[docs] Getting Started with TVM: Auto Tuning with Python by hogepodge · Pull Request #7641 · apache/tvm

hogepodge · 2021-03-11T22:37:40Z

Follows up on the TVMC tutorial, shows how to accomplish the same tasks using the python api

hogepodge · 2021-03-11T22:38:40Z

@tqchen @leandron

leandron · 2021-03-12T08:55:26Z

cc @mbaret

leandron

Looks great, and give people a great perspective on how to use the Python API. Just a few minor comments.

tutorials/get_started/auto_tuning_with_python.py

Removed extra spaces, matched file names, lazy import of timeit.

* Relax simulated qnn tests to prevent flakiness. * Change name of helper to make pytest happy.

* add support for optional args for frontends tvmc * remove unnecessary comments * Add changes suggested by Matt W. via PR Co-authored-by: Jocelyn <jocelyn@pop-os.localdomain>

* [RUNTIME] Add libbacktrace for backtraces with line numbers Co-authored-by: Robert Kimball <bobkimball@gmail.com>

…7677)

* [CPP_RPC] allow user supplied work dir * clang format

* [TFLite] Cast operator adapted for MLIR-based convertor Cast operator now can be executed in MLIR-based version. Unit test updated Change-Id: I30e5c1c9d69355116b560af8f6d0582b2d593538 * Comment added Change-Id: I3e2d29ef201283de337168d0b82679b63ca2fcf4

… no in the graph. (#7699)

* [AutoScheduler] Add function name in message * fix

…#7700) * Handle 0 box case for return_indices=False case * Add unit test for mx NMS

* [RUNTIME] Cleanup build for libbacktrace - Introduce TVM_USE_LIBBACKTRACE value macro to be consistent wth other value macros(instead of relying on disabled flag). - Introduce AUTO mode for libbacktrace - Temporary disable MacOS support in light of recent bug report. - Refactor out the libbacktrace.cmake to libs - Properly use TVM_DLL so that code is cross platform. - Fallback to the weaker dmlc impl when backtrace is disabled. * Update Logging.cmake * Update the macro check order to be consistent with the rest.

Co-authored-by: Bohan Hou <32121147+spectrometerHBH@users.noreply.github.com> Co-authored-by: Junru Shao <junrushao1994@gmail.com> Co-authored-by: Tianqi Chen <tqchen@users.noreply.github.com> Co-authored-by: Ruihang Lai <lairuihangdongdong@qq.com> Co-authored-by: Hongyi Jin <3231950289@qq.com> Co-authored-by: Wuwei Lin <wuwei@apache.org> Co-authored-by: Tristan Konolige <tristan.konolige@gmail.com> Co-authored-by: Cody Yu <comaniac0422@gmail.com>

Co-authored-by: Junru Shao <junrushao1994@gmail.com>

- Updated default Ethos-N driver stack to 21.02 - Fixed some test failures associated with this change

* [TOPI] Use fixed thread block size in unique op for Vulkan * forgot to add min for non vk backend

Set with consumers is empty during preparing auto scheduler sketches for Metal device. Added check on the size of the set. In case when the set with consumers is empty we just skip this rule.

* This adds a new '--tag' parameter so that we can build docker images on a particular tag, not only ':latest' as given by Docker * It opens up the possibility of generating "staging" images on a different tag, in the same servers as we keep the production images * By default it keeps previous behaviour of using ':latest' tag.

A diamond dependency currently generates a duplication of ancestors: A / \ B C \ / D under some conditions, this scenario leads to the following dictionnary entry: "D":[A, A] instead of "D":[A]. This results in failures when subsequently trying to transpose node states based on this data. Change-Id: I72f9b19286bbab0581b851c228b9d0e79ead400f

leandron

LGTM, with minor nit.

tutorials/get_started/auto_tuning_with_python.py

leandron

LGTM

fix minor formatting

* [docs] Getting Started with TVM: AutoTVM and Matrix Multiply This patch moves the matrix multiplcation example tuning with AutoTVM to the tutorial directory, and expands on the content. This follows and builds on the section on TE * Applying lint style * Fix license * Apply suggestions from code review Co-authored-by: Siyuan Feng <Hzfengsy@sjtu.edu.cn> * Change comparison tolerance to smaller value Co-authored-by: Siyuan Feng <Hzfengsy@sjtu.edu.cn>

* [RUNTIME][WEB] Cleanup logging for web runtime. The log(info) won't work for web runtime due to the usage of time function. - Introduce TVM_LOG_CUSTOMIZE anf TVM_LOG_STACK_TRACE. - Reorganize the log customization code to wasm_runtime(so non-gpu usecase can apply). - Update the testcase to cover the logging. * Fix windows build and address comment.

* Getting Started: Introduction and Installation The first two sections of the "Getting Started with TVM" guide. A high level introduction to TVM, and a slight introduction on installation options. Co-authored-by: Tianqi Chen <tqchen@users.noreply.github.com>

* Some docstring fixes. * Couple of small fixes: - Use `west attach` instead of `west debug` in commandline to prevent debugger from resetting device. - Fix warning on use of led_pin in zephyr-runtime/src/main.c. * Adding Zephyr demo runtime. * Cleanup of uTVM tests and demo runtime. * Working on QEMU support. Need to add board-specific prj.conf files. * Adding board-specific prj.conf files. * Some cleanup. * Lots of hacking to get ONNX model to run on QEMU and nRF5340. Added test_onnx unit test. Still need to clean up tutorial. * Adding data for unit tests. * Cleanup demo_runtime code. * Fix up tutorial. * Couple of small fixes: - Use `west attach` instead of `west debug` in commandline to prevent debugger from resetting device. - Fix warning on use of led_pin in zephyr-runtime/src/main.c. * Adding Zephyr demo runtime. * Cleanup of uTVM tests and demo runtime. * Working on QEMU support. Need to add board-specific prj.conf files. * Adding board-specific prj.conf files. * Some cleanup. * Lots of hacking to get ONNX model to run on QEMU and nRF5340. Added test_onnx unit test. Still need to clean up tutorial. * Lots of hacking to get ONNX model to run on QEMU and nRF5340. Added test_onnx unit test. Still need to clean up tutorial. * Adding data for unit tests. * Cleanup demo_runtime code. * Fix up tutorial. * Fix tutorial. * Fix tutorial and runtime. * Fix merge conflicts. * Fix merge conflict. * Remove redundant files. * Revert dep. * Fixup * Add new files to check_file_type.py. * Adding missing ONNX file. * Fixup docs. * Fix linting rule. * small fixes * Add missing file to check_file_type.py. * clang-format this file. * Fix formatting. * Black formatting. * Lint comments. * Fix path for test. * Bump CI. * Update from_onnx. * fix path * Fixing * Revert dmlc-core to 21cc7de0dc9fd6acb796e1be6181fa8e6b6c8f41 * Fix path again. * Fix tutorial to not use actual Zephyr. * Revert submodule version change * Fix bad merge. * Trying to fix this mess. * Fix formatting. * context -> device * Removing tutorial since I can't get it to pass CI. Co-authored-by: Mehrdad Hessar <mehrdad.hessar@gmail.com> Co-authored-by: Andrew Reusch <areusch@octoml.ai>

…s. (#7761)

Follows up on the TVMC tutorial, shows how to accomplish the same tasks using the python api

Removed extra spaces, matched file names, lazy import of timeit.

fix minor formatting

…o tutorial-python-resnet

hogepodge · 2021-03-29T17:46:46Z

Merge conflicts have turned this into a mess. I'm going to abandon and push a clean patch.

[docs] Getting Started with TVM: Auto Tuning with Python

759c2ab

Follows up on the TVMC tutorial, shows how to accomplish the same tasks using the python api

hogepodge mentioned this pull request Mar 11, 2021

[WIP][Docs] Getting Started With TVM Tutorial #7612

Closed

Apply patch suggestions from black

299b7f2

leandron reviewed Mar 12, 2021

View reviewed changes

hogepodge and others added 24 commits March 17, 2021 14:02

Minor corrections and updates

d5cec00

Removed extra spaces, matched file names, lazy import of timeit.

Merge branch 'main' into tutorial-python-resnet

70c9b95

[Relay][QNN] Relax simulated qnn tests to prevent flakiness. (#7684)

ab86aa6

* Relax simulated qnn tests to prevent flakiness. * Change name of helper to make pytest happy.

[Relay] Add TopPattern to nn.dropout (#7685)

4976bb2

[TVMC] Allow optional arguments to be passed to importers (#7674)

38aed59

* add support for optional args for frontends tvmc * remove unnecessary comments * Add changes suggested by Matt W. via PR Co-authored-by: Jocelyn <jocelyn@pop-os.localdomain>

[RUNTIME] Add libbacktrace for backtraces with line numbers (#7153)

c976a07

* [RUNTIME] Add libbacktrace for backtraces with line numbers Co-authored-by: Robert Kimball <bobkimball@gmail.com>

[Relay][Training][Pass] Factor out first-order AD to a module pass (#…

45442ed

…7677)

Default value for graph_runtime Init lookup_linked_param_func (#7676)

431a7d6

[CPP_RPC] allow user supplied work dir (#7670)

e467748

* [CPP_RPC] allow user supplied work dir * clang format

Free TensorRT engine and context (#7702)

570767f

Change behavior of onnx importer to throw when user provides an input…

35b43e1

… no in the graph. (#7699)

[Vulkan] Workaround for zero size allocation (#7691)

9a29141

[AutoScheduler] Add function name in message (#7703)

aa494cf

* [AutoScheduler] Add function name in message * fix

[TOPI][CUDA] Fix 0 valid boxes case for NMS when return_indices=False (…

7605f65

…#7700) * Handle 0 box case for return_indices=False case * Add unit test for mx NMS

[torch] Use try_infer_value for clamp min/max (#7712)

27f1085

[TensorIR] add TIRTextPrinter support for Block and BlockRealize (#7716)

4b528de

Co-authored-by: Junru Shao <junrushao1994@gmail.com>

[ETHOSN] Add support for Ethos-N 21.02 driver stack release. (#7628)

c4b8934

- Updated default Ethos-N driver stack to 21.02 - Fixed some test failures associated with this change

[TOPI] Use fixed thread block size in unique op for Vulkan (#7718)

21fc3bb

* [TOPI] Use fixed thread block size in unique op for Vulkan * forgot to add min for non vk backend

Fix auto scheduler crash when set with consumers is empty (#7708)

318c650

Set with consumers is empty during preparing auto scheduler sketches for Metal device. Added check on the size of the set. In case when the set with consumers is empty we just skip this rule.

Fix missing <cassert> header, caused compilation failure. (#7740)

67a4a8b

leandron approved these changes Mar 25, 2021

View reviewed changes

tutorials/get_started/auto_tuning_with_python.py Outdated Show resolved Hide resolved

Added link to tutorial

480278d

leandron approved these changes Mar 25, 2021

View reviewed changes

tqchen and others added 21 commits March 26, 2021 14:25

[CI] Temp disable rust docs build (#7743)

8c8971a

[Refactor] Rename TVMContext to Device (#7721)

fbfeee4

[Bugfix] Fix usages of logging-related macros (#7748)

f449070

Bump ci-cpu and ci-arm container versions. (#7745)

f74dac7

Update auto_tuning_with_python.py

99d68e8

fix minor formatting

[ARITH] detect iter affine map with predicate (#7752)

dc81767

[Logging] Bring back the stack size optimization (#7756)

18fa495

Make Autopad static when available (#7755)

34deb80

[VTA] Make more explicit error message during sim lib loading failure…

e8752c9

…s. (#7761)

[docs] Getting Started with TVM: Auto Tuning with Python

583c9ac

Follows up on the TVMC tutorial, shows how to accomplish the same tasks using the python api

Apply patch suggestions from black

0d05c8e

Minor corrections and updates

1cbbe31

Removed extra spaces, matched file names, lazy import of timeit.

Fix linting

ad413e7

Added link to tutorial

fc5ea26

Update auto_tuning_with_python.py

0c3abd2

fix minor formatting

Update context to device

f9b94a4

Merge branch 'tutorial-python-resnet' of github.com:hogepodge/tvm int…

5171f13

…o tutorial-python-resnet

hogepodge requested a review from yzhliu as a code owner March 29, 2021 17:44

hogepodge closed this Mar 29, 2021

hogepodge mentioned this pull request Mar 29, 2021

[docs] Getting Started with TVM: Auto Tuning with Python #7767

Merged

hogepodge deleted the tutorial-python-resnet branch June 24, 2021 23:32

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[docs] Getting Started with TVM: Auto Tuning with Python#7641

[docs] Getting Started with TVM: Auto Tuning with Python#7641
hogepodge wants to merge 62 commits intoapache:mainfrom
hogepodge:tutorial-python-resnet

hogepodge commented Mar 11, 2021

Uh oh!

hogepodge commented Mar 11, 2021

Uh oh!

leandron commented Mar 12, 2021

Uh oh!

leandron left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

leandron left a comment

Uh oh!

Uh oh!

leandron left a comment

Uh oh!

hogepodge commented Mar 29, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants

Comments

Conversation

hogepodge commented Mar 11, 2021

Uh oh!

hogepodge commented Mar 11, 2021

Uh oh!

leandron commented Mar 12, 2021

Uh oh!

leandron left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

leandron left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

leandron left a comment

Choose a reason for hiding this comment

Uh oh!

hogepodge commented Mar 29, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants

Comments