #9728: [BugFix] Fix a wrong use of `std::move()` in cross-thread reduction lowering

MasterJH5574 commented on pull request #9728: URL: https://github.com/apache/tvm/pull/9728#issuecomment-992661562 :joy: Given the [undefined behavior of evaluation order](https://stackoverflow.com/questions/2934904/order-of-evaluation-in-c-function-parameters), we decided to remove

[GitHub] [tvm] junrushao1994 commented on pull request #9728: [BugFix] Fix a wrong use of `std::move()` in cross-thread reduction lowering

junrushao1994 commented on pull request #9728: URL: https://github.com/apache/tvm/pull/9728#issuecomment-992658390 Looks like I triggered an undefined behavior of C++ 梁 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use

[tvm] branch main updated (5557b8c -> cb34604)

This is an automated email from the ASF dual-hosted git repository. mbrookhart pushed a change to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git. from 5557b8c Improve tvmc error message from lazy-loading frontend imports (#9074) add cb34604 [TOPI] Add generic

[GitHub] [tvm] manupa-arm commented on a change in pull request #9704: [USMP] Hill Climb allocator

manupa-arm commented on a change in pull request #9704: URL: https://github.com/apache/tvm/pull/9704#discussion_r767889703 ## File path: src/tir/usmp/algo/hill_climb.cc ## @@ -0,0 +1,356 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more

[GitHub] [tvm] mbrookhart merged pull request #9694: [TOPI] Add generic batch norm

mbrookhart merged pull request #9694: URL: https://github.com/apache/tvm/pull/9694 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail:

[GitHub] [tvm] manupa-arm commented on a change in pull request #9704: [USMP] Hill Climb allocator

manupa-arm commented on a change in pull request #9704: URL: https://github.com/apache/tvm/pull/9704#discussion_r767851543 ## File path: src/tir/usmp/algo/hill_climb.cc ## @@ -0,0 +1,349 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more

[GitHub] [tvm] daniperfer commented on issue #7971: Runtime error when tracing maskrcnn model: Only tensors, lists, tuples of tensors, or dictionary of tensors can be output from traced functions

daniperfer commented on issue #7971: URL: https://github.com/apache/tvm/issues/7971#issuecomment-992612913 Yes, it's in #7990. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific

[GitHub] [tvm-rfcs] Lunderberg commented on pull request #39: [RFC][TIR] Layout transformations on buffer access

Lunderberg commented on pull request #39: URL: https://github.com/apache/tvm-rfcs/pull/39#issuecomment-992608499 Sounds good! I've updated with examples of scheduling with the returned new axes, which work in the implementation posted in

[GitHub] [tvm] MasterJH5574 opened a new pull request #9728: [BugFix] Fix a wrong use of `std::move()` in cross-thread reduction lowering

MasterJH5574 opened a new pull request #9728: URL: https://github.com/apache/tvm/pull/9728 Find a wrong use of `std::move()`. Please see the changes for detail. The wrong use leads to undefined Var of a buffer's `data` field. Given C++ evaluates parameter expressions from the end to

[GitHub] [tvm] Lunderberg commented on pull request #9727: [TE][TIR] Implement layout transformations, non-flat memory buffers

Lunderberg commented on pull request #9727: URL: https://github.com/apache/tvm/pull/9727#issuecomment-992578126 @vinx13 @csullivan @tmoreau89 @adstraw @cconvey -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL

[GitHub] [tvm] Lunderberg opened a new pull request #9727: [TE][TIR] Implement layout transformations, non-flat memory buffers

Lunderberg opened a new pull request #9727: URL: https://github.com/apache/tvm/pull/9727 This is an implementation of [RFC#0039](https://github.com/apache/tvm-rfcs/pull/39), which adds layout transformations to both flat and non-flat memory buffers. The commits in this PR are split

[GitHub] [tvm] abdulazizm commented on issue #8683: Error when using `partition_for_vitis_ai` with YOLOX ONNX model

abdulazizm commented on issue #8683: URL: https://github.com/apache/tvm/issues/8683#issuecomment-992570402 > Hi @comaniac @jlamperez , > > I am following the

[GitHub] [tvm] abdulazizm commented on issue #7971: Runtime error when tracing maskrcnn model: Only tensors, lists, tuples of tensors, or dictionary of tensors can be output from traced functions

abdulazizm commented on issue #7971: URL: https://github.com/apache/tvm/issues/7971#issuecomment-992568025 > I will reformulate the question in a new issue, since this one is already closed... Did you raise the issue for this query? -- This is an automated message from the Apache

[GitHub] [tvm] gromero opened a new pull request #9726: Fix typos in runtime comments

gromero opened a new pull request #9726: URL: https://github.com/apache/tvm/pull/9726 Fix typos in comments about signal handling and entries in the runtime code. Signed-off-by: Gustavo Romero Thanks for contributing to TVM! Please refer to guideline

[GitHub] [tvm] chunit-quic edited a comment on pull request #9723: [Frontend] Add Span filling for frontends to Relay

chunit-quic edited a comment on pull request #9723: URL: https://github.com/apache/tvm/pull/9723#issuecomment-992451852 Hi @lixiaoquan, It's a good advice and easy to implement. :D If tag the original name to the final expression is a better philosophy for users, I can change

[GitHub] [tvm] chunit-quic commented on pull request #9723: [Frontend] Add Span filling for frontends to Relay

chunit-quic commented on pull request #9723: URL: https://github.com/apache/tvm/pull/9723#issuecomment-992451852 Hi @lixiaoquan, It's a good advice and easy to implement. :D If it tag the original name to the final expression is a better philosophy for users, I can change to

[GitHub] [tvm] lucagrementieri edited a comment on issue #9237: [Bug] [PYTORCH] [FRONTEND] Torch and tvm output different

lucagrementieri edited a comment on issue #9237: URL: https://github.com/apache/tvm/issues/9237#issuecomment-943090057 I am facing a similar problem running the script `from_torch.py` published with the tutorial https://tvm.apache.org/docs/how_to/compile_models/from_pytorch.html.

[GitHub] [tvm] hgt312 commented on issue #9237: [Bug] [PYTORCH] [FRONTEND] Torch and tvm output different

hgt312 commented on issue #9237: URL: https://github.com/apache/tvm/issues/9237#issuecomment-992428813 The issue is about boolean indexing, see https://numpy.org/doc/stable/reference/arrays.indexing.html#boolean-array-indexing, while relay's op `adv_index` is integer array indexing.

  1   2   >