dcslin opened a new pull request #548: DNNL(upgrade of MKLDNN) integration and
add softmax
URL: https://github.com/apache/incubator-singa/pull/548
This is an automated message from the Apache Git Service.
To respond to the
chrishkchris opened a new pull request #547: SINGA-493 Add an additional build
path
URL: https://github.com/apache/incubator-singa/pull/547
Add an webpage build path in build.sh
This is an automated message from the Apache
dcslin edited a comment on issue #541: added 4d test on batchnorm
URL: https://github.com/apache/incubator-singa/pull/541#issuecomment-540867039
HI @joddiy , could you share how did you call singa api? the test you
mentioned is tested in here
dcslin commented on issue #541: added 4d test on batchnorm
URL: https://github.com/apache/incubator-singa/pull/541#issuecomment-540867039
HI @joddiy , could you share how did you call singa api? the test you
mentioned is tested in here
joddiy edited a comment on issue #541: added 4d test on batchnorm
URL: https://github.com/apache/incubator-singa/pull/541#issuecomment-540600410
hi, shicong, the result still has a little error, please check with the
following case:
```
def _batchnorm_test_mode(x, s, bias, mean, var,
joddiy commented on issue #541: added 4d test on batchnorm
URL: https://github.com/apache/incubator-singa/pull/541#issuecomment-540600410
hi, shicong, the result still has a little error, please check with the
following case:
```
def _batchnorm_test_mode(x, s, bias, mean,
joddiy commented on issue #540: added softmax with axis
URL: https://github.com/apache/incubator-singa/pull/540#issuecomment-540567614
ready to merge
This is an automated message from the Apache Git Service.
To respond to the
dcslin closed pull request #538: [WIP] add Cudnn mode as optional param to
batch norm and softmax
URL: https://github.com/apache/incubator-singa/pull/538
This is an automated message from the Apache Git Service.
To respond
dcslin commented on issue #538: [WIP] add Cudnn mode as optional param to batch
norm and softmax
URL: https://github.com/apache/incubator-singa/pull/538#issuecomment-540511768
split to https://github.com/apache/incubator-singa/pull/540 and
nudles merged pull request #1: SINGA-493 Update SINGA website
URL: https://github.com/apache/incubator-singa-site/pull/1
This is an automated message from the Apache Git Service.
To respond to the message, please log on to
nudles merged pull request #546: SINGA-491 Remaining LGTM Python Code Cleaning
URL: https://github.com/apache/incubator-singa/pull/546
This is an automated message from the Apache Git Service.
To respond to the message,
chrishkchris opened a new pull request #546: SINGA-491 Remaining LGTM Python
Code Cleaning
URL: https://github.com/apache/incubator-singa/pull/546
This is to clean some remaining LGTM Python code alerts that I am capable to
clean. This is the last PR for my current code cleaning, because
dcslin commented on issue #540: added softmax with axis
URL: https://github.com/apache/incubator-singa/pull/540#issuecomment-539904174
Hi @joddiy, thank you for the review. kindly find the updated version which
fixed the mentioned issue.
dcslin commented on issue #540: added softmax with axis
URL: https://github.com/apache/incubator-singa/pull/540#issuecomment-539628035
> > > still has some problems, the output of multiple dimension inputs is
not correct.
> > > please check:
> > > ```
> > > x_0 = np.array([[0, 1,
dcslin commented on issue #541: added 4d test on batchnorm
URL: https://github.com/apache/incubator-singa/pull/541#issuecomment-539449929
HI @joddiy , could you please help to review this test on batchnorm value?
which also showing that the required params/output by ONNX are given.
joddiy opened a new pull request #545: SINGA-494 Singa autograd improvement
URL: https://github.com/apache/incubator-singa/pull/545
## Background: some autograd ops cannot satisfy the onnx demand, as
following:
### conv, averagepool, maxpool
- only support 2d input, i.e,
joddiy commented on issue #540: added softmax with axis
URL: https://github.com/apache/incubator-singa/pull/540#issuecomment-538984488
> > still has some problems, the output of multiple dimension inputs is not
correct.
> > please check:
> > ```
> > x_0 = np.array([[0, 1, 2, 3],
dcslin commented on issue #540: added softmax with axis
URL: https://github.com/apache/incubator-singa/pull/540#issuecomment-538955169
> still has some problems, the output of multiple dimension inputs is not
correct.
> please check:
>
> ```
> x_0 = np.array([[0, 1, 2, 3],
chrishkchris opened a new pull request #1: SINGA-493 Update SINGA website
URL: https://github.com/apache/incubator-singa-site/pull/1
This is an automated message from the Apache Git Service.
To respond to the message, please
nudles merged pull request #544: SINGA-493 Update news on SINGA tutorial
URL: https://github.com/apache/incubator-singa/pull/544
This is an automated message from the Apache Git Service.
To respond to the message, please log
chrishkchris edited a comment on issue #544: SINGA-493 Update news on SINGA
tutorial
URL: https://github.com/apache/incubator-singa/pull/544#issuecomment-538370010
I forked the incubator-singa-site and put the preview of website at
chrishkchris commented on issue #544: SINGA-493 Update news on SINGA tutorial
URL: https://github.com/apache/incubator-singa/pull/544#issuecomment-538370010
I forked the incubator-singa-site and put the preview of website at
chrishkchris opened a new pull request #544: SINGA-493 Update news on singa
tutorial
URL: https://github.com/apache/incubator-singa/pull/544
The webpage look like this when I use Sphinx to build:
joddiy commented on issue #541: added 4d test on batchnorm
URL: https://github.com/apache/incubator-singa/pull/541#issuecomment-538211845
the return of _CpuBatchNormForwardInference_ should be y, var, mean, not
just y.
could you fix it?
joddiy commented on issue #540: added softmax with axis
URL: https://github.com/apache/incubator-singa/pull/540#issuecomment-538211628
still has some problems, the output of multiple dimension inputs is not
correct.
please check:
```
x_0 = np.array([[0, 1, 2, 3], [1, 10001,
nudles merged pull request #543: SINGA-491 Use const reference for CopyData and
ResetLike from Tensor Input
URL: https://github.com/apache/incubator-singa/pull/543
This is an automated message from the Apache Git Service.
chrishkchris opened a new pull request #543: SINGA-491 Use const reference when
we use CopyData and ResetLike from Tensor Input
URL: https://github.com/apache/incubator-singa/pull/543
This is to fix the remaining C language alert from LGTM:
There are some functions using ResetLike
lzjpaul opened a new pull request #542: SINGA-492 add Food(lg) news in the
documentation
URL: https://github.com/apache/incubator-singa/pull/542
-add Food(lg) news in the SINGA homepage
This is an automated message from the
dcslin opened a new pull request #541: added 4d test on batchnorm
URL: https://github.com/apache/incubator-singa/pull/541
added batchnorm test adapted from `/test/python/test_onnx_backend.py` and
test passed.
Hi @joddiy , If current Batchnorm API could fulfil ONNX batchnorm
dcslin opened a new pull request #540: added softmax with axis
URL: https://github.com/apache/incubator-singa/pull/540
splitting https://github.com/apache/incubator-singa/pull/538 into two parts,
this PR is for softmax with axis, which is compatible to onnx api
nudles merged pull request #539: SINGA-491 Python Code Cleaning
URL: https://github.com/apache/incubator-singa/pull/539
This is an automated message from the Apache Git Service.
To respond to the message, please log on to
chrishkchris opened a new pull request #539: SINGA-491 Python Code Cleaning
URL: https://github.com/apache/incubator-singa/pull/539
Only Python code is cleaned at this time that is likely to remove around
hundred alerts in python code.
Since there are lots of changes in the folders
nudles merged pull request #537: SINGA-488 Change the path of source code for CI
URL: https://github.com/apache/incubator-singa/pull/537
This is an automated message from the Apache Git Service.
To respond to the message,
dcslin opened a new pull request #538: add Cudnn mode as optional param to
batch norm and softmax
URL: https://github.com/apache/incubator-singa/pull/538
This is an automated message from the Apache Git Service.
To respond
chrishkchris opened a new pull request #537: SINGA-488 Change the path of
source code for CI
URL: https://github.com/apache/incubator-singa/pull/537
This will fix the issue in "SINGA-488 Travis CI always build from Apache
master branch"
nudles merged pull request #528: SINGA 475 - add logical operator: and, or,
xor, not & negative, reciprocal
URL: https://github.com/apache/incubator-singa/pull/528
This is an automated message from the Apache Git Service.
nudles merged pull request #536: SINGA-491 Code Cleaning with the Reference of
LGTM Analysis Result
URL: https://github.com/apache/incubator-singa/pull/536
This is an automated message from the Apache Git Service.
To
dcslin commented on issue #514: SINGA-482 tc comprehension integration
URL: https://github.com/apache/incubator-singa/pull/514#issuecomment-534047644
Integration is done except support CPU.
CPU support development is done, however the behavior was not expected even
in clean TC env.
chrishkchris commented on issue #536: SINGA-491 Code Cleaning with the
Reference of LGTM Analysis Result
URL: https://github.com/apache/incubator-singa/pull/536#issuecomment-533585944
Thanks so much for your comments. It is a great idea concerning the CI.
moazreyad commented on issue #536: SINGA-491 Code Cleaning with the Reference
of LGTM Analysis Result
URL: https://github.com/apache/incubator-singa/pull/536#issuecomment-533497942
Thank you.
In future we should not accept only the compile output alone, we should also
run the unit
chrishkchris opened a new pull request #536: SINGA-491 Code Cleaning with the
Reference of LGTM Analysis Result
URL: https://github.com/apache/incubator-singa/pull/536
Since LGTM has been applied for our code analysis (see SINGA-484), I have
cleaned up many of the obvious issues alerted
nudles merged pull request #529: SINGA-484 Code analysis with LGTM
URL: https://github.com/apache/incubator-singa/pull/529
This is an automated message from the Apache Git Service.
To respond to the message, please log on to
nudles merged pull request #533: fixed operation add assertion for convolution
URL: https://github.com/apache/incubator-singa/pull/533
This is an automated message from the Apache Git Service.
To respond to the message,
nudles merged pull request #535: SINGA-490 Optimize performance of stochastic
gradient descent (SGD)
URL: https://github.com/apache/incubator-singa/pull/535
This is an automated message from the Apache Git Service.
To
dcslin commented on issue #533: fixed operation add assertion for convolution
URL: https://github.com/apache/incubator-singa/pull/533#issuecomment-533383280
Hi @nudles , this is ready for merge.
thanks to @chrishkchris
chrishkchris commented on a change in pull request #535: SINGA-490 Optimize
performance of stochastic gradient descent (SGD)
URL: https://github.com/apache/incubator-singa/pull/535#discussion_r326004398
##
File path: src/core/tensor/tensor_math_cuda.h
##
@@ -324,12 +324,8
chrishkchris commented on a change in pull request #535: SINGA-490 Optimize
performance of stochastic gradient descent (SGD)
URL: https://github.com/apache/incubator-singa/pull/535#discussion_r326006074
##
File path: src/core/tensor/tensor_math_cuda.h
##
@@ -324,12 +324,8
chrishkchris commented on a change in pull request #535: SINGA-490 Optimize
performance of stochastic gradient descent (SGD)
URL: https://github.com/apache/incubator-singa/pull/535#discussion_r326004398
##
File path: src/core/tensor/tensor_math_cuda.h
##
@@ -324,12 +324,8
nudles commented on a change in pull request #535: SINGA-490 Optimize
performance of stochastic gradient descent (SGD)
URL: https://github.com/apache/incubator-singa/pull/535#discussion_r326002562
##
File path: src/core/tensor/tensor_math_cuda.h
##
@@ -324,12 +324,8 @@
chrishkchris edited a comment on issue #535: SINGA-490 Optimize performance of
stochastic gradient descent (SGD)
URL: https://github.com/apache/incubator-singa/pull/535#issuecomment-532944885
Since I have done some optimization of the framework, I need to retest again
the distributed
chrishkchris commented on issue #535: SINGA-490 Optimize performance of
stochastic gradient descent (SGD)
URL: https://github.com/apache/incubator-singa/pull/535#issuecomment-532944885
Finally, I test the distributed training in AWS p2.x8large, after adding the
Sync() in the SGD loop of
chrishkchris edited a comment on issue #535: SINGA-490 Optimize performance of
stochastic gradient descent (SGD)
URL: https://github.com/apache/incubator-singa/pull/535#issuecomment-532723250
Next, I further improve the time performance by using in-place elementwise
multiply in many
chrishkchris commented on issue #535: SINGA-490 Optimize performance of
stochastic gradient descent (SGD)
URL: https://github.com/apache/incubator-singa/pull/535#issuecomment-532723250
Next, I further improve the time performance by using in-place elementwise
multiply in many functions
chrishkchris opened a new pull request #535: SINGA-490 Optimize performance of
stochastic gradient descent (SGD)
URL: https://github.com/apache/incubator-singa/pull/535
I have fused the small operations of momentum SGD so as to increase GPU
computation efficiency and decrease latency.
pinpom commented on a change in pull request #528: SINGA 475 - add logical
operator: and, or, xor, not & negative, reciprocal
URL: https://github.com/apache/incubator-singa/pull/528#discussion_r325471648
##
File path: python/singa/autograd.py
##
@@ -2522,3 +2523,144 @@
joddiy commented on a change in pull request #528: SINGA 475 - add logical
operator: and, or, xor, not & negative, reciprocal
URL: https://github.com/apache/incubator-singa/pull/528#discussion_r325462186
##
File path: python/singa/autograd.py
##
@@ -2522,3 +2523,144 @@
pinpom commented on a change in pull request #528: SINGA 475 - add logical
operator: and, or, xor, not & negative, reciprocal
URL: https://github.com/apache/incubator-singa/pull/528#discussion_r325461738
##
File path: python/singa/autograd.py
##
@@ -2522,3 +2523,144 @@
joddiy commented on a change in pull request #528: SINGA 475 - add logical
operator: and, or, xor, not & negative, reciprocal
URL: https://github.com/apache/incubator-singa/pull/528#discussion_r325123236
##
File path: python/singa/autograd.py
##
@@ -2522,3 +2523,144 @@
joddiy commented on a change in pull request #528: SINGA 475 - add logical
operator: and, or, xor, not & negative, reciprocal
URL: https://github.com/apache/incubator-singa/pull/528#discussion_r325107434
##
File path: python/singa/autograd.py
##
@@ -2522,3 +2523,144 @@
joddiy commented on a change in pull request #528: SINGA 475 - add logical
operator: and, or, xor, not & negative, reciprocal
URL: https://github.com/apache/incubator-singa/pull/528#discussion_r325108637
##
File path: python/singa/autograd.py
##
@@ -2522,3 +2523,144 @@
dcslin commented on a change in pull request #533: fixed operation add
assertion for convolution
URL: https://github.com/apache/incubator-singa/pull/533#discussion_r324502113
##
File path: src/core/tensor/tensor_math_cpp.h
##
@@ -868,6 +868,25 @@ void RowMax(const Tensor&
dcslin commented on a change in pull request #533: fixed operation add
assertion for convolution
URL: https://github.com/apache/incubator-singa/pull/533#discussion_r324502113
##
File path: src/core/tensor/tensor_math_cpp.h
##
@@ -868,6 +868,25 @@ void RowMax(const Tensor&
joddiy commented on issue #462: SINGA-333 Add ONNX backend test
URL: https://github.com/apache/incubator-singa/pull/462#issuecomment-531490830
> But if we can't run onnx backend test cases, then we simply can't claim
that we support onnx backend.
>
> We don't have to pass all the
moazreyad commented on issue #462: SINGA-333 Add ONNX backend test
URL: https://github.com/apache/incubator-singa/pull/462#issuecomment-531487825
But if we can't run onnx backend test cases, then we simply can't claim that
we support onnx backend.
We don't have to pass all the test
moazreyad commented on a change in pull request #529: SINGA-484 Code analysis
with LGTM
URL: https://github.com/apache/incubator-singa/pull/529#discussion_r324426747
##
File path: include/singa/io/transformer.h
##
@@ -56,12 +56,12 @@ class ImageTransformer: public
nudles merged pull request #531: SINGA-485 Add code coverage with codecov
URL: https://github.com/apache/incubator-singa/pull/531
This is an automated message from the Apache Git Service.
To respond to the message, please
nudles commented on a change in pull request #529: SINGA-484 Code analysis with
LGTM
URL: https://github.com/apache/incubator-singa/pull/529#discussion_r324411358
##
File path: include/singa/io/transformer.h
##
@@ -56,12 +56,12 @@ class ImageTransformer: public
joddiy commented on issue #462: SINGA-333 Add ONNX backend test
URL: https://github.com/apache/incubator-singa/pull/462#issuecomment-531445928
please refer to [here](https://github.com/apache/incubator-singa/pull/534)
This is
joddiy commented on issue #534: Test onnx backend
URL: https://github.com/apache/incubator-singa/pull/534#issuecomment-531445904
Please don't merge this PR until I finish all the test cases.
This is an automated message from
joddiy opened a new pull request #534: Test onnx backend
URL: https://github.com/apache/incubator-singa/pull/534
add onnx backend test cases
This is an automated message from the Apache Git Service.
To respond to the message,
joddiy commented on issue #462: SINGA-333 Add ONNX backend test
URL: https://github.com/apache/incubator-singa/pull/462#issuecomment-531445841
I think we cannot directly use all the onnx test cases. Since, first, it's
hard to integrate the sonnx into it. Second, there are lots of test
nudles merged pull request #532: SINGA-487 Add the support of Python
Multiprocess Module
URL: https://github.com/apache/incubator-singa/pull/532
This is an automated message from the Apache Git Service.
To respond to the
nudles commented on issue #462: SINGA-333 Add ONNX backend test
URL: https://github.com/apache/incubator-singa/pull/462#issuecomment-531445209
@joddiy can we merge this one? or we need to wait until the operators are
all implemented?
nudles commented on a change in pull request #533: fixed operation add
assertion for convolution
URL: https://github.com/apache/incubator-singa/pull/533#discussion_r324410807
##
File path: src/core/tensor/tensor_math_cpp.h
##
@@ -868,6 +868,25 @@ void RowMax(const Tensor&
dcslin commented on issue #533: fixed operation add assertion for convolution
URL: https://github.com/apache/incubator-singa/pull/533#issuecomment-530344769
after the fix, testing on `python/resnet.py` shows similar result:
Before:
```
dcslin commented on issue #533: fixed operation add assertion for convolution
URL: https://github.com/apache/incubator-singa/pull/533#issuecomment-530340346
2nd commit fixed that "cudnn softmax is not used" mentioned in
dcslin opened a new pull request #533: fixed operation add assertion for
convolution
URL: https://github.com/apache/incubator-singa/pull/533
this is required to run the `autogra/resnet.py`
This is an automated message from
chrishkchris opened a new pull request #532: SINGA-487 Add the support of
Python Multiprocess Module
URL: https://github.com/apache/incubator-singa/pull/532
I have added the support for the Python Multiprocess Module for single-node
multi-gpu scanerio.
For the old MPI-based NCCL, I
nudles merged pull request #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and
moazreyad merged pull request #426: SINGA-405 Add roles to contributors
URL: https://github.com/apache/incubator-singa/pull/426
This is an automated message from the Apache Git Service.
To respond to the message, please log
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-528186487
Therefore, I suggest we may change the input shape assertion of the ADD
function in this way: if the input shapes of the two operands are the
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-528186487
Therefore, I suggest we may change the input shape assertion of the ADD
function in this way: if the input shapes of the two operands are the
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-528186487
Therefore, I suggest we may change the input shape assertion of the ADD
function in this way: if the input shapes of the two operands are the
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-528186487
Therefore, I suggest we may change the input shape assertion of the ADD
function in this way: if the input shapes of the two operands are the
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-528186487
Therefore, I suggest we may change the input shape assertion of the ADD
function in this way: if the input shapes of the two operands are the
chrishkchris commented on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-528186487
Therefore, I suggest we may change the input shape assertion of the ADD
function in this way: if the input shapes of the two operands are The the same,
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
chrishkchris edited a comment on issue #468: Distributted module
URL: https://github.com/apache/incubator-singa/pull/468#issuecomment-527824025
> I have combined all the commits into two commits. Meanwhile, I found that
the resnet.py is not compatible with the master branch modified "Add"
1 - 100 of 714 matches
Mail list logo