[GitHub] [incubator-tvm] jiangzoi opened a new pull request #6134: [Relay]Port eliminate_common_subexpr to non-recursive form

2020-07-23 Thread GitBox
jiangzoi opened a new pull request #6134: URL: https://github.com/apache/incubator-tvm/pull/6134 @zhiics Hi Zhi, I ported the pass of eliminating common subexpr to non-recursive form by replacing ExprMutator to MixedModeMutator. Thank you for your time to review that.

[GitHub] [incubator-tvm] ANSHUMAN87 commented on a change in pull request #6066: [TIR][Transform] HoistIfThenElse added

2020-07-23 Thread GitBox
ANSHUMAN87 commented on a change in pull request #6066: URL: https://github.com/apache/incubator-tvm/pull/6066#discussion_r459859824 ## File path: src/tir/transforms/hoist_if_then_else.cc ## @@ -0,0 +1,376 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one +

[GitHub] [incubator-tvm] ANSHUMAN87 commented on a change in pull request #6066: [TIR][Transform] HoistIfThenElse added

2020-07-23 Thread GitBox
ANSHUMAN87 commented on a change in pull request #6066: URL: https://github.com/apache/incubator-tvm/pull/6066#discussion_r459853815 ## File path: src/tir/transforms/hoist_if_then_else.cc ## @@ -0,0 +1,376 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one +

[GitHub] [incubator-tvm] roastduck commented on a change in pull request #6066: [TIR][Transform] HoistIfThenElse added

2020-07-23 Thread GitBox
roastduck commented on a change in pull request #6066: URL: https://github.com/apache/incubator-tvm/pull/6066#discussion_r459839683 ## File path: src/tir/transforms/hoist_if_then_else.cc ## @@ -0,0 +1,376 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one +

[GitHub] [incubator-tvm] roastduck commented on a change in pull request #6066: [TIR][Transform] HoistIfThenElse added

2020-07-23 Thread GitBox
roastduck commented on a change in pull request #6066: URL: https://github.com/apache/incubator-tvm/pull/6066#discussion_r459838669 ## File path: src/tir/transforms/hoist_if_then_else.cc ## @@ -0,0 +1,376 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one +

[GitHub] [incubator-tvm] jcf94 commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
jcf94 commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459834617 ## File path: src/auto_scheduler/transform_step.h ## @@ -194,20 +201,25 @@ void StepApplyToState(const Step& step, State* state, const ComputeDAG&

[GitHub] [incubator-tvm] anijain2305 commented on pull request #6121: support int4/int8 conv2d tensor core with HWNC layout

2020-07-23 Thread GitBox
anijain2305 commented on pull request #6121: URL: https://github.com/apache/incubator-tvm/pull/6121#issuecomment-663325799 Thanks for the contribution. I was wondering if this has been integrated with Relay so that we can use it for quantized networks.

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459828228 ## File path: src/auto_scheduler/transform_step.h ## @@ -194,20 +201,25 @@ void StepApplyToState(const Step& step, State* state, const

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459825492 ## File path: python/tvm/auto_scheduler/compute_dag.py ## @@ -126,11 +126,16 @@ def infer_bound_from_state(self, state): Returns

[GitHub] [incubator-tvm] merrymercy edited a comment on pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy edited a comment on pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#issuecomment-663320730 I opened an issue to track the unresolved todo items during upstream. #6133 However, all todos in this PR are not related to Ansor and are outside the scope

[GitHub] [incubator-tvm] merrymercy edited a comment on pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy edited a comment on pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#issuecomment-663320730 I opened an issue to track the unresolved todo items during upstream. #6133 However, all todos in this PR are not related to Ansor and are outside the scope

[GitHub] [incubator-tvm] merrymercy commented on pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#issuecomment-663320730 I opened an issue to track the unresolved todo items during upstream. However, all todos in this PR are not related to Ansor and are outside the scope of Ansor.

[GitHub] [incubator-tvm] jcf94 commented on pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
jcf94 commented on pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#issuecomment-663319503 @merrymercy @junrushao1994 @jroesch @comaniac Comments are addressed, please take another look. I'll submit the PRs for the rest steps soon after this.

[GitHub] [incubator-tvm] merrymercy opened a new issue #6133: Ansor Stabalization Tracking

2020-07-23 Thread GitBox
merrymercy opened a new issue #6133: URL: https://github.com/apache/incubator-tvm/issues/6133 This issue is used to track all unresolved todo items in Ansor during upstream. ### Motivation and Plan Ansor has more than 13K lines of code in our internal codebase. To upstream the

[GitHub] [incubator-tvm] jcf94 commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
jcf94 commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459822242 ## File path: src/auto_scheduler/transform_step.h ## @@ -659,6 +671,153 @@ class ComputeRootStep : public Step {

[GitHub] [incubator-tvm] jcf94 commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
jcf94 commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459821958 ## File path: python/tvm/auto_scheduler/loop_state.py ## @@ -351,6 +351,72 @@ def compute_root(self, stage): self.state_object =

[GitHub] [incubator-tvm] jcf94 commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
jcf94 commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459821753 ## File path: src/auto_scheduler/transform_step.cc ## @@ -923,5 +958,272 @@ String ComputeRootStepNode::PrintAsPythonAPI(Array* stages, return

[GitHub] [incubator-tvm] leonwanghui edited a comment on pull request #5892: Add TVM application extension with WASM runtime

2020-07-23 Thread GitBox
leonwanghui edited a comment on pull request #5892: URL: https://github.com/apache/incubator-tvm/pull/5892#issuecomment-663316322 cc @tqchen @nhynes @jroesch, I think this PR is ready to go, please notice that some of `clippy` warnings in `tvm-graph-rt` module would not be addressed in

[GitHub] [incubator-tvm] leonwanghui edited a comment on pull request #5892: Add TVM application extension with WASM runtime

2020-07-23 Thread GitBox
leonwanghui edited a comment on pull request #5892: URL: https://github.com/apache/incubator-tvm/pull/5892#issuecomment-663316322 cc @tqchen @nhynes, I think this PR is ready to go, please notice that some of `clippy` warnings in `tvm-graph-rt` module would not be addressed in this PR.

[GitHub] [incubator-tvm] leonwanghui commented on pull request #5892: Add TVM application extension with WASM runtime

2020-07-23 Thread GitBox
leonwanghui commented on pull request #5892: URL: https://github.com/apache/incubator-tvm/pull/5892#issuecomment-663316322 cc @tqchen @nhynes , please notice that some of `clippy` warnings in `tvm-graph-rt` module would not be addressed in this PR.

[GitHub] [incubator-tvm] alexbooth commented on pull request #6118: Add 'get_num_inputs' to GraphRuntime

2020-07-23 Thread GitBox
alexbooth commented on pull request #6118: URL: https://github.com/apache/incubator-tvm/pull/6118#issuecomment-663309975 > The number contains both inputs and weights. Is that what you want? Yes, not just the input data but potential input weights as well. I want the length of

[GitHub] [incubator-tvm] alexbooth edited a comment on pull request #6118: Add 'get_num_inputs' to GraphRuntime

2020-07-23 Thread GitBox
alexbooth edited a comment on pull request #6118: URL: https://github.com/apache/incubator-tvm/pull/6118#issuecomment-663309975 > The number contains both inputs and weights. Is that what you want? Yes, not just the input data but potential input weights as well. I want the length

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459812514 ## File path: tests/python/unittest/test_auto_scheduler_loop_state.py ## @@ -143,6 +143,282 @@ def test_compute_at_root_inline(): assert

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459812514 ## File path: tests/python/unittest/test_auto_scheduler_loop_state.py ## @@ -143,6 +143,282 @@ def test_compute_at_root_inline(): assert

[GitHub] [incubator-tvm] comaniac opened a new pull request #6132: [TOPI] Fix CUDA Library Tuning

2020-07-23 Thread GitBox
comaniac opened a new pull request #6132: URL: https://github.com/apache/incubator-tvm/pull/6132 Although we make `conv2d_cudnn.cuda` and `dense_cublas.cuda` as AutoTVM tasks so that they can be "tuned" and compared with other implementations, there have some issues prevent us from

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459812208 ## File path: src/auto_scheduler/transform_step.h ## @@ -659,6 +671,153 @@ class ComputeRootStep : public Step {

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459812067 ## File path: src/auto_scheduler/transform_step.h ## @@ -659,6 +671,153 @@ class ComputeRootStep : public Step {

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459811861 ## File path: src/auto_scheduler/transform_step.h ## @@ -659,6 +671,153 @@ class ComputeRootStep : public Step {

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459811861 ## File path: src/auto_scheduler/transform_step.h ## @@ -659,6 +671,153 @@ class ComputeRootStep : public Step {

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459810569 ## File path: src/auto_scheduler/transform_step.cc ## @@ -923,5 +958,272 @@ String ComputeRootStepNode::PrintAsPythonAPI(Array* stages,

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459810059 ## File path: src/auto_scheduler/transform_step.cc ## @@ -923,5 +958,272 @@ String ComputeRootStepNode::PrintAsPythonAPI(Array* stages,

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459809485 ## File path: src/auto_scheduler/loop_state.h ## @@ -225,6 +238,13 @@ class StateNode : public Object { * operation. */ AttachMap

[GitHub] [incubator-tvm] jcf94 commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
jcf94 commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459809462 ## File path: src/auto_scheduler/loop_state.h ## @@ -195,6 +197,17 @@ class AttachMap : public ObjectRef { void UpdateIters(const std::vector&

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459809485 ## File path: src/auto_scheduler/loop_state.h ## @@ -225,6 +238,13 @@ class StateNode : public Object { * operation. */ AttachMap

[GitHub] [incubator-tvm] leonwanghui commented on a change in pull request #5892: Add TVM application extension with WASM runtime

2020-07-23 Thread GitBox
leonwanghui commented on a change in pull request #5892: URL: https://github.com/apache/incubator-tvm/pull/5892#discussion_r459809422 ## File path: apps/wasm-standalone/wasm-graph/src/utils.rs ## @@ -0,0 +1,48 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under

[GitHub] [incubator-tvm] jcf94 commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
jcf94 commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459809086 ## File path: tests/python/unittest/test_auto_scheduler_loop_state.py ## @@ -143,6 +143,282 @@ def test_compute_at_root_inline(): assert

[GitHub] [incubator-tvm] leonwanghui commented on a change in pull request #5892: Add TVM application extension with WASM runtime

2020-07-23 Thread GitBox
leonwanghui commented on a change in pull request #5892: URL: https://github.com/apache/incubator-tvm/pull/5892#discussion_r459808855 ## File path: apps/wasm-standalone/wasm-graph/src/utils.rs ## @@ -0,0 +1,48 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459808487 ## File path: python/tvm/auto_scheduler/loop_state.py ## @@ -351,6 +351,72 @@ def compute_root(self, stage): self.state_object =

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459808071 ## File path: python/tvm/auto_scheduler/loop_state.py ## @@ -351,6 +351,72 @@ def compute_root(self, stage): self.state_object =

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459808071 ## File path: python/tvm/auto_scheduler/loop_state.py ## @@ -351,6 +351,72 @@ def compute_root(self, stage): self.state_object =

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459808071 ## File path: python/tvm/auto_scheduler/loop_state.py ## @@ -351,6 +351,72 @@ def compute_root(self, stage): self.state_object =

[GitHub] [incubator-tvm] zhiics commented on pull request #6118: Add 'get_num_inputs' to GraphRuntime

2020-07-23 Thread GitBox
zhiics commented on pull request #6118: URL: https://github.com/apache/incubator-tvm/pull/6118#issuecomment-663303284 The number contains both inputs and weights. Is that what you want? This is an automated message from the

[GitHub] [incubator-tvm] merrymercy commented on pull request #6103: [Ansor][AutoTVM v2.0] Phase 1: Access Analyzer

2020-07-23 Thread GitBox
merrymercy commented on pull request #6103: URL: https://github.com/apache/incubator-tvm/pull/6103#issuecomment-663276179 @junrushao1994 @jcf94 @comaniac Most of the comments are addressed. I added more doc and make the name convention more consistent and meaningful. Please take another

[GitHub] [incubator-tvm] comaniac commented on a change in pull request #6131: [FIX] Fixes #6096

2020-07-23 Thread GitBox
comaniac commented on a change in pull request #6131: URL: https://github.com/apache/incubator-tvm/pull/6131#discussion_r459777668 ## File path: python/tvm/autotvm/task/relay_integration.py ## @@ -140,6 +140,11 @@ def extract_from_multiple_program(mods, params, target,

[GitHub] [incubator-tvm] jroesch commented on a change in pull request #6131: [FIX] Fixes #6096

2020-07-23 Thread GitBox
jroesch commented on a change in pull request #6131: URL: https://github.com/apache/incubator-tvm/pull/6131#discussion_r459773263 ## File path: src/relay/backend/build_module.cc ## @@ -225,6 +226,8 @@ class RelayBuildModule : public runtime::ModuleNode { targets_ =

[GitHub] [incubator-tvm] tkonolige opened a new pull request #6131: [FIX] Fixes #6096

2020-07-23 Thread GitBox
tkonolige opened a new pull request #6131: URL: https://github.com/apache/incubator-tvm/pull/6131 Clear the compile cache between module builds so that schedule changes will have an effect. Also, clear the warning cache so that schedule changes properly list untuned ops. Note: this

[GitHub] [incubator-tvm] giuseros commented on pull request #6095: Improve NHWC depthwise convolution for AArch64

2020-07-23 Thread GitBox
giuseros commented on pull request #6095: URL: https://github.com/apache/incubator-tvm/pull/6095#issuecomment-663266858 @FrozenGene , @anijain2305 , This is my last commit before holidays. I enabled only the `arm_cpu` tests (everything passes locally). The case with `dilation>1` will be

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6103: [Ansor][AutoTVM v2.0] Phase 1: Access Analyzer

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6103: URL: https://github.com/apache/incubator-tvm/pull/6103#discussion_r459750828 ## File path: src/auto_scheduler/compute_dag.cc ## @@ -114,7 +118,432 @@ Array TopoSortOps(const Array& tensors) { return ops; } -//

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6103: [Ansor][AutoTVM v2.0] Phase 1: Access Analyzer

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6103: URL: https://github.com/apache/incubator-tvm/pull/6103#discussion_r459749921 ## File path: src/auto_scheduler/compute_dag.cc ## @@ -114,7 +118,432 @@ Array TopoSortOps(const Array& tensors) { return ops; } -//

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6103: [Ansor][AutoTVM v2.0] Phase 1: Access Analyzer

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6103: URL: https://github.com/apache/incubator-tvm/pull/6103#discussion_r459749921 ## File path: src/auto_scheduler/compute_dag.cc ## @@ -114,7 +118,432 @@ Array TopoSortOps(const Array& tensors) { return ops; } -//

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6103: [Ansor][AutoTVM v2.0] Phase 1: Access Analyzer

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6103: URL: https://github.com/apache/incubator-tvm/pull/6103#discussion_r459749490 ## File path: src/auto_scheduler/compute_dag.cc ## @@ -114,7 +118,432 @@ Array TopoSortOps(const Array& tensors) { return ops; } -//

[GitHub] [incubator-tvm] ymwangg opened a new pull request #6130: [AutoTVM][BugFix] Fix autotvm on the conv2d_nchw_winograd.mali operator

2020-07-23 Thread GitBox
ymwangg opened a new pull request #6130: URL: https://github.com/apache/incubator-tvm/pull/6130 This PR fixes two bugs related to autotvm on the `conv2d_nchw_winograd.mali` operator. - KeyError when computing`kvshape` - Duplicate args returned by

[GitHub] [incubator-tvm] merrymercy commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663238657 To conclude, the current cache mechanism in relay works globally across different calls to `relay.build`. If we want to use two different schedules for the same relay

[GitHub] [incubator-tvm] jroesch commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
jroesch commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663234072 @tkonolige @jwfromm yeah the caching makes sense to me, this is why I was against it a long time ago ´ść . I think we should add to the C++ implementation of `build` to ensure

[GitHub] [incubator-tvm] giuseros commented on pull request #6095: Improve NHWC depthwise convolution for AArch64

2020-07-23 Thread GitBox
giuseros commented on pull request #6095: URL: https://github.com/apache/incubator-tvm/pull/6095#issuecomment-663233922 I probably enabled a CUDA test that was making the CI hang. I reverted the test hoping that this was the issue

[GitHub] [incubator-tvm] jwfromm commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
jwfromm commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663233439 We should probably try to upstream a PR that adds `compile_engine.get().clear()` to the end of the `relay.build` function.

[GitHub] [incubator-tvm] jwfromm commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
jwfromm commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663233076 ok that all makes sense to me, thanks for the clarification @merrymercy! This is an automated message from the

[GitHub] [incubator-tvm] merrymercy edited a comment on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy edited a comment on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663232365 @jroesch Yes, the behavior you mentioned is desired. That is designed by me to mimic the behavior of `logging.warning` in python's built-in library. The principle

[GitHub] [incubator-tvm] merrymercy commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663232365 @jroesch Yes, the behavior you mentioned is desired. That is designed by me to mimic the behavior of `logging.warning` in python's built-in library. The principle is that

[GitHub] [incubator-tvm] jroesch merged pull request #6116: [Rust] Some rust cleanups

2020-07-23 Thread GitBox
jroesch merged pull request #6116: URL: https://github.com/apache/incubator-tvm/pull/6116 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

[incubator-tvm] branch master updated (6dbc344 -> 0a1c4c2)

2020-07-23 Thread jroesch
This is an automated email from the ASF dual-hosted git repository. jroesch pushed a change to branch master in repository https://gitbox.apache.org/repos/asf/incubator-tvm.git. from 6dbc344 [RELAY][Fix] i64 indices (#5235) add 0a1c4c2 [Rust] Some rust cleanups (#6116) No new

[GitHub] [incubator-tvm] tkonolige commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
tkonolige commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663231546 @merrymercy When I tested `compile_engine.get().clear()` the first time I might have had a corrupted or incorrect tuning log. The second time I test

[incubator-tvm] branch master updated (fe76196 -> 6dbc344)

2020-07-23 Thread zhic
This is an automated email from the ASF dual-hosted git repository. zhic pushed a change to branch master in repository https://gitbox.apache.org/repos/asf/incubator-tvm.git. from fe76196 Register Shape Func for Some Operators to Handle Dynamic Shapes (#5955) add 6dbc344

[GitHub] [incubator-tvm] jroesch commented on pull request #6116: [Rust] Some rust cleanups

2020-07-23 Thread GitBox
jroesch commented on pull request #6116: URL: https://github.com/apache/incubator-tvm/pull/6116#issuecomment-663231660 Thanks Jason! This is an automated message from the Apache Git Service. To respond to the message, please

[GitHub] [incubator-tvm] zhiics merged pull request #5235: [RELAY][Fix] i64 indices

2020-07-23 Thread GitBox
zhiics merged pull request #5235: URL: https://github.com/apache/incubator-tvm/pull/5235 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

[GitHub] [incubator-tvm] zhiics commented on pull request #5235: [RELAY][Fix] i64 indices

2020-07-23 Thread GitBox
zhiics commented on pull request #5235: URL: https://github.com/apache/incubator-tvm/pull/5235#issuecomment-663231777 Thanks @hzfan @jwfromm @tqchen @kazum This is an automated message from the Apache Git Service. To

[GitHub] [incubator-tvm] jroesch commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
jroesch commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459719321 ## File path: src/auto_scheduler/loop_state.h ## @@ -358,19 +369,43 @@ class State : public ObjectRef { void compute_at(int stage_id, int

[GitHub] [incubator-tvm] jwfromm commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
jwfromm commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663229748 @merrymercy, it seems like the keys are still being saved in the DispatchContext memory even after clearing the cache. An easy to way to see this is doing build twice on the

[GitHub] [incubator-tvm] merrymercy edited a comment on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy edited a comment on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663229677 What do you mean by "remove the tuning log"? If you add `compile_engine.get().clear()`, then the result of `python3 example.py extract` will match the result of

[GitHub] [incubator-tvm] merrymercy commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663229677 What do you mean by `remove the tuning log`? If you add `compile_engine.get().clear()`, then the result of `python3 example.py extract` will match the result of `python3

[GitHub] [incubator-tvm] tkonolige commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
tkonolige commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663223033 Now it is working with `compile_engine.get().clear()`, but I had to remove the tuning log first. This is an

[GitHub] [incubator-tvm] merrymercy edited a comment on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy edited a comment on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663221391 To prevent such issues from happening again, we can force relay to clear the cache after every `relay.build`. So the cache mechanism in relay only works within a

[GitHub] [incubator-tvm] merrymercy edited a comment on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy edited a comment on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663221391 To prevent such issues from happening again, we can force relay to clear the cache after every `relay.build`

[GitHub] [incubator-tvm] merrymercy edited a comment on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy edited a comment on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663221391 To prevent such issues happening again, we can force relay to clear the cache after every `relay.build`

[GitHub] [incubator-tvm] merrymercy commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663221391 To prevent such bugs happening again, we can force relay to clear the cache after every `relay.build` This

[GitHub] [incubator-tvm] merrymercy edited a comment on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy edited a comment on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663218666 I ran the code and found adding `compile_engine.get().clear()` before `with autotvm.apply_history_best(log_filename):` acutally fixes the issue. This issue is not

[GitHub] [incubator-tvm] merrymercy edited a comment on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy edited a comment on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663218666 I ran the code and found adding `compile_engine.get().clear()` before `with autotvm.apply_history_best(log_filename):` acutally fixes the issue. This issue is not

[GitHub] [incubator-tvm] merrymercy commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
merrymercy commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663218666 I ran the code and found adding `compile_engine.get().clear()` before `with autotvm.apply_history_best(log_filename):` acutally fixes the issue. This issue is not

[GitHub] [incubator-tvm] icemelon9 merged pull request #5955: Register Shape Func for Some Operators to Handle Dynamic Shapes

2020-07-23 Thread GitBox
icemelon9 merged pull request #5955: URL: https://github.com/apache/incubator-tvm/pull/5955 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] [incubator-tvm] icemelon9 commented on pull request #5955: Register Shape Func for Some Operators to Handle Dynamic Shapes

2020-07-23 Thread GitBox
icemelon9 commented on pull request #5955: URL: https://github.com/apache/incubator-tvm/pull/5955#issuecomment-663214236 Thanks @lsy643 This is an automated message from the Apache Git Service. To respond to the message,

[incubator-tvm] branch master updated (9d34eaa -> fe76196)

2020-07-23 Thread haichen
This is an automated email from the ASF dual-hosted git repository. haichen pushed a change to branch master in repository https://gitbox.apache.org/repos/asf/incubator-tvm.git. from 9d34eaa Improve reduction schedule on arm CPUs (#6110) add fe76196 Register Shape Func for Some

[GitHub] [incubator-tvm] jwfromm commented on issue #6096: apply_history_best requires extract_from_program to be run even if log file already exists

2020-07-23 Thread GitBox
jwfromm commented on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663203960 I did a few tests with the script you provided and I'm fairly certain this has nothing to do with `apply_history_best`. I think this is instead an issue of whether tophub

[GitHub] [incubator-tvm] merrymercy commented on a change in pull request #6103: [Ansor][AutoTVM v2.0] Phase 1: Access Analyzer

2020-07-23 Thread GitBox
merrymercy commented on a change in pull request #6103: URL: https://github.com/apache/incubator-tvm/pull/6103#discussion_r459691514 ## File path: src/auto_scheduler/compute_dag.cc ## @@ -126,6 +555,7 @@ class FlopEstimator : public ExprFunctor { fail_ = true;

[GitHub] [incubator-tvm] jwfromm commented on pull request #5235: [RELAY][Fix] i64 indices

2020-07-23 Thread GitBox
jwfromm commented on pull request #5235: URL: https://github.com/apache/incubator-tvm/pull/5235#issuecomment-663174713 LGTM This is an automated message from the Apache Git Service. To respond to the message, please log on

[GitHub] [incubator-tvm] junrushao1994 commented on a change in pull request #6107: [Ansor][AutoTVM v2.0] Phase 1: Add cache_read/cache_write steps

2020-07-23 Thread GitBox
junrushao1994 commented on a change in pull request #6107: URL: https://github.com/apache/incubator-tvm/pull/6107#discussion_r459194418 ## File path: tests/python/unittest/test_auto_scheduler_loop_state.py ## @@ -143,6 +143,282 @@ def test_compute_at_root_inline(): assert

[GitHub] [incubator-tvm] junrushao1994 commented on pull request #6118: Add 'get_num_inputs' to GraphRuntime

2020-07-23 Thread GitBox
junrushao1994 commented on pull request #6118: URL: https://github.com/apache/incubator-tvm/pull/6118#issuecomment-663159381 just curious, what is it used for? This is an automated message from the Apache Git Service. To

[GitHub] [incubator-tvm] trevor-m commented on pull request #6108: Fix CUDA Compute Function For `get_valid_counts` and `nms`

2020-07-23 Thread GitBox
trevor-m commented on pull request #6108: URL: https://github.com/apache/incubator-tvm/pull/6108#issuecomment-663152986 IIRC, data arrangement was removed from get_valid_counts to improve performance because the data arrangement would be done by NMS anyway. Does this PR maintain the

[GitHub] [incubator-tvm] lhutton1 commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
lhutton1 commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459626275 ## File path: tests/python/contrib/test_arm_compute_lib/test_conv2d.py ## @@ -127,51 +241,31 @@ def test_conv2d(): device = Device()

[GitHub] [incubator-tvm] lhutton1 commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
lhutton1 commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459624243 ## File path: src/runtime/contrib/arm_compute_lib/acl_utils.cc ## @@ -108,6 +111,30 @@ arm_compute::PadStrideInfo ToACLPadStride(const

[GitHub] [incubator-tvm] zhiics commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
zhiics commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459618012 ## File path: tests/python/contrib/test_arm_compute_lib/test_conv2d.py ## @@ -127,51 +241,31 @@ def test_conv2d(): device = Device()

[GitHub] [incubator-tvm] comaniac commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
comaniac commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459616949 ## File path: tests/python/contrib/test_arm_compute_lib/test_conv2d.py ## @@ -127,51 +241,31 @@ def test_conv2d(): device = Device()

[GitHub] [incubator-tvm] csullivan commented on a change in pull request #6085: [Relay][OP] Support NMSv4

2020-07-23 Thread GitBox
csullivan commented on a change in pull request #6085: URL: https://github.com/apache/incubator-tvm/pull/6085#discussion_r459608948 ## File path: python/tvm/relay/frontend/tensorflow.py ## @@ -637,10 +637,11 @@ def _impl(inputs, attr, params, mod): iou_threshold =

[GitHub] [incubator-tvm] lhutton1 commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
lhutton1 commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459608769 ## File path: tests/python/contrib/test_arm_compute_lib/test_conv2d.py ## @@ -127,51 +241,31 @@ def test_conv2d(): device = Device()

[GitHub] [incubator-tvm] comaniac commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
comaniac commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459606833 ## File path: tests/python/contrib/test_arm_compute_lib/test_conv2d.py ## @@ -127,51 +241,31 @@ def test_conv2d(): device = Device()

[GitHub] [incubator-tvm] anijain2305 commented on pull request #6115: [Topi, x86] Using MKL blas for quantized dense

2020-07-23 Thread GitBox
anijain2305 commented on pull request #6115: URL: https://github.com/apache/incubator-tvm/pull/6115#issuecomment-663128946 @TaoLv Good point, I added the latency numbers for TVM alone. Thanks for pointing it out! This is an

[GitHub] [incubator-tvm] lhutton1 commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
lhutton1 commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459604438 ## File path: src/runtime/contrib/arm_compute_lib/acl_utils.cc ## @@ -108,6 +111,30 @@ arm_compute::PadStrideInfo ToACLPadStride(const

[GitHub] [incubator-tvm] comaniac commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
comaniac commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459604117 ## File path: src/relay/backend/contrib/arm_compute_lib/codegen.cc ## @@ -78,57 +90,83 @@ class ACLJSONSerializer : public

[GitHub] [incubator-tvm] lhutton1 commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
lhutton1 commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459603811 ## File path: tests/python/contrib/test_arm_compute_lib/test_conv2d.py ## @@ -127,51 +241,31 @@ def test_conv2d(): device = Device()

[GitHub] [incubator-tvm] lhutton1 commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
lhutton1 commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459578673 ## File path: src/relay/backend/contrib/arm_compute_lib/codegen.cc ## @@ -78,57 +90,83 @@ class ACLJSONSerializer : public

[GitHub] [incubator-tvm] lhutton1 commented on a change in pull request #6109: [BYOC][ACL] Support asymmetric per-layer quantized operators

2020-07-23 Thread GitBox
lhutton1 commented on a change in pull request #6109: URL: https://github.com/apache/incubator-tvm/pull/6109#discussion_r459571834 ## File path: src/runtime/contrib/arm_compute_lib/acl_runtime.cc ## @@ -163,24 +149,61 @@ class ACLRuntime : public JSONRuntimeBase { struct

  1   2   >