platform/upstream/tvm.git
4 years ago[ARITH][BACKPORT-0.6] fix a min/max simplify bug (#5749)
xqdan [Tue, 9 Jun 2020 15:34:27 +0000 (23:34 +0800)]
[ARITH][BACKPORT-0.6] fix a min/max simplify bug (#5749)

* fix a min/max simplify bug

* fix cpplint

* turn into oposite when c1val<0 and add more case

* fix c1=0

Co-authored-by: xqdan <danxiaoqiang@huawei.com>
4 years agoDon't add cast for TF batch norm when type isn't changing (#5731)
Trevor Morris [Mon, 8 Jun 2020 23:43:28 +0000 (16:43 -0700)]
Don't add cast for TF batch norm when type isn't changing (#5731)

4 years ago[REFACTOR][TE][TIR] Call::Halide => ProducerLoad, DSL/TIR decouple. (#5743)
Tianqi Chen [Sun, 7 Jun 2020 21:11:05 +0000 (14:11 -0700)]
[REFACTOR][TE][TIR] Call::Halide => ProducerLoad, DSL/TIR decouple. (#5743)

In the HalideIR's design, DSL components and IR are mixed together.
For example, Call::Halide can containa reference to a function which is
constructed in the tensor expression language.

While this coupled design simplifies certain aspect of the DSL construction,
it prevents the TIR to evolve as a clean standalone IR:

- The additional tensor expression provided in the function is opaque to the IR
  and may become obsolete as we transform them.
- The duplication of the information in the DSL tensor and IR makes it hard to
  design a stand-alone text format (when there are elements shared in the tensor
  expression and normal statements).

This PR aims to clearly de-couple the TIR from high-level DSL structures(tensor expression),
while still provide clear extensions to build DSLs on top of the TIR.

We introduce a DataProducer as a base class for high level tensor expressions objects
that produce data. We then introduce ProducerLoad to replace the Call::Halide usage,
so that the Call node can always be self contained and used for low-level calls.

The high-level tensor expression DSL can still generate a PrimExpr that contains a ProducerLoad.
These PrimExprs contains fragments of information that can be combined together to
generate a low-level TIR PrimFunc.

We also state clearly that DataProducer **should not** appear in any TIR PrimFunc.
Instead, the high-level DSL layer should lowered DataProducers to Buffers and TIR statements
that produces these buffers. We can further provide verifications to validate such invariance.

Changes:
- Introduce DataProducer to serve as a base class for Tensor in tensor expressions.
- Migrate use of Call::Halide to ProducerLoad
- Migrate the other usages of Calls.

We will also create follow-up PRs to migrate the remaining two DSL related IR nodes(Realize/Provide)
to use the DataProducer.

4 years agosequential cpp test (#5745)
Zhi [Sun, 7 Jun 2020 19:41:23 +0000 (12:41 -0700)]
sequential cpp test (#5745)

4 years agoAdd some docs on downstream consistency (#5742)
Junru Shao [Sat, 6 Jun 2020 22:23:20 +0000 (15:23 -0700)]
Add some docs on downstream consistency (#5742)

https://github.com/apache/incubator-tvm/pull/5730#issuecomment-639567636

4 years ago[REFACTOR][ARITH] Remove legacy compute_expr.h (#5738)
Tianqi Chen [Sat, 6 Jun 2020 20:23:31 +0000 (13:23 -0700)]
[REFACTOR][ARITH] Remove legacy compute_expr.h (#5738)

Replaces most of the ComptuteReduce using foldl.

4 years agofix small bug about dense_grad (#5695)
handar423 [Sat, 6 Jun 2020 03:40:17 +0000 (11:40 +0800)]
fix small bug about dense_grad (#5695)

4 years agoFix the values for test_fmod since it fails way too often otherwise (#5723)
abergeron [Fri, 5 Jun 2020 21:17:41 +0000 (17:17 -0400)]
Fix the values for test_fmod since it fails way too often otherwise (#5723)

4 years ago[TEST] Fix flaky topi/tests/python/test_topi_pooling.py:test_adaptive_pool (#5736)
Tianqi Chen [Fri, 5 Jun 2020 19:13:17 +0000 (12:13 -0700)]
[TEST] Fix flaky topi/tests/python/test_topi_pooling.py:test_adaptive_pool (#5736)

4 years agoFix reshape usage in ARM Winograd (#5732)
Cody Yu [Fri, 5 Jun 2020 17:27:57 +0000 (10:27 -0700)]
Fix reshape usage in ARM Winograd (#5732)

4 years agoChange 'delete's in Relay VM Instruction dtor to 'delete[]'s (#5735)
akosik-anyvision [Fri, 5 Jun 2020 15:39:20 +0000 (11:39 -0400)]
Change 'delete's in Relay VM Instruction dtor to 'delete[]'s (#5735)

4 years agoROCm: Add warp shuffles and enable reductions (#5727)
Thomas Viehmann [Fri, 5 Jun 2020 09:49:37 +0000 (11:49 +0200)]
ROCm: Add warp shuffles and enable reductions (#5727)

Thank you @masahi and @wpan11nv for the feedback

4 years ago[ONNX]MaxRoiPool, Mod & Xor op support added (#5729)
Samuel [Fri, 5 Jun 2020 05:48:44 +0000 (11:18 +0530)]
[ONNX]MaxRoiPool, Mod & Xor op support added (#5729)

4 years ago[REFACTOR] Separate ArgTypeCode from DLDataTypeCode (#5730)
Tianqi Chen [Thu, 4 Jun 2020 22:04:17 +0000 (15:04 -0700)]
[REFACTOR] Separate ArgTypeCode from DLDataTypeCode (#5730)

We use a single enum(TypeCode) to represent ArgTypeCode and DLDataTypeCode.
However, as we start to expand more data types, it is clear that argument
type code(in the FFI convention) and data type code needs to evolve separately.
So that we can add first class for data types without having changing the FFI ABI.

This PR makes the distinction clear and refactored the code to separate the two.

- [PY] Separate ArgTypeCode from DataTypeCode
- [WEB] Separate ArgTypeCode from DataTypeCode
- [JAVA] Separate ArgTypeCode from DataTypeCode

4 years ago[Frontend][TFLite] Add parser support for shape and range (#5329)
Dhruva Ray [Thu, 4 Jun 2020 17:55:38 +0000 (23:25 +0530)]
[Frontend][TFLite] Add parser support for shape and range (#5329)

* [Relay][Frontend][TFLite] Add parser support for shape and range

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* Incorporated review comments and used new functions

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* Few cosmetic changes

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* Removed an extra line added by rebase...

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
4 years ago[TOPI,RELAY][TFLITE] Sparse to dense operator (#5447)
Dhruva Ray [Thu, 4 Jun 2020 17:54:56 +0000 (23:24 +0530)]
[TOPI,RELAY][TFLITE] Sparse to dense operator (#5447)

* [Relay][Frontend][TFLite] Add parser support for shape and range

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* [TOPI,RELAY][TFLITE] Sparse to dense operator

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* use param name in documentation

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* sphinx doc errors fixed

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* incorporated review comments

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* Missing a blank line...

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* use get_tensor_expr

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* Accidently removed this function in the rebase...

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* support default value for default_value

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* clang format fixes

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
* topi pylint fixes

Signed-off-by: Dhruva Ray <dhruvaray@gmail.com>
4 years agocodegen llvm: move nvptx-specific intrinsic handling into codegen_nvptx (#5726)
Thomas Viehmann [Thu, 4 Jun 2020 15:46:55 +0000 (17:46 +0200)]
codegen llvm: move nvptx-specific intrinsic handling into codegen_nvptx (#5726)

See discussion in #5600.

I'm also throwing in a pointer lifetime fix for the context held by
NVPTX because otherwise topi/tests/python/test_topi_softmax.py
would sefault for me. With the test, I can also run resnet-18 on
the nvptx target in gpu_imagenet_bench.py.

4 years agoFix runtime::String backward compatibility in JSON (#5725)
Junru Shao [Thu, 4 Jun 2020 15:29:03 +0000 (08:29 -0700)]
Fix runtime::String backward compatibility in JSON (#5725)

4 years ago[AutoTVM, Relay] Clear compile engine after task extraction (#5724)
Wuwei Lin [Thu, 4 Jun 2020 06:47:04 +0000 (02:47 -0400)]
[AutoTVM, Relay] Clear compile engine after task extraction (#5724)

4 years ago[TENSORFLOW]StatefulPartitionedCall/PartitionedCall Ops support added (#5617)
Deepak [Thu, 4 Jun 2020 04:33:42 +0000 (10:03 +0530)]
[TENSORFLOW]StatefulPartitionedCall/PartitionedCall Ops support added  (#5617)

* Implemented functionInvocation Unit Test for StatefulPartitionedCall operator(working) and initial changes for placeholder(not working as of now)

* Placeholder exercises with tvm

* placeholder interim

* SPOP Test cases structure

* New test cases for spop

* miscellaneous test cases for spop

* Placeholder samples..working with shapes explicitly passed

* Variables test case. Works with the same fix of shape_dict

* SPOP Positive test cases first iteration

* support output tensors as function args, multiple functions

* Corrected Indentation

* filewritter is only for debug purpose

* support variables in function args

* First working iteration of positive spop test cases

* Removed commented code, simplified code

* Code Reorganization- First working iteration of positive spop test cases

* corrected variable name after refactor

* Code Reorganization- First working iteration of positive spop test cases

* move code inside mapped operator function

* Removed extra line

* support variables in function args

* Removed commented code, simplified code

* move code inside mapped operator function

* Code Reorganization- First working iteration of positive spop test cases

# Conflicts:
# tests/python/frontend/tensorflow/test_forward.py

* Code Reorganization- First working iteration of positive spop test cases

* Function invocation more test cases

* Simplified & Merged different Function Invocation Test cases

* support invocation of nested callables

no need to explicitly handle paratitioned and
statefulPartitioned condition in convert_operator function

* Simplified and Uniform testcases

* support invocation of nested callables

no need to explicitly handle paratitioned and
statefulPartitioned condition in convert_operator function

* Simplified and Uniform testcases

* removed duplicate and renamed testcase

* Negative scenario added for testing operator statefulness. Only Exception to stateful operators are Partitioned & StatefulPartitionedOp which have capability to execute even stateless operators within them

* Miscellaneous reorganization changes for spop scenarios

* Miscellaneous reorganization changes for spop scenarios

* Corrected import of tensorflow modules safely using try except and other code reorganization

* Negative scenario for resource variables handled

* Documentation update for code

* SPOP change in function handling

* handle nested subgraph

* refactor

* get op def compatible with tf 1x & 2x

* Fixed liniting issues

* added doctsring and few nits

* Merged changes for positive test cases and negative test cases

* Moved StatefulPartitionedCall test case to the end of the TC list

* Fixed some typos and semantics

* dmlc-core

* dmlc-core

* fixes

* Addressing Review comments in the PR for SPOP support

* Fixed pylint errors

* Corrected tensorflow import syntax

* Placed the op_def_registry module import outside of for loop

* Removed new stateful operators list and combined these operators with missing operators to display as single list. Also removed throwing seperate exception for stateful ops

Co-authored-by: Prashant Sail <psail4444@gmail.com>
Co-authored-by: maheshambule <mahesh_ambule@persistent.com>
4 years ago[ONNX]ReduceL1, ReduceL2, ReduceSumSquare, ReduceLogSum ops added (#5721)
Samuel [Thu, 4 Jun 2020 02:06:21 +0000 (07:36 +0530)]
[ONNX]ReduceL1, ReduceL2, ReduceSumSquare, ReduceLogSum ops added (#5721)

4 years agoFix generating types like float44 and float88 (#5722)
abergeron [Wed, 3 Jun 2020 21:24:27 +0000 (17:24 -0400)]
Fix generating types like float44 and float88 (#5722)

4 years ago[Object] Restore the StrMap behavior in JSON/SHash/SEqual (#5719)
Junru Shao [Wed, 3 Jun 2020 20:34:44 +0000 (13:34 -0700)]
[Object] Restore the StrMap behavior in JSON/SHash/SEqual (#5719)

4 years ago[Object][FFI] Introduce runtime::String::CanConvertFrom (#5718)
Junru Shao [Wed, 3 Jun 2020 15:01:55 +0000 (08:01 -0700)]
[Object][FFI] Introduce runtime::String::CanConvertFrom (#5718)

* [Object][FFI] Introduce runtime::String::CanConvertFrom

* Update container.h

4 years agoAvoid downloading when TOPHUB_LOCATION is NONE (#5720)
lixiaoquan [Wed, 3 Jun 2020 15:01:10 +0000 (23:01 +0800)]
Avoid downloading when TOPHUB_LOCATION is NONE (#5720)

4 years ago[MXNET]Softmin, trunc op support added (#5715)
Samuel [Wed, 3 Jun 2020 08:29:38 +0000 (13:59 +0530)]
[MXNET]Softmin, trunc op support added (#5715)

4 years ago[Object] Unify StrMapNode and MapNode (#5687)
Junru Shao [Tue, 2 Jun 2020 23:33:09 +0000 (16:33 -0700)]
[Object] Unify StrMapNode and MapNode (#5687)

* Pass cpptest and py unittest

* fix graph runtime

* right fix

* fix a bug that runtime::String's operator < is actually compare by address

* Update container.py

* Renaming

* Address comments

* lint

* Replace ObjectHash in object.py

4 years agoRename tvm_dso_op to libtvm_dso_op (#5714)
tobe [Tue, 2 Jun 2020 16:22:42 +0000 (00:22 +0800)]
Rename tvm_dso_op to libtvm_dso_op (#5714)

4 years ago[BUGFIX][CRT] Fix Compilation Error in CRT (#5713)
Liangfu Chen [Tue, 2 Jun 2020 16:22:14 +0000 (00:22 +0800)]
[BUGFIX][CRT] Fix Compilation Error in CRT (#5713)

4 years agoRemove opengl runtime and cmake (#5712)
Tianqi Chen [Tue, 2 Jun 2020 03:51:49 +0000 (20:51 -0700)]
Remove opengl runtime and cmake (#5712)

4 years agoRemove deprecated opengl files (#5711)
Tianqi Chen [Tue, 2 Jun 2020 00:53:33 +0000 (17:53 -0700)]
Remove deprecated opengl files (#5711)

4 years ago[PYTORCH]ReplicationPad support added (#5708)
Samuel [Tue, 2 Jun 2020 00:15:41 +0000 (05:45 +0530)]
[PYTORCH]ReplicationPad support added (#5708)

4 years ago[PatternLang] Simplify Pattern API Implementations (#5703)
Cody Yu [Tue, 2 Jun 2020 00:14:33 +0000 (17:14 -0700)]
[PatternLang] Simplify Pattern API Implementations (#5703)

* Add syntatic sugar; include pattern to API docs

* fix doc warnings

4 years ago[REFACTOR][PY] relay.op.Op -> tvm.ir.Op (#5705)
Tianqi Chen [Mon, 1 Jun 2020 22:35:06 +0000 (15:35 -0700)]
[REFACTOR][PY] relay.op.Op -> tvm.ir.Op (#5705)

* [REFACTOR][PY] relay.op.Op -> tvm.ir.Op

* Improve the error check

4 years agofix typo: anchor windoes should be anchor windows (#5706)
Rand Xie [Mon, 1 Jun 2020 15:39:36 +0000 (08:39 -0700)]
fix typo: anchor windoes should be anchor windows (#5706)

4 years ago[Arith] ExtendedEuclidean merge impl to int_operator (#5625)
ANSHUMAN TRIPATHY [Mon, 1 Jun 2020 15:38:26 +0000 (21:08 +0530)]
[Arith] ExtendedEuclidean merge impl to int_operator (#5625)

4 years ago[AutoTVM][TOPI] Fix bifrost spatial packing conv2d auto tune (#5684)
Neo Chien [Sun, 31 May 2020 19:32:25 +0000 (03:32 +0800)]
[AutoTVM][TOPI] Fix bifrost spatial packing conv2d auto tune (#5684)

* [AutoTVM][TOPI] Fix bifrost spatial packing conv2d auto tune

* [AutoTVM][TOPI] Putting placeholder replacement in compute

* Fix winograd kernel replacement

* Fix sanity check: Line too long

4 years ago[PYTORCH]floor_divide support for squeezenet (#5702)
Samuel [Sat, 30 May 2020 21:25:36 +0000 (02:55 +0530)]
[PYTORCH]floor_divide support for squeezenet (#5702)

https://github.com/apache/incubator-tvm/issues/5133#issuecomment-636330705

4 years ago[REFACTOR][RELAY] Replace build_config with PassContext (#5698)
Zhi [Sat, 30 May 2020 04:59:35 +0000 (21:59 -0700)]
[REFACTOR][RELAY] Replace build_config with PassContext (#5698)

4 years ago[BYOC] Support Tuple Output in C/DNNL Codegen (#5701)
Cody Yu [Sat, 30 May 2020 02:11:24 +0000 (19:11 -0700)]
[BYOC] Support Tuple Output in C/DNNL Codegen (#5701)

* Support tuple output runtime

* fix unit test

4 years ago[ONNX] Skip ADD inside Gemm op when vector is zero (#5697)
Balint Cristian [Sat, 30 May 2020 01:10:22 +0000 (04:10 +0300)]
[ONNX] Skip ADD inside Gemm op when vector is zero (#5697)

4 years ago[PatternLang]Conditionally Embedding Constants in Partitioned Functions (#5693)
Matthew Brookhart [Sat, 30 May 2020 01:07:07 +0000 (18:07 -0700)]
[PatternLang]Conditionally Embedding Constants in Partitioned Functions (#5693)

* Embed constants in the partition function if the pattern explicity requests constants

fix rst

fix pylint

* improve comments based on Cody's feedback

4 years agoIn memory_plan, check if value is not None, instead of just checking value as boolean...
notoraptor [Fri, 29 May 2020 23:35:03 +0000 (19:35 -0400)]
In memory_plan, check if value is not None, instead of just checking value as boolean. (#5700)

4 years ago[ONNX]LpPool Support added (#5696)
Samuel [Fri, 29 May 2020 21:29:43 +0000 (02:59 +0530)]
[ONNX]LpPool Support added (#5696)

4 years agoSupport more dtypes for TVMDSOOp (#5694)
tobe [Fri, 29 May 2020 20:58:34 +0000 (04:58 +0800)]
Support more dtypes for TVMDSOOp (#5694)

4 years ago[COMMUNITY] @masahi -> PPMC (#5691)
Tianqi Chen [Fri, 29 May 2020 16:36:19 +0000 (09:36 -0700)]
[COMMUNITY] @masahi -> PPMC (#5691)

4 years ago@zhiics -> PPMC (#5692)
Tianqi Chen [Fri, 29 May 2020 16:35:47 +0000 (09:35 -0700)]
@zhiics -> PPMC (#5692)

4 years ago[REFACTOR][RELAY] move fallback_device to config (#5690)
Zhi [Fri, 29 May 2020 14:52:03 +0000 (07:52 -0700)]
[REFACTOR][RELAY] move fallback_device to config (#5690)

4 years ago[RELAY] Fix segfault in pretty print when ObjectRef is null (#5681)
lhutton1 [Fri, 29 May 2020 14:51:30 +0000 (15:51 +0100)]
[RELAY] Fix segfault in pretty print when ObjectRef is null (#5681)

* [RELAY] Fix segfault in pretty print when ObjectRef is null

Encountered when pretty printing module with function attribute equal to NullValue<ObjectRef>().

Change-Id: I2e7b304859f03038730ba9c3b9db41ebd3e1fbb5

* Add test case

Change-Id: I579b20da3f5d49054823392be80aaf78a055f596

4 years ago[Relay] Fix dataflow_pattern.rewrite() hang if Match in IR (#5680)
lixiaoquan [Fri, 29 May 2020 10:37:05 +0000 (18:37 +0800)]
[Relay] Fix dataflow_pattern.rewrite() hang if Match in IR (#5680)

rewrite() quits only if graph stop changing, but ExprMutator
  always creates new Match node. This patch fixes this.

4 years ago[PYTORCH]Minor bug fixes (#5683)
Samuel [Fri, 29 May 2020 07:16:49 +0000 (12:46 +0530)]
[PYTORCH]Minor bug fixes (#5683)

* [PYTORCH]Minor bug fixes

* Review comment fix, testcase added

* Added testcase for bert model

4 years ago[PatternLang] Add ConstantPattern (#5689)
Cody Yu [Thu, 28 May 2020 23:48:07 +0000 (16:48 -0700)]
[PatternLang] Add ConstantPattern (#5689)

* Add ConstantPattern

* update doc

4 years ago[TIR][REFACTOR] std::string -> String Migration in TIR nodes (#5596)
Neo Chien [Thu, 28 May 2020 08:56:06 +0000 (16:56 +0800)]
[TIR][REFACTOR] std::string -> String Migration in TIR nodes (#5596)

* [TIR][REFACTOR] std::string -> String Migration for Var node and SizeVar Node

* update json_compact.py

4 years ago[TFLITE]Quantize & Dequantize op (#5394)
Samuel [Thu, 28 May 2020 03:10:58 +0000 (08:40 +0530)]
[TFLITE]Quantize & Dequantize op (#5394)

* [TFLITE]Quantize & Dequantize op

* Testcases added

* Review comment fixed

4 years ago[DOC] Improve Pattern Language Docs (#5676)
Cody Yu [Thu, 28 May 2020 03:07:32 +0000 (20:07 -0700)]
[DOC] Improve Pattern Language Docs (#5676)

* [DOC] Improve Pattern Language Docs

* address comments

* address comments

4 years ago[Bugfix] Fix Python debugger segfaults with TVM built with LLVM (#5685)
Junru Shao [Wed, 27 May 2020 20:50:56 +0000 (13:50 -0700)]
[Bugfix] Fix Python debugger segfaults with TVM built with LLVM (#5685)

* Import readline before loading libtvm

* make lint happy

4 years agoFix the shift column for scale_shift_nchw and scale_shift_nhwc in C topi (#5679)
tobe [Wed, 27 May 2020 15:59:02 +0000 (23:59 +0800)]
Fix the shift column for scale_shift_nchw and scale_shift_nhwc in C topi (#5679)

4 years agoCall previous excepthook in tvm_excepthook. (#5675)
notoraptor [Wed, 27 May 2020 01:15:18 +0000 (21:15 -0400)]
Call previous excepthook in tvm_excepthook. (#5675)

* Call previous excepthook in tvm_excepthook.

* Rename prev_excepthook.

* Create a tvm_wrap_excepthook to wrap a given excepthook with tvm custom excepthook work
and call it on system previous excepthook.

* Add docstring.

4 years agoadd a testcase for #5674 (#5677)
Matthew Brookhart [Wed, 27 May 2020 01:14:58 +0000 (18:14 -0700)]
add a testcase for #5674 (#5677)

4 years ago[BYOC] Pattern Language MergeComposite (#5656)
Cody Yu [Tue, 26 May 2020 22:04:01 +0000 (15:04 -0700)]
[BYOC] Pattern Language MergeComposite (#5656)

* Pattern Language MergeComposite

* fix DNNL pattern

* Use builtin binary operator syntax for demo

* Improve unit test

4 years agoadd a check for null function attributes (#5674)
Matthew Brookhart [Tue, 26 May 2020 20:16:13 +0000 (13:16 -0700)]
add a check for null function attributes (#5674)

4 years agoadd tvm.micro pydoc to sphinx (#5661)
Andrew Reusch [Tue, 26 May 2020 18:30:19 +0000 (11:30 -0700)]
add tvm.micro pydoc to sphinx (#5661)

* add tvm.micro pydoc to sphinx

* making build pass and addressing tqchen comments

4 years ago[TF] Support TupleWrapper as direct ancestor of control flow ops (#5639)
lixiaoquan [Tue, 26 May 2020 18:29:29 +0000 (02:29 +0800)]
[TF] Support TupleWrapper as direct ancestor of control flow ops (#5639)

4 years ago[POC][PatternLang]Remove constants from partitioned functions (#5663)
Matthew Brookhart [Tue, 26 May 2020 17:26:31 +0000 (10:26 -0700)]
[POC][PatternLang]Remove constants from partitioned functions (#5663)

* remove constants from partitioned functions

* remove print statements

4 years ago[AutoTVM][TOPI] AutoTVM incorrect measurement (#5511)
Neo Chien [Tue, 26 May 2020 17:25:50 +0000 (01:25 +0800)]
[AutoTVM][TOPI] AutoTVM incorrect measurement (#5511)

* [AutoTVM][TOPI] AutoTVM incorrect measurement

* create new placeholder with converted layout

* update _schedule_winograd

4 years agoenable amd_apu device on vulkan target (#5659)
Mei Ye [Tue, 26 May 2020 16:30:38 +0000 (09:30 -0700)]
enable amd_apu device on vulkan target (#5659)

4 years ago[C++ RPC] Fix C++ RPC build problem on Linux (#5671)
Zhao Wu [Tue, 26 May 2020 15:36:44 +0000 (23:36 +0800)]
[C++ RPC] Fix C++ RPC build problem on Linux (#5671)

4 years ago[Doc] Misc doc fix (#5672)
Zhao Wu [Tue, 26 May 2020 15:15:44 +0000 (23:15 +0800)]
[Doc] Misc doc fix (#5672)

4 years ago[REFACTOR][TIR][API-Change] Migrate BuildConfig to PassContext. (#5668)
Tianqi Chen [Tue, 26 May 2020 03:50:00 +0000 (20:50 -0700)]
[REFACTOR][TIR][API-Change] Migrate BuildConfig to PassContext. (#5668)

* [REFACTOR][TIR] Migrate BuildConfig to PassContext.

This PR migrates the TIR configurations from BuildConfig to the
PassContext used by the unified IR.
Moving forward, PassContext will be the unified way to configure passes in the TVM stack.

Changes

- Refactored TVM_PASS_REGISTER_CONFIG_OPTION to take in the reference type.
- Removed BuildConfig.
- Migrated the passes to use PassContext.

* Update include/tvm/ir/attrs.h

Co-authored-by: Zhi <5145158+zhiics@users.noreply.github.com>
Co-authored-by: Zhi <5145158+zhiics@users.noreply.github.com>
4 years ago[PYTHON] Add buffer name when creating tensor bindings (#5670)
Tianqi Chen [Tue, 26 May 2020 03:19:15 +0000 (20:19 -0700)]
[PYTHON] Add buffer name when creating tensor bindings (#5670)

4 years ago[Relay][Op]Support symbolic TopK, Ones, Zeros and Full (#5459)
Yao Wang [Tue, 26 May 2020 01:09:44 +0000 (18:09 -0700)]
[Relay][Op]Support symbolic TopK, Ones, Zeros and Full (#5459)

* Support symbolic TopK, Ones, Zeros and Full

* Fix pylint

* Add docstring for topk shape func

* Fix grad

* Fix lazy_gradient_init

* Fix parser

* Fix print ir text

* Fix lint

* Improve pattern_util

* Fix topk

* Fix build

* Use Optional for attribute

* Fix clang-format

* Minot fix

* Fix pylint

* Fix build warning

* Fix parser

* Move ToScalar

* Fix lint

* Fix lint

* Make topk shape func as data independent when k is constant.

* Fix lint

* Minor fix

4 years ago[TOPI] Improve CUDA softmax scheduling (#5600)
Wei Pan [Mon, 25 May 2020 16:44:57 +0000 (09:44 -0700)]
[TOPI] Improve CUDA softmax scheduling (#5600)

- Do not use multiple kernels

- Schedule with warp reductions

- Fixed a bug on the lower warp memory pass

- Fixed warp shuffle intrinsics for the nvptx backend.

Signed-off-by: Wei Pan <weip@nvidia.com>
4 years agohandle likely in IRMutatorWithAnalyzer (#5665)
Shizhi Tang [Mon, 25 May 2020 16:39:33 +0000 (00:39 +0800)]
handle likely in IRMutatorWithAnalyzer (#5665)

4 years ago[TIR][BUILD] Remove buffer params from pass config. (#5652)
Tianqi Chen [Sat, 23 May 2020 15:38:00 +0000 (08:38 -0700)]
[TIR][BUILD] Remove buffer params from pass config. (#5652)

Buffer configurations can be passed during construction
and does not need to be part of the build config.

This is a refactor step to simplify the BuildConfig for the PassContext migration.

4 years ago[Relay,Topi][OP] affine_grid and grid_sample (#5657)
Wuwei Lin [Sat, 23 May 2020 04:57:58 +0000 (00:57 -0400)]
[Relay,Topi][OP] affine_grid and grid_sample (#5657)

* [Relay,Topi][OP] affine_grid and grid_sample

* lint

4 years ago[PatternLang] Convert PatternGrouper to do pre-order, non-recursive analysis (#5653)
Matthew Brookhart [Sat, 23 May 2020 03:17:39 +0000 (20:17 -0700)]
[PatternLang] Convert PatternGrouper to do pre-order, non-recursive analysis (#5653)

* make the PatternGrouper iterate over the input Expr in a non-recursive pre-order fasion

* add a comment

4 years agoIncrease bss section size. (#5660)
Andrew Reusch [Sat, 23 May 2020 02:41:14 +0000 (19:41 -0700)]
Increase bss section size. (#5660)

* Likely broken in PR 5590.

4 years agoUpgrade XGBoost to latest (#5658)
Philip Hyunsu Cho [Sat, 23 May 2020 02:40:46 +0000 (19:40 -0700)]
Upgrade XGBoost to latest (#5658)

4 years ago[REFACTOR][IR] Migrate IRModule ObjectRef to not-null (#5654)
ANSHUMAN TRIPATHY [Fri, 22 May 2020 21:32:59 +0000 (03:02 +0530)]
[REFACTOR][IR] Migrate IRModule ObjectRef to not-null (#5654)

4 years agoµtvm debug improvements (#5648)
Andrew Reusch [Fri, 22 May 2020 21:00:48 +0000 (14:00 -0700)]
µtvm debug improvements (#5648)

* Forever loop in UTVMDone to aid debugging

* Use parameter and callback function as a micro debug hook.

 * Previously, users had to uncomment a region of code in
   micro_session.cc and recompile to debug. Now they can pass in a
   key in the micro.Session config:

       config = tvm.micro.device....generate_config()
       config['debug_func'] = _python_launch_gdb
       with micro.Session(config) as sess:
         ....

* clang-format

* Only forever loop on device (on host this blocks unittests)

4 years ago[RUNTIME] Resolve constexpr issue in debug mode. (#5651)
Tianqi Chen [Fri, 22 May 2020 20:36:43 +0000 (13:36 -0700)]
[RUNTIME] Resolve constexpr issue in debug mode. (#5651)

static constexpr is a bit weird before c++17.
They are not inlined by default and does not have symbols after compilation.
It usually isn't a problem when they are inlined(in c++17 they are inlined by default).
But will create compilation error when passed to functions that take (const)references.
This PR fixes the problem so that we can compile on debugmode.

4 years ago[AutoTVM] Update XGBoost verbosity option (#5649)
Tianqi Chen [Fri, 22 May 2020 18:15:24 +0000 (11:15 -0700)]
[AutoTVM] Update XGBoost verbosity option (#5649)

4 years agoHG: Commit message of changeset 6281661. (#5622)
hlu1 [Fri, 22 May 2020 16:38:30 +0000 (09:38 -0700)]
HG: Commit message of changeset 6281661. (#5622)

[Relay] Move compiler_begin/end_op to local static objects

4 years ago[Relay, Topi][OP] Correlation (#5628)
Wuwei Lin [Fri, 22 May 2020 06:00:33 +0000 (02:00 -0400)]
[Relay, Topi][OP] Correlation (#5628)

* [Relay,Topi] Correlation

* fix

* move

* typo

* Update test_topi_correlation.py

4 years agoAdd a check Callback to the Pattern Paritioner (#5646)
Matthew Brookhart [Fri, 22 May 2020 04:35:35 +0000 (21:35 -0700)]
Add a check Callback to the Pattern Paritioner (#5646)

* add a check callback to the paritioner

* fix doc string

* fix unit test spelling

* add a test with types

4 years ago[RELAY][BYOC] Preserve type information in Merge Composite (#5640)
mbaret [Fri, 22 May 2020 02:24:23 +0000 (03:24 +0100)]
[RELAY][BYOC] Preserve type information in Merge Composite (#5640)

Keep the type information when extracting patterns
so that it can be used as part of 'check' functions.

Change-Id: I16cc70c3d013a794d2ceefb5bec815129c7b8825

4 years ago[CI] Add log check to the sphinx gallery docs (#5643)
Tianqi Chen [Fri, 22 May 2020 01:37:58 +0000 (18:37 -0700)]
[CI] Add log check to the sphinx gallery docs (#5643)

* [CI] Add log check to the sphinx gallery docs

This PR add log check to sphinx gallery tutorials to prevent
the case when sphinx failed to capture the error in tutorials.

* Fix the status

4 years ago[Runtime] Introduce runtime::Array (#5585)
Junru Shao [Thu, 21 May 2020 23:27:14 +0000 (16:27 -0700)]
[Runtime] Introduce runtime::Array (#5585)

* Introduce runtime::Array

* Sync with dmlc-core

* Tests added: size, capacity, empty, front, back, push_back, pop_back, insert * 2, erase * 2, resize, reserve, clear

4 years ago[CI] Allow CI_PYTEST_ADD_OPTIONS to be unbound. (#5644)
Tianqi Chen [Thu, 21 May 2020 23:01:19 +0000 (16:01 -0700)]
[CI] Allow CI_PYTEST_ADD_OPTIONS to be unbound. (#5644)

This patch allows the test script to execute normally
when CI_PYTEST_ADD_OPTIONS is not available.

4 years agoRemove unnecessary print (#5642)
Cody Yu [Thu, 21 May 2020 22:09:28 +0000 (15:09 -0700)]
Remove unnecessary print (#5642)

4 years ago[PYTORCH]Padding support (#5638)
Samuel [Thu, 21 May 2020 20:44:17 +0000 (02:14 +0530)]
[PYTORCH]Padding support (#5638)

4 years ago[RUNTIME][VULKAN] Seg fault in WorkspacePool's destructor (#5632) (#5636)
Yi Wang [Thu, 21 May 2020 19:07:54 +0000 (12:07 -0700)]
[RUNTIME][VULKAN] Seg fault in WorkspacePool's destructor (#5632) (#5636)

* [RUNTIME][VULKAN] Seg fault in WorkspacePool's destructor (#5632)
* fixed this issue by changing WorkspacePool's destruction order

* make line < 100 charactors long

4 years ago[DOCS] Fix the QNN TFLite tutorial build (#5641)
Tianqi Chen [Thu, 21 May 2020 17:22:31 +0000 (10:22 -0700)]
[DOCS] Fix the QNN TFLite tutorial build (#5641)

* [TUTORIAL] Fix execution error of TFLite quantized tutorial

* Assign TensorCore to docs build

4 years agoExtend AttrPattern to support CallNode and FunctionNode attributes (#5637)
Matthew Brookhart [Thu, 21 May 2020 03:16:40 +0000 (20:16 -0700)]
Extend AttrPattern to support CallNode and FunctionNode attributes (#5637)

* Extend AttrPattern to support CallNode and FunctionNode attributes

* Update tutorial and add breaks

* add func attr test

4 years ago[TUTORIAL]TFLite QNN Tutorial (#5595)
Samuel [Thu, 21 May 2020 00:39:14 +0000 (06:09 +0530)]
[TUTORIAL]TFLite QNN Tutorial (#5595)

* [TUTORIAL]TFLite QNN Tutorial

* Review comments

4 years ago[RELAY][PYTORCH]Resize3d, Upsample3d op support (#5633)
Samuel [Thu, 21 May 2020 00:35:28 +0000 (06:05 +0530)]
[RELAY][PYTORCH]Resize3d, Upsample3d op support (#5633)

4 years agoLabel Pattern Partitions (#5627)
Matthew Brookhart [Thu, 21 May 2020 00:28:43 +0000 (17:28 -0700)]
Label Pattern Partitions (#5627)

* Label Pattern Partitions with a default label to prevent nested partitions and an optional user supplied-label

* Add node names in topological order to Partitioned attribute

* respond to review comments

* move partition tag into const in attr namespace

4 years agoFix typo in test script (#5635)
Wuwei Lin [Wed, 20 May 2020 22:12:37 +0000 (18:12 -0400)]
Fix typo in test script (#5635)

4 years agoanother cmake fix (#5630)
Thierry Moreau [Wed, 20 May 2020 22:12:28 +0000 (15:12 -0700)]
another cmake fix (#5630)

4 years ago[NODE][PASS] Introduce config to PassContext. (#5631)
Tianqi Chen [Wed, 20 May 2020 21:54:11 +0000 (14:54 -0700)]
[NODE][PASS] Introduce config to PassContext. (#5631)

This PR introduces a new config field to the PassContext
to allow it store arbitary config values.

To make sure that the config is validated, we allow each pass
to register the config key they would expect and the corresponding types.

We also introduce a CreateObject from Map<str, Object> to allow config creation
from a json-nest(like in vscode) in python.

We added an example of UnrollLoopConfig.

Followup PR should migrate the passes to use the new config field.