[GitHub] [tvm] manupa-arm commented on a change in pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

manupa-arm commented on a change in pull request #8795: URL: https://github.com/apache/tvm/pull/8795#discussion_r695438902 ## File path: python/tvm/relay/op/contrib/ethosu.py ## @@ -0,0 +1,251 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contribut

[GitHub] [tvm] manupa-arm commented on a change in pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

manupa-arm commented on a change in pull request #8795: URL: https://github.com/apache/tvm/pull/8795#discussion_r695439000 ## File path: python/tvm/relay/op/contrib/ethosu.py ## @@ -0,0 +1,251 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contribut

[GitHub] [tvm] manupa-arm commented on a change in pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

manupa-arm commented on a change in pull request #8795: URL: https://github.com/apache/tvm/pull/8795#discussion_r695439618 ## File path: tests/python/contrib/test_ethosu/relay_ir_builder.py ## @@ -0,0 +1,295 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or

[GitHub] [tvm] manupa-arm commented on a change in pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

manupa-arm commented on a change in pull request #8795: URL: https://github.com/apache/tvm/pull/8795#discussion_r695440184 ## File path: tests/python/contrib/test_ethosu/test_legalize.py ## @@ -0,0 +1,345 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or mo

[GitHub] [tvm] manupa-arm commented on pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

manupa-arm commented on pull request #8795: URL: https://github.com/apache/tvm/pull/8795#issuecomment-905225146 @jroesch Thanks for the comments! I've addressed majority of them bar TE parts. @mbaret Thanks for the clarification with the RFC and PR strategy. Would you be able take

[GitHub] [tvm] manupa-arm edited a comment on pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

manupa-arm edited a comment on pull request #8795: URL: https://github.com/apache/tvm/pull/8795#issuecomment-905225146 @jroesch Thanks for the comments! I've addressed majority of them bar TE parts. @mbaret Thanks for the clarification with the RFC and PR strategy. Would you be ab

[GitHub] [tvm] manupa-arm edited a comment on pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

manupa-arm edited a comment on pull request #8795: URL: https://github.com/apache/tvm/pull/8795#issuecomment-905225146 @jroesch Thanks for the comments! I've addressed majority of them bar TE parts. Also regarding the python cpp split, the reason is some of the passes need to access

[GitHub] [tvm] sunjiweiswift commented on a change in pull request #8497: [Runtime] add set_output_zero_copy

sunjiweiswift commented on a change in pull request #8497: URL: https://github.com/apache/tvm/pull/8497#discussion_r695452531 ## File path: tests/cpp/build_module_test.cc ## @@ -199,3 +199,129 @@ TEST(BuildModule, Heterogeneous) { ICHECK_LT(std::fabs(p_out[i] - (i + (i + 1

[GitHub] [tvm] manupa-arm commented on a change in pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

manupa-arm commented on a change in pull request #8795: URL: https://github.com/apache/tvm/pull/8795#discussion_r695440184 ## File path: tests/python/contrib/test_ethosu/test_legalize.py ## @@ -0,0 +1,345 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or mo

[GitHub] [tvm] junrushao1994 commented on pull request #8778: [BUG] ToBasicBlockNormalForm immutability

junrushao1994 commented on pull request #8778: URL: https://github.com/apache/tvm/pull/8778#issuecomment-905245825 Yeah there is some flakiness in the CI -- 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 t

[GitHub] [tvm] masahi merged pull request #8781: [Frontend] [Torch] [ONNX] GRU layer

masahi merged pull request #8781: URL: https://github.com/apache/tvm/pull/8781 -- 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: commits-unsubscr...

[GitHub] [tvm] masahi commented on pull request #8781: [Frontend] [Torch] [ONNX] GRU layer

masahi commented on pull request #8781: URL: https://github.com/apache/tvm/pull/8781#issuecomment-905267004 Thanks @vvchernov -- 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 commen

[GitHub] [tvm] masahi commented on pull request #8810: [AMP] Disallow fp16 conversion for summation-like ops

masahi commented on pull request #8810: URL: https://github.com/apache/tvm/pull/8810#issuecomment-905289599 @AndrewZhaoLuo A possible flaky test from test_to_mixed_precision.py: `test_let_statement_simple` https://ci.tlcpack.ai/blue/organizations/jenkins/tvm/detail/PR-8810/3/pipeline

[GitHub] [tvm] manupa-arm commented on a change in pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

manupa-arm commented on a change in pull request #8795: URL: https://github.com/apache/tvm/pull/8795#discussion_r695550561 ## File path: python/tvm/relay/backend/contrib/ethosu/util.py ## @@ -0,0 +1,198 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more

[GitHub] [tvm] Meteorix commented on pull request #8777: [PyTorch][WIP]Add PyTorchTVM: compile torchscript to tvm and export as pytorch_op

Meteorix commented on pull request #8777: URL: https://github.com/apache/tvm/pull/8777#issuecomment-905320199 Update: * RFC: https://github.com/apache/tvm-rfcs/pull/25 * discuss: https://discuss.tvm.apache.org/t/rfc-pytorchtvm-compile-torchscript-to-tvm-and-use-accelerated-module-in-p

[GitHub] [tvm-rfcs] masahi commented on pull request #25: [RFC]PyTorchTVM

masahi commented on pull request #25: URL: https://github.com/apache/tvm-rfcs/pull/25#issuecomment-905322526 Is this supposed to be useful for accelerating PyTorch training? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and

[GitHub] [tvm] manupa-arm commented on a change in pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

manupa-arm commented on a change in pull request #8795: URL: https://github.com/apache/tvm/pull/8795#discussion_r695571625 ## File path: python/tvm/relay/op/contrib/ethosu.py ## @@ -0,0 +1,251 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contribut

[GitHub] [tvm] jcf94 commented on pull request #8777: [PyTorch][WIP]Add PyTorchTVM: compile torchscript to tvm and export as pytorch_op

jcf94 commented on pull request #8777: URL: https://github.com/apache/tvm/pull/8777#issuecomment-905338556 Looks great! We have ever implemented a similar approach to put a optimized tvm graph runtime back to TensorFlow and Pytorch through custom ops. For some reason we're not able to s

[GitHub] [tvm] ashutosh-arm commented on a change in pull request #8833: [2/10] CMSIS-NN code generator for softmax

ashutosh-arm commented on a change in pull request #8833: URL: https://github.com/apache/tvm/pull/8833#discussion_r695582274 ## File path: python/tvm/relay/backend/contrib/cmsisnn/codegen.py ## @@ -0,0 +1,140 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# o

[GitHub] [tvm] Mousius commented on pull request #8835: Change target string to Target object in the TE compiler and interpreter

Mousius commented on pull request #8835: URL: https://github.com/apache/tvm/pull/8835#issuecomment-905352332 Hmm, something weird is occurring here, unsure why those tests didn't run for me locally - but debugging `test_any.py` a bit I ran `tvm::Dump(per_target_module_)` and got (removed f

[GitHub] [tvm] ashutosh-arm commented on a change in pull request #8833: [2/10] CMSIS-NN code generator for softmax

ashutosh-arm commented on a change in pull request #8833: URL: https://github.com/apache/tvm/pull/8833#discussion_r695595860 ## File path: tests/python/contrib/test_cmsisnn/test_softmax.py ## @@ -0,0 +1,197 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or

[GitHub] [tvm] ashutosh-arm commented on a change in pull request #8833: [2/10] CMSIS-NN code generator for softmax

ashutosh-arm commented on a change in pull request #8833: URL: https://github.com/apache/tvm/pull/8833#discussion_r695597658 ## File path: python/tvm/relay/backend/contrib/cmsisnn/codegen.py ## @@ -0,0 +1,140 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# o

[GitHub] [tvm] ashutosh-arm commented on a change in pull request #8833: [2/10] CMSIS-NN code generator for softmax

ashutosh-arm commented on a change in pull request #8833: URL: https://github.com/apache/tvm/pull/8833#discussion_r695598599 ## File path: python/tvm/relay/backend/contrib/cmsisnn/codegen.py ## @@ -0,0 +1,140 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# o

[GitHub] [tvm] ashutosh-arm commented on a change in pull request #8833: [2/10] CMSIS-NN code generator for softmax

ashutosh-arm commented on a change in pull request #8833: URL: https://github.com/apache/tvm/pull/8833#discussion_r695599788 ## File path: src/relay/backend/contrib/cmsisnn/codegen_cmsisnn.cc ## @@ -0,0 +1,215 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under on

[GitHub] [tvm] Mousius opened a new pull request #8840: Force CMake targets in top-level Makefile to run

Mousius opened a new pull request #8840: URL: https://github.com/apache/tvm/pull/8840 This is a bug I introduced in https://github.com/apache/tvm/pull/8809, because the built binary is now named `build/cpptest` when `make` checks that artifact it finds it exists already and skips running `

[GitHub] [tvm] Mousius commented on pull request #8840: Force CMake targets in top-level Makefile to run

Mousius commented on pull request #8840: URL: https://github.com/apache/tvm/pull/8840#issuecomment-905364710 @Lunderberg could you take a look at this please? :smile_cat: -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and u

[GitHub] [tvm] mehrdadh opened a new pull request #8841: Update CI Lint Image Version

mehrdadh opened a new pull request #8841: URL: https://github.com/apache/tvm/pull/8841 Updating image version for this issue: https://github.com/apache/tvm/issues/8836 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and

[GitHub] [tvm] mehrdadh opened a new issue #8842: [Test][Possible flaky test] QEMU timeout on zephyr test_relay

mehrdadh opened a new issue #8842: URL: https://github.com/apache/tvm/issues/8842 https://ci.tlcpack.ai/blue/organizations/jenkins/tvm/detail/PR-8822/4/pipeline/#step-203-log-1971 ``` === FAILURES ===

[GitHub] [tvm] mehrdadh commented on pull request #8841: Update CI Lint Image Version

mehrdadh commented on pull request #8841: URL: https://github.com/apache/tvm/pull/8841#issuecomment-905376239 cc @leandron @jroesch @junrushao1994 -- 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 t

[GitHub] [tvm] ekalda opened a new pull request #8843: [Pattern matching] Add an option to rewrite the graph only once

ekalda opened a new pull request #8843: URL: https://github.com/apache/tvm/pull/8843 If the graph returned from the callback consists of the original pattern, the rewriter will run in the loop which is not always desired. So this patch proposes an option to run the rewriter only once.

[GitHub] [tvm] ashutosh-arm commented on a change in pull request #8833: [2/10] CMSIS-NN code generator for softmax

ashutosh-arm commented on a change in pull request #8833: URL: https://github.com/apache/tvm/pull/8833#discussion_r695627499 ## File path: src/relay/backend/contrib/cmsisnn/codegen_cmsisnn.cc ## @@ -0,0 +1,215 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under on

[GitHub] [tvm] ashutosh-arm commented on a change in pull request #8833: [2/10] CMSIS-NN code generator for softmax

ashutosh-arm commented on a change in pull request #8833: URL: https://github.com/apache/tvm/pull/8833#discussion_r695629688 ## File path: tests/python/contrib/test_cmsisnn/test_softmax.py ## @@ -0,0 +1,197 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or

[GitHub] [tvm] elvin-n commented on pull request #8793: Enable python debug runtime for exported network libraries

elvin-n commented on pull request #8793: URL: https://github.com/apache/tvm/pull/8793#issuecomment-905394016 @tkonolige addressed 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 the spe

[GitHub] [tvm] Mousius commented on a change in pull request #8833: [2/10] CMSIS-NN code generator for softmax

Mousius commented on a change in pull request #8833: URL: https://github.com/apache/tvm/pull/8833#discussion_r695636290 ## File path: python/tvm/relay/backend/contrib/cmsisnn/codegen.py ## @@ -0,0 +1,140 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or mor

[GitHub] [tvm] ekalda commented on pull request #8843: [Pattern matching] Add an option to rewrite the graph only once

ekalda commented on pull request #8843: URL: https://github.com/apache/tvm/pull/8843#issuecomment-905394703 @mbrookhart @mbaret -- 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 comm

[GitHub] [tvm] ashutosh-arm commented on a change in pull request #8833: [2/10] CMSIS-NN code generator for softmax

ashutosh-arm commented on a change in pull request #8833: URL: https://github.com/apache/tvm/pull/8833#discussion_r695640554 ## File path: python/tvm/relay/backend/contrib/cmsisnn/codegen.py ## @@ -0,0 +1,140 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# o

[GitHub] [tvm] Mousius opened a new pull request #8844: Apply CPPLint to CRT Tests

Mousius opened a new pull request #8844: URL: https://github.com/apache/tvm/pull/8844 This one was a bit trickier as there was more usage of dynamic arrays and less safe casts. I've tried to minimise the changes to just those required to passing linting. -- This is an automated mess

[GitHub] [tvm] mehrdadh commented on a change in pull request #8748: [microTVM][RVM] Add Arduino RVM

mehrdadh commented on a change in pull request #8748: URL: https://github.com/apache/tvm/pull/8748#discussion_r695677385 ## File path: apps/microtvm/reference-vm/base-box-tool.py ## @@ -439,26 +473,27 @@ def parse_args(): "--provider", choices=ALL_PROVIDERS,

[GitHub] [tvm] mehrdadh commented on pull request #8748: [microTVM][RVM] Add Arduino RVM

mehrdadh commented on pull request #8748: URL: https://github.com/apache/tvm/pull/8748#issuecomment-905432350 @guberti thanks for the fixes. Can you explain what does `platform-version` mean in the case of arduino? -- This is an automated message from the Apache Git Service. To respond t

[GitHub] [tvm] mehrdadh edited a comment on pull request #8748: [microTVM][RVM] Add Arduino RVM

mehrdadh edited a comment on pull request #8748: URL: https://github.com/apache/tvm/pull/8748#issuecomment-905432350 @guberti thanks for the fixes. Can you explain what does `platform-version` mean in the case of arduino? I think we need to document that in the README file for people who w

[GitHub] [tvm] mehrdadh commented on a change in pull request #8748: [microTVM][RVM] Add Arduino RVM

mehrdadh commented on a change in pull request #8748: URL: https://github.com/apache/tvm/pull/8748#discussion_r695680883 ## File path: apps/microtvm/reference-vm/arduino/Vagrantfile ## @@ -0,0 +1,66 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more con

[GitHub] [tvm] mehrdadh edited a comment on pull request #8748: [microTVM][RVM] Add Arduino RVM

mehrdadh edited a comment on pull request #8748: URL: https://github.com/apache/tvm/pull/8748#issuecomment-905432350 @guberti thanks for the fixes. Can you explain what does `platform-version` mean in the case of arduino? I think we need to document that in the help message. can you add fo

[GitHub] [tvm] Lunderberg commented on pull request #8840: Force CMake targets in top-level Makefile to run

Lunderberg commented on pull request #8840: URL: https://github.com/apache/tvm/pull/8840#issuecomment-905511312 Looks good to me, and thank you for the fix! Looks like I accidentally left that trap in #8317, since at the time none of the cmake rules produced a file with the same name as t

[GitHub] [tvm] Mousius commented on a change in pull request #8833: [2/10] CMSIS-NN code generator for softmax

Mousius commented on a change in pull request #8833: URL: https://github.com/apache/tvm/pull/8833#discussion_r695726706 ## File path: src/relay/backend/contrib/cmsisnn/codegen_cmsisnn.cc ## @@ -0,0 +1,199 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + *

[GitHub] [tvm] Lunderberg edited a comment on pull request #8840: Force CMake targets in top-level Makefile to run

Lunderberg edited a comment on pull request #8840: URL: https://github.com/apache/tvm/pull/8840#issuecomment-905511312 Looks good to me, and thank you for the fix! Looks like I accidentally left that trap in #8317, since at the time none of the cmake rules produced a file with the same na

[GitHub] [tvm] mehrdadh opened a new pull request #8846: [microtvm][Zephyr] Increase timeout to fix flaky tests

mehrdadh opened a new pull request #8846: URL: https://github.com/apache/tvm/pull/8846 This PR increases timeout to fix qemu flaky tests: https://github.com/apache/tvm/issues/8842 cc @Mousius @areusch @gromero -- This is an automated message from the Apache Git Service. To re

[GitHub] [tvm] hgt312 opened a new pull request #8847: [Pass] Make Module Pass always return a new IRModule

hgt312 opened a new pull request #8847: URL: https://github.com/apache/tvm/pull/8847 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-rfcs] u99127 commented on a change in pull request #11: [RFC] Arm Ethos-U Integration

u99127 commented on a change in pull request #11: URL: https://github.com/apache/tvm-rfcs/pull/11#discussion_r695811906 ## File path: rfcs/0011_Arm_Ethos-U_Integration.md ## @@ -192,42 +191,42 @@ TVM_DLL int32_t ethosu_0(TVMValue* args, int* type_code, int num_args, TVMValue*

[GitHub] [tvm-rfcs] u99127 commented on a change in pull request #11: [RFC] Arm Ethos-U Integration

u99127 commented on a change in pull request #11: URL: https://github.com/apache/tvm-rfcs/pull/11#discussion_r695811906 ## File path: rfcs/0011_Arm_Ethos-U_Integration.md ## @@ -192,42 +191,42 @@ TVM_DLL int32_t ethosu_0(TVMValue* args, int* type_code, int num_args, TVMValue*

[GitHub] [tvm] Mousius commented on a change in pull request #8795: [2/6] Arm(R) Ethos(TM)-U NPU Relay passes and Conv2D op

Mousius commented on a change in pull request #8795: URL: https://github.com/apache/tvm/pull/8795#discussion_r695768083 ## File path: python/tvm/relay/backend/contrib/ethosu/legalize.py ## @@ -0,0 +1,218 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or mor

[GitHub] [tvm] mbs-octoml commented on pull request #8835: Change target string to Target object in the TE compiler and interpreter

mbs-octoml commented on pull request #8835: URL: https://github.com/apache/tvm/pull/8835#issuecomment-905558007 Are the hash and equality on Targets legit? Also, possibly relevant, there's a bit of hackery to force shape functions to end up on the 'cpu' target which is hard coded, so it

[GitHub] [tvm] mbs-octoml commented on pull request #8835: Change target string to Target object in the TE compiler and interpreter

mbs-octoml commented on pull request #8835: URL: https://github.com/apache/tvm/pull/8835#issuecomment-905573893 (3oz of coffee later) It will be the on-the-fly Target creation for the shape functions. By switching from String to Target we no longer implicitly identify distinct targets o

[GitHub] [tvm] kparzysz-quic commented on a change in pull request #8823: [Hexagon] Rework tvm.target.hexagon() interface

kparzysz-quic commented on a change in pull request #8823: URL: https://github.com/apache/tvm/pull/8823#discussion_r695835863 ## File path: python/tvm/target/target.py ## @@ -510,23 +548,27 @@ def validate_hvx_length(codegen_hvx, sim_args): + cpu_attr["post"]

[GitHub] [tvm] kparzysz-quic commented on a change in pull request #8823: [Hexagon] Rework tvm.target.hexagon() interface

kparzysz-quic commented on a change in pull request #8823: URL: https://github.com/apache/tvm/pull/8823#discussion_r695837217 ## File path: python/tvm/target/target.py ## @@ -413,79 +413,117 @@ def bifrost(model="unknown", options=None): return Target(" ".join(["opencl"] +

[GitHub] [tvm] mbrookhart commented on a change in pull request #8845: Frontend: add onnx GlobalLpPool op

mbrookhart commented on a change in pull request #8845: URL: https://github.com/apache/tvm/pull/8845#discussion_r695841609 ## File path: python/tvm/relay/frontend/onnx.py ## @@ -961,6 +961,18 @@ def _impl_v1(cls, inputs, attr, params): return _op.power(out, reci_p)

[GitHub] [tvm] Mousius commented on pull request #8835: Change target string to Target object in the TE compiler and interpreter

Mousius commented on pull request #8835: URL: https://github.com/apache/tvm/pull/8835#issuecomment-905584474 Hmm, I picked on it calling into that block within `LowerTE` before the `shape_target` existed via this call: ``` std::pair> main_and_lowered = Prepare(mod_and_globa

[GitHub] [tvm] tqchen commented on pull request #8847: [Pass] Make Module Pass always return a new IRModule

tqchen commented on pull request #8847: URL: https://github.com/apache/tvm/pull/8847#issuecomment-905589054 Thanks @hgt312 , may I ask the motivation of the change? The original intend of the code is that the pass should return a new module if necessary, but can perform copyon write optimi

[GitHub] [tvm] leandron merged pull request #8840: Force CMake targets in top-level Makefile to run

leandron merged pull request #8840: URL: https://github.com/apache/tvm/pull/8840 -- 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: commits-unsubscr.

[GitHub] [tvm] guberti commented on a change in pull request #8748: [microTVM][RVM] Add Arduino RVM

guberti commented on a change in pull request #8748: URL: https://github.com/apache/tvm/pull/8748#discussion_r695902185 ## File path: apps/microtvm/reference-vm/base-box-tool.py ## @@ -439,26 +473,27 @@ def parse_args(): "--provider", choices=ALL_PROVIDERS,

[GitHub] [tvm] tkonolige commented on a change in pull request #8793: Enable python debug runtime for exported network libraries

tkonolige commented on a change in pull request #8793: URL: https://github.com/apache/tvm/pull/8793#discussion_r695902578 ## File path: tests/python/unittest/test_runtime_module_based_interface.py ## @@ -90,6 +90,25 @@ def test_cpu(): tvm.testing.assert_allclose(out, verif

[GitHub] [tvm] junrushao1994 commented on pull request #8847: [Pass] Make Module Pass always return a new IRModule

junrushao1994 commented on pull request #8847: URL: https://github.com/apache/tvm/pull/8847#issuecomment-905680014 I am curious too. Isn’t copy-on-write good enough? -- 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] hgt312 commented on pull request #8847: [Pass] Make Module Pass always return a new IRModule

hgt312 commented on pull request #8847: URL: https://github.com/apache/tvm/pull/8847#issuecomment-905689897 The motivation is just to prevent processing one module with the same pass multiple times, it makes the result unexpected and I found that function pass always return a new Module.

[GitHub] [tvm] junrushao1994 commented on pull request #8841: Update CI Lint Image Version

junrushao1994 commented on pull request #8841: URL: https://github.com/apache/tvm/pull/8841#issuecomment-905693422 hey would you mind also updating ci-cpu/gpu? Thanks a lot! -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and

[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-905696426 Adding @cconvey to the table who is working on `QLinearMatMul` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use

[GitHub] [tvm] comaniac commented on a change in pull request #8843: [Pattern matching] Add an option to rewrite the graph only once

comaniac commented on a change in pull request #8843: URL: https://github.com/apache/tvm/pull/8843#discussion_r695929809 ## File path: tests/python/relay/test_dataflow_pattern.py ## @@ -1793,3 +1825,4 @@ def test_partition_constant_embedding(): test_IfPattern() test_m

[GitHub] [tvm] tmoreau89 merged pull request #8822: [Hexagon] Reuse Hexagon SDK analysis across cmake files

tmoreau89 merged pull request #8822: URL: https://github.com/apache/tvm/pull/8822 -- 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: commits-unsubscr

[GitHub] [tvm] tmoreau89 commented on pull request #8822: [Hexagon] Reuse Hexagon SDK analysis across cmake files

tmoreau89 commented on pull request #8822: URL: https://github.com/apache/tvm/pull/8822#issuecomment-905702785 Thank you @kparzysz-quic @tkonolige @Lunderberg - 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] [tvm] tmoreau89 commented on pull request #8822: [Hexagon] Reuse Hexagon SDK analysis across cmake files

tmoreau89 commented on pull request #8822: URL: https://github.com/apache/tvm/pull/8822#issuecomment-905702962 Flaky test issue is being tracked here: https://github.com/apache/tvm/issues/8842 -- This is an automated message from the Apache Git Service. To respond to the message, please

[GitHub] [tvm] tmoreau89 commented on issue #8842: [Test][Possible flaky test] QEMU timeout on zephyr test_relay

tmoreau89 commented on issue #8842: URL: https://github.com/apache/tvm/issues/8842#issuecomment-905703540 Thanks for starting this issue @mehrdadh. For reference this happened in this PR. The issue was circumvented by re-triggering CI. Changes are likely unrelated to QEMU or uTVM altogethe

[GitHub] [tvm] jverma-quic commented on a change in pull request #8773: Add support for QLinearMul ONNX op

jverma-quic commented on a change in pull request #8773: URL: https://github.com/apache/tvm/pull/8773#discussion_r695936244 ## File path: python/tvm/relay/frontend/onnx.py ## @@ -3279,6 +3279,40 @@ def get_scalar(x, dtype="float32"): return _qnn.op.quantize(out, c_scal

[GitHub] [tvm] mehrdadh commented on pull request #8841: Update CI Lint Image Version

mehrdadh commented on pull request #8841: URL: https://github.com/apache/tvm/pull/8841#issuecomment-905708492 @junrushao1994 sure, since this is almost done, I'll send a separate PR. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to G

[GitHub] [tvm] mbrookhart commented on a change in pull request #8773: Add support for QLinearMul ONNX op

mbrookhart commented on a change in pull request #8773: URL: https://github.com/apache/tvm/pull/8773#discussion_r695939489 ## File path: python/tvm/relay/frontend/onnx.py ## @@ -3279,6 +3279,40 @@ def get_scalar(x, dtype="float32"): return _qnn.op.quantize(out, c_scale

[GitHub] [tvm] electriclilies commented on pull request #8835: Change target string to Target object in the TE compiler and interpreter

electriclilies commented on pull request #8835: URL: https://github.com/apache/tvm/pull/8835#issuecomment-905711627 @mbs-octoml I was doing a bit of debugging yesterday and saw that behavior too-- there are some cases where we try to look up a target that is LLVM CPU in a map but we don't

[GitHub] [tvm] electriclilies edited a comment on pull request #8835: Change target string to Target object in the TE compiler and interpreter

electriclilies edited a comment on pull request #8835: URL: https://github.com/apache/tvm/pull/8835#issuecomment-905711627 @mbs-octoml I was doing a bit of debugging yesterday and saw that behavior too-- there are some cases where we try to look up a target that is LLVM CPU in a map but we

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

sergey-grovety commented on issue #8717: URL: https://github.com/apache/tvm/issues/8717#issuecomment-905720857 Hello @Mousius, I would like to clarify something about the part of the problem related to casting to int16. Please look at `/tvm/python/tvm/relay/qnn/op/legalizations.py

[GitHub] [tvm] ccjoechou opened a new issue #8848: pooling convert layout bug in pooling.cc and in test_pass_convert_op_layout.py?

ccjoechou opened a new issue #8848: URL: https://github.com/apache/tvm/issues/8848 We are prototyping a BYOC flow and have a need to convert the layout of each nn.max_pool2d call in IR graph to use the attrs["layout"] = "NHWC" format. Based on our analysis, we found 2 possible issues bel

[GitHub] [tvm] csullivan commented on a change in pull request #8823: [Hexagon] Rework tvm.target.hexagon() interface

csullivan commented on a change in pull request #8823: URL: https://github.com/apache/tvm/pull/8823#discussion_r695962970 ## File path: python/tvm/target/target.py ## @@ -413,79 +413,118 @@ def bifrost(model="unknown", options=None): return Target(" ".join(["opencl"] + opt

[GitHub] [tvm] zxybazh commented on pull request #8823: [Hexagon] Rework tvm.target.hexagon() interface

zxybazh commented on pull request #8823: URL: https://github.com/apache/tvm/pull/8823#issuecomment-905732933 LGTM. -- 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 unsub

[GitHub] [tvm] zxybazh removed a comment on pull request #8823: [Hexagon] Rework tvm.target.hexagon() interface

zxybazh removed a comment on pull request #8823: URL: https://github.com/apache/tvm/pull/8823#issuecomment-905732933 LGTM. -- 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.

[GitHub] [tvm] junrushao1994 commented on pull request #8835: Change target string to Target object in the TE compiler and interpreter

junrushao1994 commented on pull request #8835: URL: https://github.com/apache/tvm/pull/8835#issuecomment-905745045 Hey Lily, I could imagine there are tons of challenges here and thank you for the really hard work! I feel like option 1 might be the most convenient way we could go wit

[GitHub] [tvm] manupa-arm opened a new pull request #8849: [5/6] Arm(R) Ethos(TM)-U NPU codegen integration

manupa-arm opened a new pull request #8849: URL: https://github.com/apache/tvm/pull/8849 This commit integrates the codegen for Arm® Ethos™-U. * Adding Conv2D tests and a mobilenet_v1 conv2d offload test. This PR is blocked on merging #8795 , #8806 and #8811. Since this is built o

[GitHub] [tvm] kparzysz-quic commented on a change in pull request #8823: [Hexagon] Rework tvm.target.hexagon() interface

kparzysz-quic commented on a change in pull request #8823: URL: https://github.com/apache/tvm/pull/8823#discussion_r695984580 ## File path: python/tvm/target/target.py ## @@ -413,79 +413,118 @@ def bifrost(model="unknown", options=None): return Target(" ".join(["opencl"] +

[GitHub] [tvm] jverma-quic commented on a change in pull request #8773: Add support for QLinearMul ONNX op

jverma-quic commented on a change in pull request #8773: URL: https://github.com/apache/tvm/pull/8773#discussion_r695992594 ## File path: python/tvm/relay/frontend/onnx.py ## @@ -3279,6 +3279,40 @@ def get_scalar(x, dtype="float32"): return _qnn.op.quantize(out, c_scal

[GitHub] [tvm] comaniac commented on a change in pull request #8497: [Runtime] add set_output_zero_copy

comaniac commented on a change in pull request #8497: URL: https://github.com/apache/tvm/pull/8497#discussion_r696007010 ## File path: tests/cpp/build_module_test.cc ## @@ -199,3 +199,129 @@ TEST(BuildModule, Heterogeneous) { ICHECK_LT(std::fabs(p_out[i] - (i + (i + 1.0) -

[GitHub] [tvm] tqchen merged pull request #8382: [Pre-commit] Add pre-commit configuration to perform minimal checks locally

tqchen merged pull request #8382: URL: https://github.com/apache/tvm/pull/8382 -- 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: commits-unsubscr...

[GitHub] [tvm] tqchen commented on pull request #8382: [Pre-commit] Add pre-commit configuration to perform minimal checks locally

tqchen commented on pull request #8382: URL: https://github.com/apache/tvm/pull/8382#issuecomment-905781697 Thanks @mikepapadim ! This is now merged -- 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

[GitHub] [tvm] guberti commented on a change in pull request #8748: [microTVM][RVM] Add Arduino RVM

guberti commented on a change in pull request #8748: URL: https://github.com/apache/tvm/pull/8748#discussion_r695902185 ## File path: apps/microtvm/reference-vm/base-box-tool.py ## @@ -439,26 +473,27 @@ def parse_args(): "--provider", choices=ALL_PROVIDERS,

[GitHub] [tvm] guberti commented on a change in pull request #8748: [microTVM][RVM] Add Arduino RVM

guberti commented on a change in pull request #8748: URL: https://github.com/apache/tvm/pull/8748#discussion_r695902185 ## File path: apps/microtvm/reference-vm/base-box-tool.py ## @@ -439,26 +473,27 @@ def parse_args(): "--provider", choices=ALL_PROVIDERS,

[GitHub] [tvm] guberti commented on a change in pull request #8748: [microTVM][RVM] Add Arduino RVM

guberti commented on a change in pull request #8748: URL: https://github.com/apache/tvm/pull/8748#discussion_r696031380 ## File path: apps/microtvm/reference-vm/arduino/Vagrantfile ## @@ -0,0 +1,66 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more cont

[GitHub] [tvm] guberti commented on pull request #8748: [microTVM][RVM] Add Arduino RVM

guberti commented on pull request #8748: URL: https://github.com/apache/tvm/pull/8748#issuecomment-905798037 Thanks for the second round of comments Mehrdad! They should be addressed by 32882c4. -- This is an automated message from the Apache Git Service. To respond to the message, pleas

[GitHub] [tvm] mbs-octoml commented on pull request #8835: Change target string to Target object in the TE compiler and interpreter

mbs-octoml commented on pull request #8835: URL: https://github.com/apache/tvm/pull/8835#issuecomment-905810621 I'm fine with 1, and in the future: 4. Ensure all Targets are created at the top level and shared throughout the code rather than created on-the-fly for various 'default' o

[GitHub] [tvm] AndrewZhaoLuo opened a new pull request #8850: [AMP] Bump up tolerance on flaky test

AndrewZhaoLuo opened a new pull request #8850: URL: https://github.com/apache/tvm/pull/8850 https://ci.tlcpack.ai/blue/organizations/jenkins/tvm/detail/PR-8810/3/pipeline <-- flakiness here -- This is an automated message from the Apache Git Service. To respond to the message, p

[GitHub] [tvm] AndrewZhaoLuo commented on pull request #8850: [AMP] Bump up tolerance on flaky test

AndrewZhaoLuo commented on pull request #8850: URL: https://github.com/apache/tvm/pull/8850#issuecomment-905827889 @masahi -- 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.

[GitHub] [tvm] csullivan commented on a change in pull request #8823: [Hexagon] Rework tvm.target.hexagon() interface

csullivan commented on a change in pull request #8823: URL: https://github.com/apache/tvm/pull/8823#discussion_r696091607 ## File path: python/tvm/target/target.py ## @@ -413,79 +413,118 @@ def bifrost(model="unknown", options=None): return Target(" ".join(["opencl"] + opt

[GitHub] [tvm] tmoreau89 merged pull request #8841: Update CI Lint Image Version

tmoreau89 merged pull request #8841: URL: https://github.com/apache/tvm/pull/8841 -- 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: commits-unsubscr

[GitHub] [tvm] tqchen commented on pull request #8847: [Pass] Make Module Pass always return a new IRModule

tqchen commented on pull request #8847: URL: https://github.com/apache/tvm/pull/8847#issuecomment-905909951 I think it was due to the fact that we are designing the Module to look like a mutable interface while in nature we want something immutable. As a result some of the passes are writt

[GitHub] [tvm] tqchen edited a comment on pull request #8847: [Pass] Make Module Pass always return a new IRModule

tqchen edited a comment on pull request #8847: URL: https://github.com/apache/tvm/pull/8847#issuecomment-905909951 I think it was due to the fact that we are designing the Module to look like a mutable interface while in nature we want something immutable. As a result some of the passes ar

[GitHub] [tvm] trevor-m commented on a change in pull request #8808: [BYOC][TensorRT] Add TensorRT own int8 calibration support to TensorRT BYOC integration

trevor-m commented on a change in pull request #8808: URL: https://github.com/apache/tvm/pull/8808#discussion_r695107832 ## File path: src/runtime/contrib/tensorrt/tensorrt_builder.cc ## @@ -238,14 +253,15 @@ void TensorRTBuilder::CleanUp() { #endif builder_->destroy();

[GitHub] [tvm] junrushao1994 merged pull request #8778: [BUG] ToBasicBlockNormalForm immutability

junrushao1994 merged pull request #8778: URL: https://github.com/apache/tvm/pull/8778 -- 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: commits-unsu

[GitHub] [tvm] junrushao1994 commented on pull request #8778: [BUG] ToBasicBlockNormalForm immutability

junrushao1994 commented on pull request #8778: URL: https://github.com/apache/tvm/pull/8778#issuecomment-905920923 Finally the CI is green! -- 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 sp

  1   2   3   4   5   6   7   8   9   10   >