anirudh2290 commented on a change in pull request #11320: Static alloc for
hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#discussion_r195918965
##
File path: tests/python/unittest/test_gluon.py
##
@@ -1137,6 +1137,60 @@ def test_zero_grad():
eric-haibin-lin closed issue #10947: Flaky test:
test_tutorials.test_onnx_inference_on_onnx_model
URL: https://github.com/apache/incubator-mxnet/issues/10947
This is an automated message from the Apache Git Service.
To
ThomasDelteil commented on a change in pull request #11320: Static alloc for
hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#discussion_r195918147
##
File path: tests/python/unittest/test_gluon.py
##
@@ -1137,6 +1137,60 @@ def test_zero_grad():
ThomasDelteil commented on issue #10947: Flaky test:
test_tutorials.test_onnx_inference_on_onnx_model
URL:
https://github.com/apache/incubator-mxnet/issues/10947#issuecomment-397854557
@eric-haibin-lin this should be solved by
https://github.com/apache/incubator-mxnet/pull/11181 please
ThomasDelteil commented on issue #9669: Gateway timeout when downloading in
test_contrib_text.test_download_embed
URL:
https://github.com/apache/incubator-mxnet/issues/9669#issuecomment-397854475
@KellenSunderland this should be solved with
asitstands commented on issue #10951: [MXNET-545] Fix broken cython build
URL: https://github.com/apache/incubator-mxnet/pull/10951#issuecomment-397175348
```bash
make
make cython
cd python
# `pip install -e .` or
export PYTHONPATH=my_installation_dir
python setup.py
yajiedesign commented on issue #11100: Installation instructions for Windows
don't work
URL:
https://github.com/apache/incubator-mxnet/issues/11100#issuecomment-397852731
download base package in https://github.com/yajiedesign/mxnet/releases
yajiedesign commented on issue #11100: Installation instructions for Windows
don't work
URL:
https://github.com/apache/incubator-mxnet/issues/11100#issuecomment-397852731
https://github.com/yajiedesign/mxnet/releases base package
ThomasDelteil commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397852586
Thanks @szhengac for testing out the optimisation and glad to see it is
bringing such drastic improvements on multi-gpu variable-sized
yajiedesign commented on issue #11321: Failure to build for Windows (CUDA or
MKL-DNN)
URL:
https://github.com/apache/incubator-mxnet/issues/11321#issuecomment-397852559
do you install cuda visual studio integration?
This is
asitstands commented on issue #10951: [MXNET-545] Fix broken cython build
URL: https://github.com/apache/incubator-mxnet/pull/10951#issuecomment-397852430
I added the cython build to Python3 CentOS 7 CPU and GPU environments. I
have made an intentional bug in the cython module and
asitstands commented on issue #10951: [MXNET-545] Fix broken cython build
URL: https://github.com/apache/incubator-mxnet/pull/10951#issuecomment-397852430
I added the cython build to Python3 CentOS 7 CPU and GPU environments. I
have made an intentional bug in the cython module and
asitstands commented on issue #10951: [MXNET-545] Fix broken cython build
URL: https://github.com/apache/incubator-mxnet/pull/10951#issuecomment-397852430
I added the cython build to Python3 CentOS 7 CPU and GPU environments. I
have made an intentional bug in the cython module and
pengzhao-intel commented on issue #10835: Lazy sparse param initialization for
distributed training
URL:
https://github.com/apache/incubator-mxnet/issues/10835#issuecomment-397851719
mark to come back soon
This is an
TaoLv commented on a change in pull request #11262: [MXNET-542] Fix mkldnn
performance regression + improve test logging
URL: https://github.com/apache/incubator-mxnet/pull/11262#discussion_r195916774
##
File path: src/operator/nn/mkldnn/mkldnn_base.cc
##
@@ -77,26 +77,35
anirudh2290 commented on a change in pull request #11320: Static alloc for
hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#discussion_r195916761
##
File path: tests/python/unittest/test_gluon.py
##
@@ -1137,6 +1137,60 @@ def test_zero_grad():
TaoLv commented on a change in pull request #11262: [MXNET-542] Fix mkldnn
performance regression + improve test logging
URL: https://github.com/apache/incubator-mxnet/pull/11262#discussion_r195916640
##
File path: src/operator/nn/mkldnn/mkldnn_base.cc
##
@@ -77,26 +77,35
szha commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397850582
Thanks, @szhengac
This is an automated message from the Apache Git Service.
To
haojin2 commented on issue #11021: [MXNET-380] count_include_pad argument for
Avg Pooling
URL: https://github.com/apache/incubator-mxnet/pull/11021#issuecomment-397850554
@TaoLv All previous tests were still performed even with those changes. If I
do separate the tests, wouldn't it be a
TaoLv commented on issue #11021: [MXNET-380] count_include_pad argument for Avg
Pooling
URL: https://github.com/apache/incubator-mxnet/pull/11021#issuecomment-397850228
I feel like adding new tests for this change would be better, compared with
changing the existing ones.
This is an automated email from the ASF dual-hosted git repository.
zhasheng pushed a commit to branch asf-site
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet-site.git
The following commit(s) were added to refs/heads/asf-site by this push:
new 2975e99 Bump the publish
szha commented on a change in pull request #11090: Define build target for
mkldnn lib build to fix 'make clean USE_MKMLDNN=1' issue
URL: https://github.com/apache/incubator-mxnet/pull/11090#discussion_r195915423
##
File path: mkldnn.mk
##
@@ -0,0 +1,66 @@
+# Licensed to
zheng-da commented on a change in pull request #10451: [MXNET-432] Add Foreach
URL: https://github.com/apache/incubator-mxnet/pull/10451#discussion_r195915415
##
File path: src/operator/control_flow.cc
##
@@ -0,0 +1,409 @@
+/*
+ * Licensed to the Apache Software Foundation
eric-haibin-lin commented on issue #10835: Lazy sparse param initialization for
distributed training
URL:
https://github.com/apache/incubator-mxnet/issues/10835#issuecomment-397847014
I'm not aware if anyone has spare time to work on this. If you'd like to
contribute, I'm happy to
zheng-da commented on a change in pull request #10451: [MXNET-432] Add Foreach
URL: https://github.com/apache/incubator-mxnet/pull/10451#discussion_r195915356
##
File path: src/operator/subgraph_op_common.cc
##
@@ -0,0 +1,250 @@
+/*
+ * Licensed to the Apache Software
casscw opened a new issue #10835: Lazy sparse param initialization for
distributed training
URL: https://github.com/apache/incubator-mxnet/issues/10835
## Description
In the model wide,**a categorical feature 'sex' has three values
'0,1,2'.** Then feed it into SparseEmbedding layer,
zheng-da commented on a change in pull request #10451: [MXNET-432] Add Foreach
URL: https://github.com/apache/incubator-mxnet/pull/10451#discussion_r195915304
##
File path: src/operator/subgraph_op_common.h
##
@@ -0,0 +1,104 @@
+/*
+ * Licensed to the Apache Software
zheng-da commented on a change in pull request #10451: [MXNET-432] Add Foreach
URL: https://github.com/apache/incubator-mxnet/pull/10451#discussion_r195915287
##
File path: python/mxnet/symbol/contrib.py
##
@@ -91,3 +98,154 @@ def rand_zipfian(true_classes, num_sampled,
szha commented on issue #11090: Define build target for mkldnn lib build to fix
'make clean USE_MKMLDNN=1' issue
URL: https://github.com/apache/incubator-mxnet/pull/11090#issuecomment-397846800
@ashokei this procedure isn't documented anywhere, so whoever is aware of
this approach would
eric-haibin-lin commented on issue #11299: row_sparse_pull,push row_sparse
gradient is too slow,it has 10+ times difference
URL:
https://github.com/apache/incubator-mxnet/issues/11299#issuecomment-397846747
For minibatch training usually you derive rowids from the sparse data in the
eric-haibin-lin commented on a change in pull request #11316:
MXNET_FORCE_ADDTAKEGRAD to disable AddTakeGradLargeBatchCaller
URL: https://github.com/apache/incubator-mxnet/pull/11316#discussion_r195915212
##
File path: src/operator/tensor/indexing_op.h
##
@@ -598,7
eric-haibin-lin commented on a change in pull request #11316:
MXNET_FORCE_ADDTAKEGRAD to disable AddTakeGradLargeBatchCaller
URL: https://github.com/apache/incubator-mxnet/pull/11316#discussion_r195915212
##
File path: src/operator/tensor/indexing_op.h
##
@@ -598,7
szhengac commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397846636
In my experiment on gluon, this static_alloc option is very important for
inputs with variable shapes. Without setting this to true, I cannot
szha commented on issue #11304: Added Learning Rate Finder tutorial
URL: https://github.com/apache/incubator-mxnet/pull/11304#issuecomment-397846476
@aaronmarkham would you provide some input on the writing? @astonzhang could
you help review the code when you have time?
This is an automated email from the ASF dual-hosted git repository.
zhasheng pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git
The following commit(s) were added to refs/heads/master by this push:
new 7210c2c bump up rtol for fp16 case
szha closed pull request #11246: [MXNET-453] Fix flaky test_sgd
URL: https://github.com/apache/incubator-mxnet/pull/11246
This is a PR merged from a forked repository.
As GitHub hides the original diff on merge, it is displayed below for
the sake of provenance:
As this is a foreign pull
sandeep-krishnamurthy commented on a change in pull request #11274: [MXNET-547]
Tutorial explaining how to use the profiler
URL: https://github.com/apache/incubator-mxnet/pull/11274#discussion_r195915012
##
File path: docs/tutorials/python/profiler.md
##
@@ -0,0
szha commented on a change in pull request #11316: MXNET_FORCE_ADDTAKEGRAD to
disable AddTakeGradLargeBatchCaller
URL: https://github.com/apache/incubator-mxnet/pull/11316#discussion_r195914883
##
File path: src/operator/tensor/indexing_op.h
##
@@ -598,7 +598,12 @@ void
szha commented on a change in pull request #11317: Fix the inaccruate
instruction on installing nose
URL: https://github.com/apache/incubator-mxnet/pull/11317#discussion_r195914841
##
File path: python/README.md
##
@@ -10,12 +10,12 @@ To install MXNet Python package,
szha commented on issue #11307: Memory Management
URL:
https://github.com/apache/incubator-mxnet/issues/11307#issuecomment-397845251
weak ref doesn't really help pushing GC to work, unfortunately.
This is an automated
marcoabreu commented on issue #10451: [MXNET-432] Add Foreach
URL: https://github.com/apache/incubator-mxnet/pull/10451#issuecomment-397845037
Thanks a lot for adding a benchmark!
This is an automated message from the Apache
szha commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397844635
@szhengac could you comment on the speed improvement of static memory
allocation in your transformer models?
zheng-da commented on a change in pull request #11262: [MXNET-542] Fix mkldnn
performance regression + improve test logging
URL: https://github.com/apache/incubator-mxnet/pull/11262#discussion_r195909929
##
File path: src/operator/nn/mkldnn/mkldnn_base-inl.h
##
@@ -327,7
zheng-da commented on issue #10451: [MXNET-432] Add Foreach
URL: https://github.com/apache/incubator-mxnet/pull/10451#issuecomment-397843936
benchmark results on m4.4xlarge.
```
| | Batch size: 1 (second) | Batch size: 32 (second) |
| |
zheng-da commented on issue #10451: [MXNET-432] Add Foreach
URL: https://github.com/apache/incubator-mxnet/pull/10451#issuecomment-397843936
```
| | Batch size: 1 (second) | Batch size: 32 (second) |
| | -- |
zheng-da commented on issue #10451: [MXNET-432] Add Foreach
URL: https://github.com/apache/incubator-mxnet/pull/10451#issuecomment-397843936
```
Batch size: 1 (second) | Batch size: 32 (second)
-- | --
Imperative inference | 0.61 | 0.64
Hybrid inference | 0.21 | 0.23
marcoabreu commented on issue #11302: [MXNET-553] Restructure dockcross
dockerfiles to fix caching
URL: https://github.com/apache/incubator-mxnet/pull/11302#issuecomment-397841350
That's an excellent idea! Maybe just printing the cache hit/miss statistics
after a build (not globally using
KellenSunderland commented on issue #11302: [MXNET-553] Restructure dockcross
dockerfiles to fix caching
URL: https://github.com/apache/incubator-mxnet/pull/11302#issuecomment-397841274
@marcoabreu Gotcha. I think what would be useful would be to include some
more information about what
zheng-da commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397840627
@ThomasDelteil thanks for your benchmark script. I modified your script to
exclude the first epoch and can verify that the fix in
zheng-da commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397840627
@ThomasDelteil thanks for your benchmark script. I modified your script to
exclude the first epoch and can verify that the fix in
ThomasDelteil commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397840263
- agreed that MNIST might not be representative of real-world workload. I
have left a commented out "image.resize_short" if we feel like
ThomasDelteil commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397840263
- agreed that MNIST might not be representative of real-world workload. I
have left a commented out "image.resize_short" if we feel like
ThomasDelteil commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397840263
- agreed that MNIST might not be representative of real-world workload. I
have left a commented out "image.resize_short" if we feel like
tqchen commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397839131
@ThomasDelteil a few things thoughts, please feel free to comment
- Bench-marking against mnist is a good reference but may not reflect
KellenSunderland commented on a change in pull request #11302: [MXNET-553]
Restructure dockcross dockerfiles to fix caching
URL: https://github.com/apache/incubator-mxnet/pull/11302#discussion_r195911652
##
File path: ci/docker/install/deb_ubuntu_ccache.sh
##
@@ -43,7
absalama commented on issue #11200: Training Imagenet on AlexNet failed with
corrupted JPEG format images
URL:
https://github.com/apache/incubator-mxnet/issues/11200#issuecomment-397834739
@sandeep-krishnamurthy
I figured out where the problem comes from. Streaming huge record
gigasquid commented on issue #11205: Clojure Contrib Package
URL: https://github.com/apache/incubator-mxnet/pull/11205#issuecomment-397834692
Thanks for taking the time to review @larroy I'll get to work on
incorporating your feedback.
gigasquid commented on a change in pull request #11205: Clojure Contrib Package
URL: https://github.com/apache/incubator-mxnet/pull/11205#discussion_r195910572
##
File path: contrib/clojure-package/README.md
##
@@ -0,0 +1,197 @@
+# Clojure MXNet
+
+A clojure package to the
ThomasDelteil commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397833059
@tqchen, thanks I will explicitly hit "Request Change" next time.
Here is an easy to reproduce benchmark script
zheng-da commented on a change in pull request #11262: [MXNET-542] Fix mkldnn
performance regression + improve test logging
URL: https://github.com/apache/incubator-mxnet/pull/11262#discussion_r195909929
##
File path: src/operator/nn/mkldnn/mkldnn_base-inl.h
##
@@ -327,7
zheng-da commented on a change in pull request #11262: [MXNET-542] Fix mkldnn
performance regression + improve test logging
URL: https://github.com/apache/incubator-mxnet/pull/11262#discussion_r195910062
##
File path: src/operator/nn/mkldnn/mkldnn_base.cc
##
@@ -77,26
tqchen commented on a change in pull request #11320: Static alloc for
hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#discussion_r195909156
##
File path: python/mxnet/gluon/block.py
##
@@ -502,8 +502,16 @@ def hybridize(self, active=True, **kwargs):
tqchen commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397830554
I think the specific comment of @ThomasDelteil on the documents are
actionable and should be addressed. The time difference in terms of resnet50,
marcoabreu commented on issue #11302: [MXNET-553] Restructure dockcross
dockerfiles to fix caching
URL: https://github.com/apache/incubator-mxnet/pull/11302#issuecomment-397830325
Yes, I got the feeling that your changes are causing some builds to not use
ccache anymore. It'd be good if
ThomasDelteil commented on a change in pull request #11320: Static alloc for
hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#discussion_r195908700
##
File path: python/mxnet/gluon/block.py
##
@@ -502,8 +502,16 @@ def hybridize(self, active=True,
ThomasDelteil commented on a change in pull request #11320: Static alloc for
hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#discussion_r195908757
##
File path: python/mxnet/gluon/block.py
##
@@ -502,8 +502,16 @@ def hybridize(self, active=True,
tqchen commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397828737
@ThomasDelteil can you copy your list of concerns over and make comments on
the code? It is good to comment on the code and request changes, so
tqchen commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397828737
@ThomasDelteil can you copy your list of concerns over and make comments on
the code?
azai91 commented on a change in pull request #11262: [MXNET-542] Fix mkldnn
performance regression + improve test logging
URL: https://github.com/apache/incubator-mxnet/pull/11262#discussion_r195908349
##
File path: src/operator/nn/mkldnn/mkldnn_base.cc
##
@@ -77,26
azai91 commented on a change in pull request #11262: [MXNET-542] Fix mkldnn
performance regression + improve test logging
URL: https://github.com/apache/incubator-mxnet/pull/11262#discussion_r195908368
##
File path: src/operator/nn/mkldnn/mkldnn_base.cc
##
@@ -77,26
azai91 commented on a change in pull request #11262: [MXNET-542] Fix mkldnn
performance regression + improve test logging
URL: https://github.com/apache/incubator-mxnet/pull/11262#discussion_r195908291
##
File path: src/operator/nn/mkldnn/mkldnn_base.cc
##
@@ -77,26
azai91 commented on a change in pull request #11262: [MXNET-542] Fix mkldnn
performance regression + improve test logging
URL: https://github.com/apache/incubator-mxnet/pull/11262#discussion_r195908109
##
File path: src/operator/nn/mkldnn/mkldnn_base.cc
##
@@ -77,26
ThomasDelteil commented on issue #11320: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11320#issuecomment-397824212
@piiswrong could you please respond to the concerns that I listed there:
https://github.com/apache/incubator-mxnet/pull/11318 ?
I
aluo-x opened a new issue #11321: Failure to build for Windows (CUDA or MKL-DNN)
URL: https://github.com/apache/incubator-mxnet/issues/11321
Currently attempting to build on Windows 1803 with visual studio 2017
(15.6.7).
Also present on the device is CUDA 9.2.88 and 9.2.88.1, along
larroy commented on a change in pull request #11205: Clojure Contrib Package
URL: https://github.com/apache/incubator-mxnet/pull/11205#discussion_r195905637
##
File path:
contrib/clojure-package/examples/pre-trained-models/download-caltech.sh
##
@@ -0,0 +1,25 @@
larroy commented on a change in pull request #11205: Clojure Contrib Package
URL: https://github.com/apache/incubator-mxnet/pull/11205#discussion_r195905634
##
File path:
contrib/clojure-package/examples/pre-trained-models/download-resnet-152.sh
##
@@ -0,0 +1,27 @@
larroy commented on a change in pull request #11205: Clojure Contrib Package
URL: https://github.com/apache/incubator-mxnet/pull/11205#discussion_r195905544
##
File path: contrib/clojure-package/README.md
##
@@ -0,0 +1,197 @@
+# Clojure MXNet
+
+A clojure package to the
larroy commented on a change in pull request #11205: Clojure Contrib Package
URL: https://github.com/apache/incubator-mxnet/pull/11205#discussion_r195905489
##
File path: contrib/clojure-package/examples/cnn-text-classification/get_data.sh
##
@@ -0,0 +1,29 @@
+#!/bin/bash
larroy commented on a change in pull request #11302: [MXNET-553] Restructure
dockcross dockerfiles to fix caching
URL: https://github.com/apache/incubator-mxnet/pull/11302#discussion_r195772709
##
File path: ci/docker/install/deb_ubuntu_ccache.sh
##
@@ -43,7 +43,9 @@ git
KellenSunderland commented on issue #11302: [MXNET-553] Restructure dockcross
dockerfiles to fix caching
URL: https://github.com/apache/incubator-mxnet/pull/11302#issuecomment-397803518
@larroy Sounds good, we can also chat offline about if it in more detail if
I've misunderstood.
marcoabreu commented on issue #11319: Static alloc for hybridblock
URL: https://github.com/apache/incubator-mxnet/pull/11319#issuecomment-397797110
We will prioritize to get this PR back in as soon as possible. Sorry for the
caused inconveniences.
BenBBear commented on issue #8392: symbol/operator type feature request
URL:
https://github.com/apache/incubator-mxnet/issues/8392#issuecomment-397795979
```python
dct = json.loads(sym.tojson())
node = dct['nodes'][-1]
print(node['op'])
```
hope it helps, inefficient but
82 matches
Mail list logo