Update #7

…dded a unit test (#2487)

Signed-off-by: Zhebin Jin <zhebin.jzb@alibaba-inc.com>

* TFLite frontend tutorial * Modify as suggestion

* [Relay][Frontend] Add CoreML Support * pip install six in CI * remove triggering nnvm coreml test * set opt_level=2 for nnvm coreml test case

) * nms fixed for gpu, tested on cuda and opencl devices, ssd now can run fully on the gpu * sort updated to use virtual thread * typo fixed * fix lint * fix lint * add support when batch_size > 1 * intel graphics conv2d bugs fixed for inception_v3 * intel conv2d api updated, nn input size 4 condition added * review addressed * move conv_tags to attributes * opencl ctx fixed * nms_ir index simplified

…odify TFLite whl files path (#2529)

* [QUANTIZE] Quantization implementation. * Update. * Update. * Update. * Update.

@FrozenGene

* [COMMUNITY] @FrozenGene -> Reviewer * Fix

* [Relay][Frontend] Add Caffe2 Support * [Relay][Frontend] Add Caffe2 Support (fix unsed import) * [Relay][Frontend] Add Caffe2 Support (fix caffe2 model import) * [Relay][Frontend] Add Caffe2 Support (fix model install and reflect code reviews) * [Relay][Frontend] Add Caffe2 Support (fix caffe2 model import) * [Relay][Frontend] Add Caffe2 Support (fix caffe2 model import) * [Relay][Frontend] Add Caffe2 Support (fix caffe2 model import) * [Relay][Frontend] Add Caffe2 Support (fix caffe2 frontend import) * [Relay][Frontend] Add Caffe2 Support (rename function name in test_forward) * [Relay][Frontend] Add Caffe2 Support (fix caffe2 model import) * [Relay][Frontend] Add Caffe2 Support (fix caffe2 model import) * [Doc] Caffe2 frontend tutorial * [Doc] Caffe2 frontend tutorial * [Doc] Caffe2 frontend tutorial * [Relay][Frontend] Add Caffe2 Support (remove unsed file)

* Enable reverse in reshape * Fix lint and typo * Put reverse_reshape into a separate op * Fix pylint

* [RELAY][FRONTEND] Tensorflow frontend support. * * LSTM removed for a while. * * basic ops are good. * * nn wip * * wip * * python2.7 corrections. * * NN ops are good. * * e2e models working good * * all good except LSTM * * rebase, tutorials and CI trigger. * * CI errors. * * enable opt_level=3 * * Docstrings cleanup. testing.tf utils moved to relay from nnvm. * * tutorials update. * * LSTM work good now. * * Rebase * * CI error * * enable PTB. * * rebase. * * tutorials * Update python/tvm/relay/frontend/tensorflow.py Co-Authored-By: srkreddy1238 <sivar.b@huawei.com> * * review comments. * CI fix. * * review comments.

* [TVM][BUGFIX] Fix reductions in split axes * A test case for the problem * Fix the fix: skip loops that are related to reduction AND are unrelated to axis

…ich causes atomic contention (#2576)

TF parser: return the consistent error message to error handler

Remove exception trail in tf parser error message Fix lint Fix comments

…es (#2565)

thanks @icemelon9, this is merged

…#2594)

@vinx13

* a preliminary version is done? * we no longer need the redundant hybrid/api.py * support assert stmt * cast supported * intrin -> runtime; util is mainly in charge of compilation time * assert statement * fix python lint * fix cpp lint * on the way to module * rollback .cc * fix typo, no direct expose then * @vinx13 ceil is added i guess? * wip... * temp commit * fix import * i preliminary version is done? * on the way to build hybrid module * nearly fixed... * dumped python are equiv as original python * on the way to bootstrap * cpu bootstrap done * bootstrap! * fix lint * fix doc * resolve some review concerns * support load/save * fix lint * thanks to xqdan fixed my typo * fix build, make dump non-optional * add vthread * jesus why i added this

* [TOPI][CUDA] Add faster-rcnn proposal op * Fix doc * Add global barrier * Use vthread in argsort * Update sort and nms ir * Fix lint * Update sort ir in ssd nms

* fix storage_rewrite bug when input is big * cast when necessary * simplification * simplification * int64->uint32 * revert uint32->int64

) * update titles to reflect tutorial content (nnvm vs. relay) * move things around * fix typo

* move fix test fix lint fix test add more code fix lint better type infer ability * fix build * address comment

* Version 0.5 * update version.py * update news * update news * update news

* First pass on ADTs * Add doc string for tag field * Visit constructors in TypeVisitor for TypeData * Add to description of type call * Add type call to type solving and unification * Make type mutator for typecall consistent with others (only create new node if there's a change) * Ensure kindchecking can handle type calls and typedata * Fix bad nesting in module constructor * Correctly construct call in typecall test * Add call override for ordinary vars (do we want this?) * Remove generalization hack from type inference because it was breaking ADT constructors * Check that there are no free type vars in exprs after inferring type * Free var checks need module because of ADT constructors * Typecall test can't have unbound type var, make it global * Uncomment tmap test and remove comments about failing to infer ret type; those work now * Put in dummy visits for ADTs in graph runtime codegen to placate pylint * Fix Relay type infer test module constructor * Mark override for TypeCallNode in type solver * Ensure free vars check treats patern vars as bound * Run interpreter in more ADT test cases * Refactor kind check to return the kind, like typechecking * Fix invalid typecall in test * Add kind check to type inference, do not use nulls in func_type_annotation()! * Redundant whitespace * Make TypeData a separate kind * Make ADT handles a separate kind too, document calling convention better * Remove nats and tree from prelude, move to test, document prelude * Restore and document nat and tree to prelude, add more tree tests * Add alpha equality tests for match cases, fix variable binding bug * Add more kind check tests for ADTs * Add more tests for finding free or bound vars in match exprs * Add unification tests for type call * Update main() for alpha equality tests * Add simple type inference test cases for match exprs and ADT constructors * Add more ADT interpreter tests * Allow incomplete types when typechecking match cases * Type inference for pattern vars should use the type annotation if it's there * Two more specific test cases for ADT matching * Add option ADT to prelude * Fix broken reference to kind enum * Fix rebase snags * Do not attach checked types to constructors * More docstrings for module fields * Use proper wrapper for indexing into module type data * checked_type for constructors is not populated * Expand type call docstring * Rename PatternConstructor con field * Use error reporter for pattern constructor case * Condense error reporting in kind check, use error reporter * Expand docstrings and rename ADT fields * Rename 'option' ADT to 'optional' for consistency with Python * Add various list iterators and utility functions to prelude * Add smoke tests for new iterators in prelude * Add concat to prelude * Add smoke test for concat * Correct docstrings in prelude * Ensure that type defs are written in module initialization * Various requested renamings * Correct rebase snags * Add kind check tests for ref types * Update the main() for kind checking tests

* check in * update build and run

* alter_op_layout for x86 * cleanup * cleanup * fix lint * fix lint * fix lint * fix lint * change support level * change other support levels

* nms data race solved * tst_topi_vision reference results are gonna be updated in PR #2353 * proposal nms_ir updated

) * Add eager simplication for FloatImm * fix * fix lint * Fix gcc warning * fix * Add test case

@hlu1

Thanks @hlu1 and @tqchen, this is now merged.

* add a relay pass to count #macs of a model * remove unnecessary includes * fix end-of-file issues * address review comments * remove test resnet * address more review comments * use data layout string to locate the input channel * fix bug in conv 2d output tensor checking * C must exist

* [Tutorial][Frontend] move from_keras tutorial to frontend * remove tutorial/nnvm/from_keras.py

…2613)

… pass (#2646) * check in * fix typo * fix typo * change message * change message * typo * lint

Recent pylint warngs about import renames with no effect. Remove them.

…2645)

* Add arange op * Update docs * Fix bug * add sanity check in relay and mxnet frontend mapping * lint * nits * pylint * don't allow empty output from arange * Remove empty test for arange * Fix bug and update doc

* i think it works for now? * fix lint * fix 2/3 compat * fix py2 again * fine, i gave up

…d create free variable). (#2665)

* Add CONCATENATION to tflite frontend * fix typo * Fix codestyle * Fix code style * simplify convert map * Update

)

…2676)

* [DOC] CoreML frontend tutorial * Update tutorials/frontend/from_coreml.py Co-Authored-By: kazum <morita.kazutaka@lab.ntt.co.jp> * Update tutorials/frontend/from_coreml.py Co-Authored-By: kazum <morita.kazutaka@lab.ntt.co.jp> * Addressed comments and added the original author

Recent pylint introduced support for the useless-return diagnostic. This patch remove the useless returns.

tshape.strip('()[]').split(',') will make a list ['1001',''] but the empty one isn't needed.

* check in * lint * cleanup * Update util.py

* move layout.h & layout.cc from relay to tvm * change ConvertLayout in relay to bijectiveLayout->Forward/backward * add first test case * add LayoutAxis * add LayoutAxis struct and compiles * simplify BijectiveLayout rule consturct * polish func name for Layout, move impl to .cc, remove Layout::defined(), add defined() checker * partially add layout py support * add layout test cases * add doc for tvm.layout & tvm.bijective_layout * fix lint * fix lint * fix layout name generation bug * fix layout typo * address comments and add topi.layout_transform * layout.h->data_layout.h, test_lang_layout.py->test_lang_data_layout.py

* added relay quick start * added relay/using_external_lib.py * update using_external_lib * Update using_external_lib.py * update tvm/make/config.mk -> cmake/config.cmake * Fixed: result mismatched when lowering relay with cudnn support at opt level 2 * setting opt_level=2 and out_channels=16 for consistency of original tutorial * Fixed some typos

* [DOCKER] Switch from yes|apt-get to apt-get -y The yes | apt-get idom guarantees that the 'yes' process always exists with exit code 141 (pipe broken). This is fine while the script generally ignores failures but won't work when the script behaviour is tightened to robustly catch errors. * [DOCKER] Turn down the wget/curl volume

* do * address comment

…2707)

* [SCRIPT] Refactor grep for multiple patterns Tidy up the use of grep. Use -E rather than run multiple grep instances. * [SCRIPT] Refactor grep use in pipeline. Prefer to use stdin redirection rather than create a pipeline. * [SCRIPT] Refactor placement and cleanup of temporary files. Place temporary files in the conventional /tmp location. Avoid poisoning file name space by using $$. Ensure the temporary files get cleaned up, even when the script fails / exits early. * [SCRIPT] Improve robustness of task_lint.sh error handling. Ensure script failures are caught and propagated. Rather than trying to explicitly catch and propagate failures with explicit "|| exit" annotations, use the "set -e" idom from docker/install scripts and have the shell catch and propagate errors in the general case and special case the grep instances where non zero exit is permitted and should be ignored.

* [DOCKER] Make all install .sh scripts directly executable. * [DOCKER] Use curl -L consistently. Make the use of the curl -L option in docker build scripts consistent. * [DOCKER] Drop use of --force-yes The --force-yes option is generally not recommend, it can leave systems in an undefined state. The use of --allow-* options is preferred. In this particular case the --force-yes option appears to serve no purpose. Dropping it. * [DOCKER] Drop superflous repeated apt-get update. The "apt-get update && apt-get install" idiom is necessary and specific to Dockerfile. In shell the repeated apt-get update is superflous. Drop the duplicates. * [DOCKER] Robustness -e -u -o pipefail The install scripts used to construct docker environments do not, in general, propagate errors. Some of the scripts use adhoc && directives to chain together short sequences of commands but there are numerous failure modes which are silently ignored. This patch puts in place some consistent, basic, shell error trapping across all of the install scripts. Note this is a step forward towards more robust scripts but it is not a complete solution. * [DOCKER] Shallow clone. Use shallow clone to reduce bandwidth requirements of repeated docker (re)-builds. * [DOCKER] Use clone --branch rather than clone then checkout Use the git clone --branch idiom rather than git clone && git checkout. This paves the way for using --depth=1

…2693)

@ZihengJiang

* Defined a common base class for TensorComputeOp and ComputeOp * Made changes requested by @ZihengJiang * added a testcase to assert that `tensorize` does not have any effect on TensorComputeOp ops.

* Add batch_dot and cpu schedule * Add relay support for batch_dot * Rename batch_dot to batch_matmul * nits * Add missing file * Put batch_matmul and dense x86 schedule in separate files * Fix pylint * Remove unused import * Add cuda schedule for batch_matmul * Add test case with larger batch size * Add batch_matmul in api doc * Fix quantize pass rounding error * Fix pylint and minor change * bug fix

…2719)

* Add slice axis op in mxnet converter * Fix lint

* fix docments * delete e

… with control flow (#2496) add test remove dead code stash do it add more test

* use LLVM linker * error message improved in case of filenotfound * linting error fixed

* [RELAY][OP] Proposal * Fix * Fix test

…is negative (#2739) * bug fix * trigger ci

…ded stack op (#2729) * add stack op frontend * concate moved * topi stack added * stack added * fix stack bugs and tested * conv2d alterlayout udpated for relay * fix pylint * fix cmake warnings * cmake warnings fixed

…2744) * Fix bug in mxnet converter for slick_like * More tolerance for topi_conv2d_NCHWc

* Improve NNVM to Relay conversion * fix pylint * support __lshift_scalar__, abs, ceil, floor, and trunc to pass CI

) * [Relay][Quantization] Speed-aware quantization scheme improvement * Add comment * Add use_stop_fusion to qconfig * Update comment

…2741)

#2720) * tile and repeat operator added in rely * fix pylint * fix make warnings * comments addressed * fix lint error * comment addressed

* [relay][frontend] TensorFlow saved model support * Add Examples section * keep one copy of tensorflow_parser in relay

@tqchen

* initial commit * add python frontend and module tests * add unit tests for function pass and optimize interface * add ExprPass * remove PassState and pass context for run * add required_passes * return module * remove move * fix minor reviews * remove optimizer, optimizer->pass_manager, make pass a the base class of all * remove deleted files * move resolvedependency to sequential pass, use ir_pass namespace * add todo * add disabled passes in sequetialpass * fix minor * fix currying doc * remove pass_kind from passnode * remove pass kind from test * fix doc * fix per @tqchen's comments * remove pass_manager.py create separate classes * simplify pass_func * inline using passfunc * update doc * disable test_quantize_pass for now * create PassInfo class to contain the meta data * flatten passinfo for interface * retrigger ci * remove required method * make Pass python class lighter * create pass -> decorator * make the api consistent for all classes

#2780)

* start adding reverse * reverse updated * reverse uses topi::flip * typo fixed * comment addressed * exp simplified

* Add shapeof op in topi * Add relay shape_of op * Add constant folding for shape_of * Allow shape op to specify dtype * Add mxnet converter for shape_array * lint * lint * Add doc

git clone --branch=xxx won't take a hash, switch from the hash to the tag that represents that hash.

* adding _contrib_BilinearResize2D op from mxnet * error fixed * use resize instead of upsample

Code like this can't be built with NV OpenCL, and it needs an explicit type converison for ternary expression if return type is uchar. uchar i = 0, j = 0; uchar t = max((uchar)j, ((i > 0) ? (uchar)1 : (uchar)0));

This patch reverts one of my earlier patches (squashed in #2710) to reduce bandwidth requirements of git clone, in this particular case we are checking out a specific hash rather than a tag or branch name. The --branch option to git clone permits tags or branches but does not permit a specific hash.

* [FRONTEND][TENSORFLOW] Enhance with left over patches from NNVM. commit 76188a4 Author: Siva sivar.b@huawei.com [NNVM][TENSORFLOW] bugfix. (#2444) commit 6737739 Author: Ashutosh Parkhi ashutosh.parkhi@imgtec.com [Tensorflow] Support for Crop (#2285) commit f6c3f99 Author: Alexey Romanov alexey.v.romanov@gmail.com [FRONTEND][TENSORFLOW] Use input shapes directly instead of 1-element lists (#2242) commit e5d92e1 Author: Dominic Symes 36929632+dominicsymes@users.noreply.github.com [FRONTEND][TENSORFLOW] Bugfix (#2326) commit 00d509d Author: Alexey Romanov alexey.v.romanov@gmail.com [FRONTEND][TENSORFLOW] Support Unstack and Split (#2105) commit df9d3ad Author: Siva sivar.b@huawei.com [FRONTEND][TENSORFLOW] Bugfix (#2267) commit d1a0c90 Author: Zhebin Jin zhebin.jzb@alibaba-inc.com [FRONTEND][TENSORFLOW]Add Split and realdiv op support (#2123) * Add Split and realdiv op support * Fix the pad calculation in the case of dilated convolution * * review comments * * resnet fix. * * review comments

…2854) * [Relay][Frontend][keras] added interpolation method of Upsampling2D. * added testcase * small fixes

* [AlterLayout] NCHW upsampling * [Relay][Pass] Fix Depthwise AlterLayout

* Add ops in mxnet converter * trigger ci

The dict which is used to initilize OrderDict is not ordered, so metadata may not be at the end.

* improved `multiply_rewrite` for gluoncv_ssd quantization * add a check to attach_simulated_quantize below to prevent quantizing lhs to INPUT if lhs_kind is INPUT;

* Fix error reporting for missing axis * Update data_layout.cc

…2879)

* decompile tf control flow * Add docs * remove import relay * move tests under tensorflow frontend * minor fix

* [Relay][Op] Add group conv2d dispatch to topi function * Rerun tests

* Add dense schedules to __init__ for cpu * Add documentation for topi::shape * Add additional imports to topi CPU __init__.

A number of test scripts use the '|| exit 1' idiom. This has two issues, first process exit codes are defined to be in the range 0-255. Second, more importantly, the idiom is fragile because it requires that every possible failure point be explicitly coded. This patch removes the idiom in favour of "set -e" as used in the docker scripts as a more robust mechanism to ensure that script failures are always caught and propagated by default.

…yout issue. (#2861) * Fix Relay ARM CPU spatial pack depthwise alter op layout issue. * Update tune_relay_arm.py

Adopt the "set -u" idiom from the docker scripts as a mechanism to improve future robustness.

* Allow converting keras.layers.Sequential * Use existing new_var function * Only update expr when missing * Add test

Commits on Feb 10, 2019

[IR] Update HalideIR (#2582 )

tqchen authored Feb 10, 2019

Configuration menu

View commit details

Copy full SHA for 89deaa6

Browse repository at this point

Copy the full SHA

89deaa6 View commit details

Browse the repository at this point in the history

Update #7

Update #7

Commits on Jan 24, 2019

Commits on Jan 25, 2019

Commits on Jan 29, 2019

Commits on Jan 30, 2019

Commits on Jan 31, 2019

Commits on Feb 1, 2019

Commits on Feb 2, 2019

Commits on Feb 3, 2019

Commits on Feb 4, 2019

Commits on Feb 5, 2019

Commits on Feb 6, 2019

Commits on Feb 7, 2019

Commits on Feb 8, 2019

Commits on Feb 9, 2019

Commits on Feb 10, 2019

Commits on Feb 12, 2019

Commits on Feb 13, 2019

Commits on Feb 14, 2019

Commits on Feb 15, 2019

Commits on Feb 16, 2019

Commits on Feb 17, 2019

Commits on Feb 18, 2019

Commits on Feb 19, 2019

Commits on Feb 20, 2019

Commits on Feb 21, 2019

Commits on Feb 22, 2019

Commits on Feb 23, 2019

Commits on Feb 24, 2019

Commits on Feb 25, 2019

Commits on Feb 26, 2019

Commits on Feb 27, 2019

Commits on Feb 28, 2019

Commits on Mar 1, 2019

Commits on Mar 2, 2019

Commits on Mar 3, 2019

Commits on Mar 4, 2019

Commits on Mar 5, 2019

Commits on Mar 6, 2019

Commits on Mar 7, 2019

Commits on Mar 8, 2019

Commits on Mar 9, 2019

Commits on Mar 10, 2019

Commits on Mar 11, 2019

Commits on Mar 12, 2019

Commits on Mar 13, 2019

Commits on Mar 14, 2019

Commits on Mar 15, 2019

Commits on Mar 16, 2019

Commits on Mar 17, 2019

Commits on Mar 18, 2019

Commits on Mar 19, 2019

Commits on Mar 20, 2019

Commits on Mar 21, 2019

Commits on Mar 23, 2019

Commits on Mar 24, 2019

Commits on Mar 25, 2019

Commits on Mar 26, 2019

Commits on Mar 27, 2019

Commits on Mar 28, 2019