[GitHub] [tvm] chunit-quic closed pull request #8893: [Doc] Current passes

chunit-quic closed pull request #8893: URL: https://github.com/apache/tvm/pull/8893 -- 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] tmoreau89 commented on issue #8838: [Tracking Issue][ONNX] Quantized operator support in ONNX importer

tmoreau89 commented on issue #8838: URL: https://github.com/apache/tvm/issues/8838#issuecomment-921430465 Also happy to share that `com.microsoft.QLinearGlobalAveragePool` and `com.microsoft.QLinearAveragePool` are supported, thanks @quic-sanirudh for the support -- This is an

[GitHub] [tvm] tmoreau89 commented on issue #8838: [Tracking Issue][ONNX] Quantized operator support in ONNX importer

tmoreau89 commented on issue #8838: URL: https://github.com/apache/tvm/issues/8838#issuecomment-921430240 Thanks @arangasa - I've added it to the table! -- 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

[GitHub] [tvm] tmoreau89 commented on pull request #9033: [Hexagon] Don't use {} initialization with FastRPC structures

tmoreau89 commented on pull request #9033: URL: https://github.com/apache/tvm/pull/9033#issuecomment-921419859 Thank you @kparzysz-quic @csullivan the PR has been merged -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and

[tvm] branch main updated: [Hexagon] Don't use {} initialization with FastRPC structures (#9033)

This is an automated email from the ASF dual-hosted git repository. moreau 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 1fd8f61 [Hexagon] Don't use {} initialization with

[GitHub] [tvm] tmoreau89 merged pull request #9033: [Hexagon] Don't use {} initialization with FastRPC structures

tmoreau89 merged pull request #9033: URL: https://github.com/apache/tvm/pull/9033 -- 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 edited a comment on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

masahi edited a comment on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921412813 > Can someone explain how the current Stage scope, which ends up in ProducerRealize nodes, would be coherent with the Allocate scope? @mbs-octoml The PR

[GitHub] [tvm] masahi edited a comment on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

masahi edited a comment on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921412813 > Can someone explain how the current Stage scope, which ends up in ProducerRealize nodes, would be coherent with the Allocate scope? @mbs-octoml The PR

[GitHub] [tvm] masahi commented on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

masahi commented on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921412813 > Can someone explain how the current Stage scope, which ends up in ProducerRealize nodes, would be coherent with the Allocate scope? @mbs-octoml The PR

[GitHub] [tvm] masahi commented on a change in pull request #9028: [ONNX][#8838] QLinearSigmoid contrib op and Bug Fix for DequantizeLinear

masahi commented on a change in pull request #9028: URL: https://github.com/apache/tvm/pull/9028#discussion_r710634533 ## File path: python/tvm/relay/frontend/onnx.py ## @@ -3176,7 +3176,7 @@ def _impl_v10(cls, inputs, attr, params): @classmethod def _impl_v13(cls,

[tvm] branch main updated (8ced546 -> af61d1e)

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 8ced546 Fix line break in `setup.py` (#9029) add af61d1e [Onnx] Add SoftmaxCrossEntropyLoss (#8906) No new

[GitHub] [tvm] masahi merged pull request #8906: [Onnx] Add SoftmaxCrossEntropyLoss

masahi merged pull request #8906: URL: https://github.com/apache/tvm/pull/8906 -- 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 (a74ef13 -> 8ced546)

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 a74ef13 [ONNX] QLinearAveragePool and QLinearGlobalAveragePool contrib op (#9017) add 8ced546 Fix line break in

[GitHub] [tvm] masahi merged pull request #9029: Fix line break in `setup.py`

masahi merged pull request #9029: URL: https://github.com/apache/tvm/pull/9029 -- 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] lazycal commented on issue #8505: [Bug] Legalization for nn.dense on cuda does not account for change of units

lazycal commented on issue #8505: URL: https://github.com/apache/tvm/issues/8505#issuecomment-921370695 Closed as it's already fixed in this PR: https://github.com/apache/tvm/pull/9030 -- This is an automated message from the Apache Git Service. To respond to the message, please log on

[GitHub] [tvm] lazycal closed issue #8505: [Bug] Legalization for nn.dense on cuda does not account for change of units

lazycal closed issue #8505: URL: https://github.com/apache/tvm/issues/8505 -- 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] mbrookhart commented on pull request #9017: [ONNX] QLinearAveragePool and QLinearGlobalAveragePool contrib op

mbrookhart commented on pull request #9017: URL: https://github.com/apache/tvm/pull/9017#issuecomment-921318067 Thanks @quic-sanirudh @anwang2009 @AndrewZhaoLuo -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the

[tvm] branch main updated (db78d96 -> a74ef13)

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 db78d96 [CUDA] Fix dense tensorcore legalize type error when units is specified (#9030) add a74ef13 [ONNX]

[GitHub] [tvm] mbrookhart merged pull request #9017: [ONNX] QLinearAveragePool and QLinearGlobalAveragePool contrib op

mbrookhart merged pull request #9017: URL: https://github.com/apache/tvm/pull/9017 -- 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] mbrookhart commented on a change in pull request #9017: [ONNX] QLinearAveragePool and QLinearGlobalAveragePool contrib op

mbrookhart commented on a change in pull request #9017: URL: https://github.com/apache/tvm/pull/9017#discussion_r710553130 ## File path: python/tvm/relay/frontend/onnx.py ## @@ -654,6 +676,40 @@ def _impl_v1(cls, inputs, attr, params): ) +class

[GitHub] [tvm-rfcs] junrushao1994 edited a comment on pull request #22: [RFC][TIR] TIR Non-scalar Constants

junrushao1994 edited a comment on pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#issuecomment-921297958 Thanks @manupa-arm for the reply! These are all good points  Again, I am supportive of this change, and the reason I am asking so many questions is that it is

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

mbs-octoml commented on a change in pull request #9012: URL: https://github.com/apache/tvm/pull/9012#discussion_r710539137 ## File path: src/runtime/logging.cc ## @@ -166,10 +167,127 @@ namespace tvm { namespace runtime { namespace detail { +std::unordered_map

[GitHub] [tvm] mbs-octoml commented on pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

mbs-octoml commented on pull request #9012: URL: https://github.com/apache/tvm/pull/9012#issuecomment-921300444 PTAL. -- 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

[GitHub] [tvm-rfcs] junrushao1994 commented on pull request #22: [RFC][TIR] TIR Non-scalar Constants

junrushao1994 commented on pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#issuecomment-921298252 Also CC @tqchen for comments -- 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

[GitHub] [tvm-rfcs] junrushao1994 commented on pull request #22: [RFC][TIR] TIR Non-scalar Constants

junrushao1994 commented on pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#issuecomment-921297958 Thanks @manupa-arm for the reply! These are all good points. Again, I am supportive of this change, and the reason I am asking so many questions is that it is a change

[GitHub] [tvm] comaniac merged pull request #9030: [CUDA] Fix dense tensorcore legalize type error when units is specified

comaniac merged pull request #9030: URL: https://github.com/apache/tvm/pull/9030 -- 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 #9030: [CUDA] Fix dense tensorcore legalize type error when units is specified

comaniac commented on pull request #9030: URL: https://github.com/apache/tvm/pull/9030#issuecomment-921271385 Thanks @masahi @vinx13 -- 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

[tvm] branch main updated (be37923 -> db78d96)

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 be37923 Implementation of relay_to_tir target hook (#8423) add db78d96 [CUDA] Fix dense tensorcore legalize type

[GitHub] [tvm] comaniac commented on a change in pull request #9030: [CUDA] Fix dense tensorcore legalize type error when units is specified

comaniac commented on a change in pull request #9030: URL: https://github.com/apache/tvm/pull/9030#discussion_r710507437 ## File path: python/tvm/topi/cuda/tensorcore_alter_op.py ## @@ -176,6 +176,12 @@ def _dense_legalize(attrs, inputs, arg_types): x_ = relay.nn.pad(x,

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

mbs-octoml commented on a change in pull request #9012: URL: https://github.com/apache/tvm/pull/9012#discussion_r710502051 ## File path: src/runtime/logging.cc ## @@ -166,10 +167,127 @@ namespace tvm { namespace runtime { namespace detail { +std::unordered_map

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

mbs-octoml commented on a change in pull request #9012: URL: https://github.com/apache/tvm/pull/9012#discussion_r710501694 ## File path: src/runtime/logging.cc ## @@ -166,10 +167,127 @@ namespace tvm { namespace runtime { namespace detail { +std::unordered_map

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

mbs-octoml commented on a change in pull request #9012: URL: https://github.com/apache/tvm/pull/9012#discussion_r710500504 ## File path: include/tvm/runtime/logging.h ## @@ -409,6 +416,68 @@ inline bool DebugLoggingEnabled() { return state == 1; } +/*! \brief Helpers for

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

mbs-octoml commented on a change in pull request #9012: URL: https://github.com/apache/tvm/pull/9012#discussion_r710497929 ## File path: include/tvm/runtime/logging.h ## @@ -409,6 +416,68 @@ inline bool DebugLoggingEnabled() { return state == 1; } +/*! \brief Helpers for

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

mbs-octoml commented on a change in pull request #9012: URL: https://github.com/apache/tvm/pull/9012#discussion_r710497333 ## File path: include/tvm/runtime/logging.h ## @@ -129,8 +132,9 @@ * a = ... * b = ... * // if quit_on_assertion is true, if a==b, continue,

[GitHub] [tvm] masahi commented on a change in pull request #9030: [CUDA] Fix dense tensorcore legalize type error when units is specified

masahi commented on a change in pull request #9030: URL: https://github.com/apache/tvm/pull/9030#discussion_r710496654 ## File path: python/tvm/topi/cuda/tensorcore_alter_op.py ## @@ -176,6 +176,12 @@ def _dense_legalize(attrs, inputs, arg_types): x_ = relay.nn.pad(x,

[GitHub] [tvm] masahi commented on a change in pull request #9030: [CUDA] Fix dense tensorcore legalize type error when units is specified

masahi commented on a change in pull request #9030: URL: https://github.com/apache/tvm/pull/9030#discussion_r710495653 ## File path: python/tvm/topi/cuda/tensorcore_alter_op.py ## @@ -176,6 +176,12 @@ def _dense_legalize(attrs, inputs, arg_types): x_ = relay.nn.pad(x,

[GitHub] [tvm] mbs-octoml commented on a change in pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

mbs-octoml commented on a change in pull request #9012: URL: https://github.com/apache/tvm/pull/9012#discussion_r710487260 ## File path: include/tvm/runtime/logging.h ## @@ -409,6 +416,68 @@ inline bool DebugLoggingEnabled() { return state == 1; } +/*! \brief Helpers for

[GitHub] [tvm] huajsj commented on a change in pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

huajsj commented on a change in pull request #9012: URL: https://github.com/apache/tvm/pull/9012#discussion_r710458479 ## File path: src/runtime/logging.cc ## @@ -166,10 +167,127 @@ namespace tvm { namespace runtime { namespace detail { +std::unordered_map

[GitHub] [tvm] mbrookhart commented on pull request #9028: [ONNX][#8838] QLinearSigmoid contrib op and Bug Fix for DequantizeLinear

mbrookhart commented on pull request #9028: URL: https://github.com/apache/tvm/pull/9028#issuecomment-921233630 The de-quantize change did cause those tests to fail in CI. I think we need a better solution there, but otherwise, it looks good. -- This is an automated message from the

[GitHub] [tvm] tqchen edited a comment on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

tqchen edited a comment on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921220669 Thanks @manupa-arm . I understand that proposal R4 can also work by having a pass to convert "global" to something more specialize as a pass (essentially R1 and R4 are not that

[GitHub] [tvm] tqchen edited a comment on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

tqchen edited a comment on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921220669 Thanks @manupa-arm . I understand that proposal R4 can also work by having a pass to convert "global" to something more specialize as a pass (essentially R1 and R4 are not that

[GitHub] [tvm] tqchen edited a comment on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

tqchen edited a comment on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921220669 Thanks @manupa-arm . I understand that proposal R4 can also work by having a pass to convert "global" to something more specialize as a pass (essentially R1 and R4 are not that

[GitHub] [tvm] tqchen edited a comment on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

tqchen edited a comment on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921220669 Thanks @manupa-arm . I understand that proposal R4 can also work by having a pass to convert "global" to something more specialize as a pass (essentially R1 and R4 are not that

[GitHub] [tvm] tqchen edited a comment on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

tqchen edited a comment on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921220669 Thanks @manupa-arm . I understand that proposal R4 can also work by having a pass to convert "global" to something more specialize as a pass (essentially R1 and R4 are not that

[GitHub] [tvm] huajsj commented on a change in pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

huajsj commented on a change in pull request #9012: URL: https://github.com/apache/tvm/pull/9012#discussion_r710451509 ## File path: include/tvm/runtime/logging.h ## @@ -409,6 +416,68 @@ inline bool DebugLoggingEnabled() { return state == 1; } +/*! \brief Helpers for \p

[GitHub] [tvm] tqchen commented on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

tqchen commented on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921220669 Thanks @manupa-arm . I understand that proposal R4 can also work by having a pass to convert "global" to something more specialize as a pass (essentially R1 and R4 are not that different

[GitHub] [tvm] mehrdadh commented on a change in pull request #9018: [microTVM][autoTVM] Follow up fixes to #9003

mehrdadh commented on a change in pull request #9018: URL: https://github.com/apache/tvm/pull/9018#discussion_r710443115 ## File path: tutorials/micro/micro_autotune.py ## @@ -147,8 +149,8 @@ #build_func=tvm.micro.autotvm_build_func, #) #runner =

[GitHub] [tvm-rfcs] manupa-arm commented on pull request #22: [RFC][TIR] TIR Non-scalar Constants

manupa-arm commented on pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#issuecomment-921211511 @junrushao1994 , For A.) The reason is constants represent something intimate to the compute and requiring space. Moreover, in the scheduling passes where we want

[GitHub] [tvm] huajsj commented on a change in pull request #9018: [microTVM][autoTVM] Follow up fixes to #9003

huajsj commented on a change in pull request #9018: URL: https://github.com/apache/tvm/pull/9018#discussion_r710435576 ## File path: python/tvm/micro/testing.py ## @@ -0,0 +1,33 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license

[GitHub] [tvm] mbs-octoml commented on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

mbs-octoml commented on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921207626 + @csullivan who also needs finer grained scopes Can someone explain how the current Stage scope, which ends up in ProducerRealize nodes, would be coherent with the Allocate

[GitHub] [tvm] mbs-octoml edited a comment on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

mbs-octoml edited a comment on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921207626 plus @csullivan who also needs finer grained scopes Can someone explain how the current Stage scope, which ends up in ProducerRealize nodes, would be coherent with the

[GitHub] [tvm] kparzysz-quic commented on pull request #9033: [Hexagon] Don't use {} initialization with FastRPC structures

kparzysz-quic commented on pull request #9033: URL: https://github.com/apache/tvm/pull/9033#issuecomment-921201575 @csullivan : This could fix the signature verification error you were seeing. -- This is an automated message from the Apache Git Service. To respond to the message, please

[GitHub] [tvm] kparzysz-quic opened a new pull request #9033: [Hexagon] Don't use {} initialization with FastRPC structures

kparzysz-quic opened a new pull request #9033: URL: https://github.com/apache/tvm/pull/9033 The data members in FastRPC structures aren't guaranteed to remain in the same order. Replace aggregate initialization with direct, member-by-member initialization. -- This is an automated

[GitHub] [tvm-rfcs] manupa-arm commented on a change in pull request #22: [RFC][TIR] TIR Non-scalar Constants

manupa-arm commented on a change in pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#discussion_r710429880 ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR:

[GitHub] [tvm-rfcs] comaniac commented on a change in pull request #28: Command Line Composition from Internal Registry

comaniac commented on a change in pull request #28: URL: https://github.com/apache/tvm-rfcs/pull/28#discussion_r710416817 ## File path: rfcs/0028-command-line-registry-composition.md ## @@ -0,0 +1,108 @@ +- Feature Name: Command Line Composition from Internal Registry +- Start

[GitHub] [tvm-rfcs] comaniac commented on a change in pull request #28: Command Line Composition from Internal Registry

comaniac commented on a change in pull request #28: URL: https://github.com/apache/tvm-rfcs/pull/28#discussion_r710414110 ## File path: rfcs/0028-command-line-registry-composition.md ## @@ -0,0 +1,108 @@ +- Feature Name: Command Line Composition from Internal Registry +- Start

[GitHub] [tvm] manupa-arm commented on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

manupa-arm commented on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921152468 Thanks @tqchen . In the R4, I was not suggesting to change TOPI but saying we could just do a Pass to change the storage scope. What are your thoughts about making the pass to

[GitHub] [tvm-rfcs] Mousius commented on a change in pull request #28: Command Line Composition from Internal Registry

Mousius commented on a change in pull request #28: URL: https://github.com/apache/tvm-rfcs/pull/28#discussion_r710375485 ## File path: rfcs/0028-command-line-registry-composition.md ## @@ -0,0 +1,108 @@ +- Feature Name: Command Line Composition from Internal Registry +- Start

[GitHub] [tvm] tqchen commented on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

tqchen commented on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921134533 Thanks @manupa-arm . Trying to capture some of the discussions. - Right now the "global" scope translate to something that can be accessed by CPU, and there was no requirement of

[GitHub] [tvm-rfcs] Mousius commented on a change in pull request #28: Command Line Composition from Internal Registry

Mousius commented on a change in pull request #28: URL: https://github.com/apache/tvm-rfcs/pull/28#discussion_r710363613 ## File path: rfcs/0028-command-line-registry-composition.md ## @@ -0,0 +1,108 @@ +- Feature Name: Command Line Composition from Internal Registry +- Start

[GitHub] [tvm] tqchen edited a comment on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

tqchen edited a comment on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921134533 Thanks @manupa-arm . Trying to capture some of the discussions. - Right now the "global" scope translate to something that can be accessed by CPU, and there was no

[GitHub] [tvm] AndrewZhaoLuo commented on pull request #9023: [TE] Support negative indices

AndrewZhaoLuo commented on pull request #9023: URL: https://github.com/apache/tvm/pull/9023#issuecomment-92599 > I think I sort of understand the usecase here: some negative indices are not known to be negative until in runtime. This forces us to defer the conversion from compile-time

[GitHub] [tvm] junrushao1994 commented on pull request #9016: Support match pvar with dtype constraint

junrushao1994 commented on pull request #9016: URL: https://github.com/apache/tvm/pull/9016#issuecomment-921095868 CC @spectrometerHBH @tqchen @Hzfengsy would you guys review this patch? -- This is an automated message from the Apache Git Service. To respond to the message, please log

[GitHub] [tvm] areusch commented on issue #8717: Fusion of operations and cast in mobilenet v1 conv2d causing large feature maps

areusch commented on issue #8717: URL: https://github.com/apache/tvm/issues/8717#issuecomment-921095159 Adding a pointer to the implementation: https://github.com/sergey-grovety/tvm/commit/6d7bdaa3dd3c5077d9efdc57c691987df6bc9cfa -- This is an automated message from the Apache Git

[GitHub] [tvm] areusch commented on a change in pull request #9012: [Relay] VLOG for finer grained control of hyper-detailed logging

areusch commented on a change in pull request #9012: URL: https://github.com/apache/tvm/pull/9012#discussion_r710243501 ## File path: include/tvm/runtime/logging.h ## @@ -409,6 +416,68 @@ inline bool DebugLoggingEnabled() { return state == 1; } +/*! \brief Helpers for \p

[GitHub] [tvm] comaniac commented on a change in pull request #9030: [CUDA] Fix dense tensorcore legalize type error when units is specified

comaniac commented on a change in pull request #9030: URL: https://github.com/apache/tvm/pull/9030#discussion_r710330599 ## File path: python/tvm/topi/cuda/tensorcore_alter_op.py ## @@ -176,6 +176,12 @@ def _dense_legalize(attrs, inputs, arg_types): x_ = relay.nn.pad(x,

[GitHub] [tvm] vinx13 edited a comment on pull request #9032: [Auto-Schedule][Fix] Fix hang while tune model through rpc

vinx13 edited a comment on pull request #9032: URL: https://github.com/apache/tvm/pull/9032#issuecomment-921088786 > But I have to check it. Maybe the good solution will be to regenerate such arguments which were not predefined for each run. I think this is already the case.

[GitHub] [tvm] vinx13 edited a comment on pull request #9032: [Auto-Schedule][Fix] Fix hang while tune model through rpc

vinx13 edited a comment on pull request #9032: URL: https://github.com/apache/tvm/pull/9032#issuecomment-921088786 > But I have to check it. Maybe the good solution will be to regenerate such arguments which were not predefined for each run. I think this is already the case.

[GitHub] [tvm-rfcs] comaniac commented on a change in pull request #28: Command Line Composition from Internal Registry

comaniac commented on a change in pull request #28: URL: https://github.com/apache/tvm-rfcs/pull/28#discussion_r710315602 ## File path: rfcs/0028-command-line-registry-composition.md ## @@ -0,0 +1,108 @@ +- Feature Name: Command Line Composition from Internal Registry +- Start

[GitHub] [tvm] vinx13 commented on pull request #9032: [Auto-Schedule][Fix] Fix hang while tune model through rpc

vinx13 commented on pull request #9032: URL: https://github.com/apache/tvm/pull/9032#issuecomment-921088786 > But I have to check it. Maybe the good solution will be to regenerate such arguments which were not predefined for each run. I think this is already the case. Currently, in

[GitHub] [tvm-rfcs] manupa-arm commented on a change in pull request #22: [RFC][TIR] TIR Non-scalar Constants

manupa-arm commented on a change in pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#discussion_r710308724 ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR:

[GitHub] [tvm] anwang2009 commented on a change in pull request #9028: [ONNX][#8838] QLinearSigmoid contrib op and Bug Fix for DequantizeLinear

anwang2009 commented on a change in pull request #9028: URL: https://github.com/apache/tvm/pull/9028#discussion_r710307985 ## File path: python/tvm/relay/frontend/onnx.py ## @@ -3176,7 +3176,7 @@ def _impl_v10(cls, inputs, attr, params): @classmethod def

[GitHub] [tvm] anwang2009 commented on a change in pull request #9028: [ONNX][#8838] QLinearSigmoid contrib op and Bug Fix for DequantizeLinear

anwang2009 commented on a change in pull request #9028: URL: https://github.com/apache/tvm/pull/9028#discussion_r710298764 ## File path: python/tvm/relay/frontend/onnx.py ## @@ -3351,6 +3351,28 @@ def _impl_v10(cls, inputs, attr, params): return _qnn.op.quantize(out,

[GitHub] [tvm-rfcs] manupa-arm commented on a change in pull request #22: [RFC][TIR] TIR Non-scalar Constants

manupa-arm commented on a change in pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#discussion_r710301937 ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR:

[GitHub] [tvm-rfcs] manupa-arm commented on a change in pull request #22: [RFC][TIR] TIR Non-scalar Constants

manupa-arm commented on a change in pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#discussion_r710298700 ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR:

[GitHub] [tvm] mbrookhart commented on a change in pull request #9028: [ONNX][#8838] QLinearSigmoid contrib op and Bug Fix for DequantizeLinear

mbrookhart commented on a change in pull request #9028: URL: https://github.com/apache/tvm/pull/9028#discussion_r710293142 ## File path: python/tvm/relay/frontend/onnx.py ## @@ -3176,7 +3176,7 @@ def _impl_v10(cls, inputs, attr, params): @classmethod def

[GitHub] [tvm] AndrewZhaoLuo commented on a change in pull request #9017: [ONNX] QLinearAveragePool and QLinearGlobalAveragePool contrib op

AndrewZhaoLuo commented on a change in pull request #9017: URL: https://github.com/apache/tvm/pull/9017#discussion_r710291824 ## File path: tests/python/frontend/onnx/test_forward.py ## @@ -3056,6 +3056,152 @@ def verify_global_pooling(x_shape, mode):

[GitHub] [tvm] AndrewZhaoLuo commented on a change in pull request #9017: [ONNX] QLinearAveragePool and QLinearGlobalAveragePool contrib op

AndrewZhaoLuo commented on a change in pull request #9017: URL: https://github.com/apache/tvm/pull/9017#discussion_r710288421 ## File path: python/tvm/relay/frontend/onnx.py ## @@ -351,6 +366,13 @@ class AveragePool(Pool): name = "avg_pool" +class

[GitHub] [tvm] echuraev commented on pull request #9032: [Auto-Schedule][Fix] Fix hang while tune model through rpc

echuraev commented on pull request #9032: URL: https://github.com/apache/tvm/pull/9032#issuecomment-921056388 > Is it because `args` not being properly managed? I think so. In case when the `args` is a global dictionary then during the tuning some object hold the `rpc_endpoint` and

[GitHub] [tvm] vinx13 commented on pull request #9032: [Auto-Schedule][Fix] Fix hang while tune model through rpc

vinx13 commented on pull request #9032: URL: https://github.com/apache/tvm/pull/9032#issuecomment-921047912 Sorry I’m a bit confused. Is it because `args` not being properly managed? It should be a python array with numpy ndarray as elements. It will be pickled and sent to workers and

[GitHub] [tvm] manupa-arm commented on issue #9022: [Bug] BuiltinLower does not use alloca for storage on kDLCPU target devices

manupa-arm commented on issue #9022: URL: https://github.com/apache/tvm/issues/9022#issuecomment-921044001 Thanks @tqchen for summarizing the ideas and presenting possible resolutions. The two needs seems very valid. For N0, The operators should really be tagged with 'local'

[GitHub] [tvm] mehrdadh commented on a change in pull request #9018: [microTVM][autoTVM] Follow up fixes to #9003

mehrdadh commented on a change in pull request #9018: URL: https://github.com/apache/tvm/pull/9018#discussion_r710269680 ## File path: python/tvm/micro/testing.py ## @@ -0,0 +1,36 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license

[GitHub] [tvm] echuraev edited a comment on pull request #9032: [Auto-Schedule][Fix] Fix hang while tune model through rpc

echuraev edited a comment on pull request #9032: URL: https://github.com/apache/tvm/pull/9032#issuecomment-921032101 > `args` was added because some workloads like sparse require predefined inputs that shouldn’t be randomly generated. These inputs are looked up from a global dictionary

[GitHub] [tvm] mehrdadh commented on a change in pull request #9018: [microTVM][autoTVM] Follow up fixes to #9003

mehrdadh commented on a change in pull request #9018: URL: https://github.com/apache/tvm/pull/9018#discussion_r710266722 ## File path: src/runtime/crt/host/Makefile ## @@ -38,7 +42,7 @@ CRT_LIBS = $(patsubst %, $(BUILD_DIR)/crt/lib%.a, $(CRT_LIB_NAMES)) CRT_INCLUDES = $(glob

[GitHub] [tvm] jwfromm commented on pull request #8636: [OpenCL] Add vectorization to cuda conv2d_nhwc schedule

jwfromm commented on pull request #8636: URL: https://github.com/apache/tvm/pull/8636#issuecomment-921037478 @jcf94 can you please take another look at this? It's currently being blocked by your request for changes. -- This is an automated message from the Apache Git Service. To respond

[GitHub] [tvm] tmoreau89 commented on pull request #9028: [ONNX][#8838] QLinearSigmoid contrib op and Bug Fix for DequantizeLinear

tmoreau89 commented on pull request #9028: URL: https://github.com/apache/tvm/pull/9028#issuecomment-921036873 CC - @anwang2009 @mbrookhart @AndrewZhaoLuo -- 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

[GitHub] [tvm] echuraev commented on pull request #9032: [Auto-Schedule][Fix] Fix hang while tune model through rpc

echuraev commented on pull request #9032: URL: https://github.com/apache/tvm/pull/9032#issuecomment-921032101 > `args` was added because some workloads like sparse require predefined inputs that shouldn’t be randomly generated. These inputs are looked up from a global dictionary that

[GitHub] [tvm] areusch commented on a change in pull request #8951: [3/10] Moved TIR generation from Python to C++ for CMSIS-NN

areusch commented on a change in pull request #8951: URL: https://github.com/apache/tvm/pull/8951#discussion_r710228149 ## File path: src/relay/backend/contrib/cmsisnn/relay_to_tir.cc ## @@ -0,0 +1,140 @@ + +/* + * Licensed to the Apache Software Foundation (ASF) under one + *

[GitHub] [tvm] grant-arm commented on a change in pull request #8849: [5/6] Arm(R) Ethos(TM)-U NPU codegen integration

grant-arm commented on a change in pull request #8849: URL: https://github.com/apache/tvm/pull/8849#discussion_r710235296 ## File path: src/relay/backend/contrib/ethosu/source_module.cc ## @@ -0,0 +1,320 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + *

[GitHub] [tvm] vinx13 edited a comment on pull request #9032: [Auto-Schedule][Fix] Fix hang while tune model through rpc

vinx13 edited a comment on pull request #9032: URL: https://github.com/apache/tvm/pull/9032#issuecomment-921004809 `args` was added because some workloads like sparse require predefined inputs that shouldn’t be randomly generated. These inputs are looked up from a global dictionary that

[GitHub] [tvm] vinx13 commented on pull request #9032: [Auto-Schedule][Fix] Fix hang while tune model through rpc

vinx13 commented on pull request #9032: URL: https://github.com/apache/tvm/pull/9032#issuecomment-921004809 `args` was added because some workloads like sparse require predefined inputs that shouldn’t be randomly generated. These inputs are looked up from a global dictionary that won’t be

[tvm] branch main updated (02fbaf0 -> be37923)

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 02fbaf0 [Onnx] Fix NLL Loss tests (#8971) add be37923 Implementation of relay_to_tir target hook (#8423) No new

[GitHub] [tvm] areusch merged pull request #8423: Implementation of relay_to_tir target hook

areusch merged pull request #8423: URL: https://github.com/apache/tvm/pull/8423 -- 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 commented on a change in pull request #8423: Implementation of relay_to_tir target hook

areusch commented on a change in pull request #8423: URL: https://github.com/apache/tvm/pull/8423#discussion_r710224938 ## File path: cmake/modules/contrib/ExampleTargetHooks.cmake ## @@ -0,0 +1,19 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more

[GitHub] [tvm] areusch commented on a change in pull request #9018: [microTVM][autoTVM] Follow up fixes to #9003

areusch commented on a change in pull request #9018: URL: https://github.com/apache/tvm/pull/9018#discussion_r710209423 ## File path: src/runtime/rpc/rpc_endpoint.cc ## @@ -692,12 +692,12 @@ void RPCEndpoint::Init() { */ std::shared_ptr RPCEndpoint::Create(std::unique_ptr

[GitHub] [tvm] mehrdadh commented on a change in pull request #9018: [microTVM][autoTVM] Follow up fixes to #9003

mehrdadh commented on a change in pull request #9018: URL: https://github.com/apache/tvm/pull/9018#discussion_r710195264 ## File path: python/tvm/micro/testing.py ## @@ -0,0 +1,36 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license

[GitHub] [tvm] echuraev commented on pull request #9032: [Auto-Schedule][Fix] Fix hang while tune model through rpc

echuraev commented on pull request #9032: URL: https://github.com/apache/tvm/pull/9032#issuecomment-920961268 @shingjan please take a look on this PR. After #8492 the tuning doesn't work in proper way. It hangs. The problem is in the resources management and global dictionary with

[GitHub] [tvm] Mousius commented on issue #8717: Fusion of operations and cast in mobilenet v1 conv2d causing large feature maps

Mousius commented on issue #8717: URL: https://github.com/apache/tvm/issues/8717#issuecomment-920936729 This looks great @Alex-grovety, a huge improvement! As far as next steps, ideally from here:

[GitHub] [tvm] ashutosh-arm commented on a change in pull request #8951: [3/10] Moved TIR generation from Python to C++ for CMSIS-NN

ashutosh-arm commented on a change in pull request #8951: URL: https://github.com/apache/tvm/pull/8951#discussion_r710125210 ## File path: src/relay/backend/contrib/cmsisnn/codegen_cmsisnn.cc ## @@ -16,190 +16,36 @@ * specific language governing permissions and limitations

[GitHub] [tvm] Mousius commented on a change in pull request #8951: [3/10] Moved TIR generation from Python to C++ for CMSIS-NN

Mousius commented on a change in pull request #8951: URL: https://github.com/apache/tvm/pull/8951#discussion_r710119903 ## File path: src/relay/backend/contrib/cmsisnn/codegen_cmsisnn.cc ## @@ -16,190 +16,36 @@ * specific language governing permissions and limitations *

  1   2   >