Enable tvm_op for ci #15889

yzhliu · 2019-08-14T08:31:49Z

Description

(Brief description on what this PR is about)

Checklist

Essentials

Please feel free to remove inapplicable items for your PR.

The PR title starts with [MXNET-$JIRA_ID], where $JIRA_ID refers to the relevant JIRA issue created (except PRs with tiny changes)
Changes are complete (i.e. I finished coding on this PR)
All changes have test coverage:
Unit tests are added for small changes to verify correctness (e.g. adding a new operator)
Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore)
Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL)
Code is well-documented:
For user-facing API changes, API doc string has been updated.
For new C++ functions in header files, their functionalities and arguments are documented.
For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable
Check the API doc at http://mxnet-ci-doc.s3-accelerate.dualstack.amazonaws.com/PR-$PR_ID/$BUILD_ID/index.html
To the my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change

Changes

Feature1, tests, (and when applicable, API doc)
Feature2, tests, (and when applicable, API doc)

Comments

If this change is a backward incompatible change, why must this change be made.
Interesting edge cases to note here

marcoabreu · 2019-08-14T19:18:56Z

tests/python/unittest/test_tvm_op.py

@@ -24,14 +24,13 @@

 @with_seed()
 def test_tvm_broadcast_add():
-    if _features.is_enabled("TVM_OP"):


Lets keep that if possible

yes I will, thanks for reminding. could you guide me where to pip install package in docker? ci failed with "ImportError: No module named 'decorator'" during compiling

yzhliu · 2019-08-30T18:13:45Z

@gigasquid Could you take a look at the following failure? I have no idea what does it mean.

:~/incubator-mxnet/contrib/clojure-package/examples/profiler/test$ lein test
INFO  MXNetJVM: Try loading mxnet-scala from native path.
WARN  MXNetJVM: MXNet Scala native library not found in path. Copying native library from the archive. Consider installing the library somewhere in the path (for Windows: PATH, for Linux: LD_LIBRARY_PATH), or specifying by Java cmd option -Djava.library.path=[lib path].
WARN  MXNetJVM: LD_LIBRARY_PATH=null
WARN  MXNetJVM: java.library.path=/usr/java/packages/lib/amd64:/usr/lib/x86_64-linux-gnu/jni:/lib/x86_64-linux-gnu:/usr/lib/x86_64-linux-gnu:/usr/lib/jni:/lib:/usr/lib

lein test core_test
[18:10:20] src/executor/graph_executor.cc:2025: Subgraph backend MKLDNN is activated.

lein test :only core_test/run-profiler

FAIL in (run-profiler) (core_test.clj:30)
expected: (.exists new-file)
  actual: false

lein test :only core_test/run-profiler

FAIL in (run-profiler) (core_test.clj:31)
expected: (> 10000 (- (System/currentTimeMillis) (.lastModified new-file)))
  actual: (not (> 10000 1567188620414))

Ran 1 tests containing 2 assertions.
2 failures, 0 errors.
INFO  org.apache.mxnet.util.NativeLibraryLoader: Deleting /tmp/mxnet158431027493995547/mxnet-scala
INFO  org.apache.mxnet.util.NativeLibraryLoader: Deleting /tmp/mxnet158431027493995547/libtvm_runtime.so
INFO  org.apache.mxnet.util.NativeLibraryLoader: Deleting /tmp/mxnet158431027493995547/libmkldnn.so.0
INFO  org.apache.mxnet.util.NativeLibraryLoader: Deleting /tmp/mxnet158431027493995547/libiomp5.so
INFO  org.apache.mxnet.util.NativeLibraryLoader: Deleting /tmp/mxnet158431027493995547/libmxnet.so
INFO  org.apache.mxnet.util.NativeLibraryLoader: Deleting /tmp/mxnet158431027493995547/libmklml_intel.so
INFO  org.apache.mxnet.util.NativeLibraryLoader: Deleting /tmp/mxnet158431027493995547
Tests failed.

gigasquid · 2019-08-30T22:27:05Z

@yzhliu Sure - I'll pull and take a look this weekend. The test is trying to see if the profiler is working by checking that it wrote a file. Not sure why it would fail.

yzhliu · 2019-08-30T23:12:00Z

Thanks @gigasquid I saw another PR also fails this test: #16046

gigasquid · 2019-08-31T15:31:17Z

@yzhliu I created a PR to fix the flaky test #16058. If you cherry pick it into your branch, it should resolve the issues. Please ping me if you have any other problems.

…BRARY_PATH

* enable tvm_op for ci * specify python3 bin * move rpath to top * move tvm op dep forward * add ldd debug info * add libtvm_runtime.so to mx_lib_cython * add ldd debug for py3 * fix libtvm_runtime path for cmake * cp libtvm_runtime.so when make rpkg * add libtvm_runtime.so to scala-pkg * add python3 to cmake in unix-gpu build * hack: add cuda to ld_path in cmake * add LD_LIBRARY_PATH into cmake tvm op * add /usr/local/cuda/compat to Dockerfile.build.ubuntu_gpu_cu101 LD_LIBRARY_PATH * remove unused codes * remove USE_TVM_OP from build_ubuntu_cpu_large_tensor

leezu · 2020-03-16T20:03:02Z

ci/docker/Dockerfile.build.ubuntu_gpu_cu101

@@ -80,3 +80,4 @@ RUN /work/ubuntu_adduser.sh
 COPY runtime_functions.sh /work/

 WORKDIR /work/mxnet
+ENV LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/local/cuda/compat


@yzhliu what is the purpose and need of this line?

marcoabreu reviewed Aug 14, 2019

View reviewed changes

yzhliu force-pushed the enable-tvm-op branch from a7833c9 to 74b4799 Compare August 16, 2019 08:07

roywei added the CI label Aug 19, 2019

yzhliu force-pushed the enable-tvm-op branch from 74b4799 to 8af7685 Compare August 20, 2019 04:12

yzhliu requested a review from szha as a code owner August 20, 2019 04:12

yzhliu force-pushed the enable-tvm-op branch 4 times, most recently from 041b07e to 0ee82ca Compare August 21, 2019 09:11

yzhliu requested a review from nswamy as a code owner August 21, 2019 09:11

yzhliu force-pushed the enable-tvm-op branch from 0ee82ca to c243064 Compare August 21, 2019 09:57

enable tvm_op for ci

299010b

yzhliu force-pushed the enable-tvm-op branch 3 times, most recently from 1da8878 to 51f09f9 Compare August 23, 2019 04:58

merge from upstream

995b817

yzhliu force-pushed the enable-tvm-op branch from 51f09f9 to 995b817 Compare August 23, 2019 05:59

specify python3 bin

ab31df2

yzhliu force-pushed the enable-tvm-op branch from 94f0aa0 to ab31df2 Compare August 25, 2019 22:18

move rpath to top

b885179

yzhliu force-pushed the enable-tvm-op branch from 8093fc3 to b885179 Compare August 26, 2019 01:30

move tvm op dep forward

0858222

yzhliu force-pushed the enable-tvm-op branch 3 times, most recently from 9583719 to d1d1519 Compare August 26, 2019 20:57

add ldd debug info

f1cb270

yzhliu force-pushed the enable-tvm-op branch 2 times, most recently from c4b6891 to 741443f Compare August 27, 2019 22:03

add libtvm_runtime.so to mx_lib_cython

5efc745

yzhliu force-pushed the enable-tvm-op branch from 741443f to 5efc745 Compare August 28, 2019 02:53

yzhliu force-pushed the enable-tvm-op branch from cfd5081 to 1c7df0f Compare August 28, 2019 23:14

cp libtvm_runtime.so when make rpkg

a12fde7

yzhliu force-pushed the enable-tvm-op branch from aaf6d92 to a12fde7 Compare August 29, 2019 20:18

yzhliu added 2 commits August 29, 2019 16:25

add libtvm_runtime.so to scala-pkg

9fecd90

merge from upstream

7ea6a5c

gigasquid mentioned this pull request Aug 31, 2019

Fix flaky clojure profile test #16058

Merged

4 tasks

yzhliu added 2 commits September 1, 2019 12:28

Merge remote-tracking branch 'origin/master' into enable-tvm-op

6a2b752

add python3 to cmake in unix-gpu build

d4ce2e4

yzhliu force-pushed the enable-tvm-op branch from 18b9545 to 8112529 Compare September 3, 2019 05:31

hack: add cuda to ld_path in cmake

5f3ea79

yzhliu force-pushed the enable-tvm-op branch 3 times, most recently from 1cc199c to e64c8ec Compare September 3, 2019 20:09

add LD_LIBRARY_PATH into cmake tvm op

1a5ae86

yzhliu force-pushed the enable-tvm-op branch from e64c8ec to 1a5ae86 Compare September 4, 2019 04:21

yzhliu added 3 commits September 4, 2019 00:14

add /usr/local/cuda/compat to Dockerfile.build.ubuntu_gpu_cu101 LD_LI…

8a44b72

…BRARY_PATH

remove unused codes

ad396ea

Merge remote-tracking branch 'origin/master' into enable-tvm-op

b2c3871

yzhliu changed the title ~~[DO NOT MERGE] enable tvm_op for ci~~ Enable tvm_op for ci Sep 4, 2019

yzhliu added 2 commits September 4, 2019 14:20

remove USE_TVM_OP from build_ubuntu_cpu_large_tensor

a06af1e

Merge remote-tracking branch 'origin/master' into enable-tvm-op

de26970

reminisce approved these changes Sep 5, 2019

View reviewed changes

reminisce merged commit b7071c4 into apache:master Sep 5, 2019

leezu reviewed Mar 16, 2020

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Enable tvm_op for ci #15889

Enable tvm_op for ci #15889

yzhliu commented Aug 14, 2019

marcoabreu Aug 14, 2019

yzhliu Aug 16, 2019 •

edited

Loading

yzhliu commented Aug 30, 2019

gigasquid commented Aug 30, 2019

yzhliu commented Aug 30, 2019 •

edited

Loading

gigasquid commented Aug 31, 2019

leezu Mar 16, 2020

Enable tvm_op for ci #15889

Enable tvm_op for ci #15889

Conversation

yzhliu commented Aug 14, 2019

Description

Checklist

Essentials

Changes

Comments

marcoabreu Aug 14, 2019

Choose a reason for hiding this comment

yzhliu Aug 16, 2019 • edited Loading

Choose a reason for hiding this comment

yzhliu commented Aug 30, 2019

gigasquid commented Aug 30, 2019

yzhliu commented Aug 30, 2019 • edited Loading

gigasquid commented Aug 31, 2019

leezu Mar 16, 2020

Choose a reason for hiding this comment

yzhliu Aug 16, 2019 •

edited

Loading

yzhliu commented Aug 30, 2019 •

edited

Loading