You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@singa.apache.org by jo...@apache.org on 2020/09/24 04:00:28 UTC
[singa] branch master updated (a00a23b -> 4de1590)
This is an automated email from the ASF dual-hosted git repository.
joddiyzhang pushed a change to branch master
in repository https://gitbox.apache.org/repos/asf/singa.git.
from a00a23b Merge pull request #711 from agnesnatasya/squeezenet
new 9c65c99 Implement vgg19 model from onnx model zoo
new 319358f Merge pull request #717 from Alvinnyk/vgg19
new f334f6b Shufflenet v1
new 456ef8b Merge pull request #718 from Shashankwer/shufflenetv1
new 3ce2966 Implement densenet121 from ONNX
new ace84fe Change device to GPU
new c5769f1 Merge pull request #721 from agnesnatasya/densenet
new db1846d Merge pull request #689 from chrishkchris/dev
new efd7cba SINGA-510 Add Time-Profiling Function for Single GPU
new 989abdc SINGA-510 Address review comments
new dc5dd4c SINGA-510 Use int for Verobsity and remove profiling mode
new 84de1af Merge pull request #694 from chrishkchris/SINGA-510
new dd2fb3f SINGA-510 Skip initial iterations
new e21f833 SINGA-510 Minimize measure time overhead
new 0f3beb2 SINGA-510 Use one Sync function only
new 7af0125 SINGA-510 Add Set Skip Iteration
new ae98f94 SINGA-510 Address Review Comment
new bec1964 Merge pull request #708 from chrishkchris/SINGA-510
new b37e950 Update version of CUDNN
new ca13bc2 Update meta.yaml
new 4cae51b add option for cuda10.2
new dd18aff Merge pull request #709 from chrishkchris/CUDNN765
new be28824 create proof of concept script for testing
new 37188ae update decorator
new 9af54bd fix bugs of buffering operations
new a57b39f separate layers from autograd.py into layer.py
new a4bc5f6 rename operation to operator
new fb69039 fix the usage errors of layers
new 522a271 fix alerts
new 765a771 rename module to model
new 5688178 create aliases for Operator and Tensor
new ad32bf9 create aliases for layers for backward compatibility
new 4b7ec13 make Model(renamed from module) a subclass of Layer compile / do_init: separate the initialzation and forward propagation for Linear, create name of each layer save_states/load_states/set_states: recursive logic
new 5fc48e3 renamed initialize method according to latest API design; auto check inputsize from x in linear and conv2d layer, add maintaining backward compatibility
new 8b3ca83 set a unique name for every layer
new 5788005 support setting and getting params for each layer
new 362f9aa setup load save states framework
new 546f3d9 complete setting and getting params and states
new 5833136 allow list of tensor as input of initialize() Fix conv2d in channels check Fix rnn_base override method Fix LSTM rnn initialise args, __init__ method, params
new 20a9b6b fix creator assignment for intermediate tensor when training is not used
new ed1b7f4 rename some parameters and move all init code into initialize
new 880de7a fix bugs of models in train.py, rm files related to the old layer.py
new ca474ef fix test cases & add gemm layer
new 72ef70d fix layer to suit onnx logic
new 3569a3c fixed CURAND_STATUS_LAUNCH_FAILURE which is due to dev not sanitized across all the params/states
new 4d36dbb refine save load states for model, detailed testing added
new 55eb2b5 pass test cases in test_operations except conv1d and pooling1d
new f0708b5 fix conv1d and pool1d
new ee5607c update some features for onnx test cases
new 74ddf8f fix bug for conv dilation
new da1290d new implemenation for get_params, set_params, get_states, set_states
new feb4365 Restore creator settings for interim tensors; update device_check() after new get states implementation; fix typo in LSTM; fix test case for save model and load model, as new impl does not keep top level layer name; remove proof of concept script.
new 8b1a49d fix bugs in test_model.py, train_mpi.py, layer.py and module.py, rm converter.py, lsgan.py and vanilla.py
new ab0cb13 fixed batchnorm backward compatability
new 4fd4f19 support register layers, fix bugs in xceptionnet and resnet
new 8e94319 some modifications based on review: 1. fix typos in train.py 2. fix training bugs in CharRNN, treat hx, cx as states 3. call intialize before forward 4. remove set_attribute 5. remove on_device, get device info from input tensors 6. support adding activation layer after conv2d 7. add some layers for operators, update models in examples 8. no longer change the default device when creating devices
new ede4a3e Merge pull request #697 from dcslin/py-api
new ad65d62 implemented cudnn lstm, revised api compared to previous version, lstm layer take 1 tensor as input. added lstm model testing, as api is different from previous rnn impl
new 59f7970 add cossim operator
new f52804c Merge remote-tracking branch 'jod/cos_similarity' into cudnn-lstm
new c60fc69 added bilstm model for qa dataset, added customized loss function operator, added gensim embedding transform, implemented bilstm cos sim model, implement mlp model for comparison
new 9ebf331 added support on varying batchsize, seqeunce length for cudnn lstm
new 684580a qabot model updated 2nd data pipeline and model to speed up process
new 8e3c130 updated qabot train on eval and testing top1hits
new 88ffd81 refactor soonx backend
new 77eaf45 refactor sonnx test cases
new 43addc7 refactor sonnx examples
new 038e2df Merge pull request #703 from joddiy/sonnx_new_api
new 4fecdd1 fix soonx test cases
new 43a48be Merge pull request #723 from joddiy/sonnx_new_api
new c4ac130 impl new opt api, decay scheduler constant class, optimzer class, sgd momentum, dampening, added unit test, backward compatibility tested
new 0b0348b optimize opt.py, add new Axpy
new bb1c78f Updates on opt.py Added tensor math floor, and testings Fixed exponential decay errors, and test cases Added tensor floor div Added SGD const LR, momentum, weight decay, nesterov test cases.
new 63c4349 fix bugs of updating params, test examples in cnn
new b9de65f added non staircase expo decay scheduler test, other opt test passes except nesterov on gpu
new 799283f fix some alerts, fix bugs in test_model.py, opt.py and tensor.cc, update some code to use new opt apis
new 735cafb update distopt and argument parsers
new 67304d5 Merge pull request #713 from dcslin/opt_api
new c9981a2 disable compiling py3.6
new a109c46 add dependency to conda build
new d1857c8 fix bugs for transfer learning
new d3d48c5 Merge pull request #726 from XJDKC/dev
new 7725e70 Merge branch 'dev' into cudnn-lstm
new 4d7ac74 Merge branch 'cudnn-lstm' of https://github.com/dcslin/singa into cudnn-lstm
new cb98151 fix rnn cpp device execution lambda pass by value
new d3105fe Merge pull request #725 from joddiy/cos_similarity
new 6156b01 Merge branch 'dev' into cudnn-lstm
new ccf6545 Merge branch 'cudnn-lstm' of https://github.com/dcslin/singa into cudnn-lstm
new fc8c7fc update to use persistent tensor for data container
new 0801593 Merge pull request #722 from dcslin/cudnn-lstm
new de7003d SINGA-510 Move the communicator cuda streams to context
new 218b5ec SINGA-510 Distributed Training Time Profiling
new 3576886 SINGA-510 Separate each stream operation to exclude waiting time
new 3effa89 SINGA-510 Exclude waiting time in synch allreduce
new 6021e70 SINGA-510 Make backward time compatible to distributed training
new 8f284d7 SINGA-510 Address review comment
new 042b081 Merge pull request #716 from chrishkchris/SINGA-510_2
new fb1db9b Update the initializer module to add more initialization methods.
new 78764af Compile the project via setuptools
new 3eb6504 add setup.py.in back, which is used by conda build and native build
new e42b203 Update the initializer module to add more initialization methods.
new e4afa64 merge apache/singa:dev branch
new fde8224 successfully run aduitwheel over the generated wheel inside the docker container from Dockerfile.manylinux2014
new 2ef3d86 finished debugging; to add documentation for building wheel packages and uploading them to pypi repo
new 4c7ff5a added imdb sentiment binary classification model, self-contained data preparation scripts, baseline model added tensor contiguous method, add args to imdb train, added cudnn rnn Ex operations with different seq lengths
new 054f276 Add ShufflenetV2 model
new ae748a3 add expand operator
new ca698bf add expand operator to sonnx
new 1ffc6a9 add pad operator
new 8c47edf add pad operator to sonnx
new 5b5f33a fix linting error by changing repeated file names
new 4e6f9e5 fix onnx backend cuda cases
new 9dd09fa Merge pull request #742 from joddiy/fix-cuda-cases
new 9764217 Merge branch 'dev' of https://github.com/apache/singa into dev
new 182519d Merge pull request #740 from chrishkchris/fix_filename
new 2d28b11 fix gaussian CURAND_STATUS_LENGTH_NOT_MULTIPLE
new 8656a4c Merge pull request #729 from dcslin/gaussian
new 7ed0822 some updates: 1. fix the bug of extra separator before the parameter name 2. add comments for layer and model 3. fix bugs in test_model
new 2857ec7 move the test cases in test_module to test_model
new 84e029c set the tensor name automatically when setting attr
new 8d39068 set the layer name automatically when setting attr
new 35d6cef Merge pull request #731 from XJDKC/param_name
new f7c8b30 cudnn lstm test case update
new f418722 Merge pull request #727 from dcslin/cudnn-lstm
new bd14143 fix sonnx frontend
new 3a4d092 Merge pull request #732 from joddiy/fix-onnx-frontend
new 6978291 add upsample operator
new 0706b04 add upsample operator to soonx
new 78bfc16 Merge pull request #737 from joddiy/add-upsample-operator
new 3006787 set precision
new e500978 Merge pull request #743 from chrishkchris/precision
new 3ce6503 Merge pull request #739 from nudles/dev
new 7d7d02c Merge branch 'dev' into add-pad-operator
new f4a4148 change uniform to he_uniform at example
new b027bf7 Merge pull request #734 from joddiy/add-pad-operator
new 3060382 Merge branch 'dev' into add-expand-operator
new 867c0a3 Merge pull request #736 from joddiy/add-expand-operator
new bff229c Merge branch 'dev' into add-ShufflenetV2-model
new 2593046 Merge pull request #735 from joddiy/add-ShufflenetV2-model
new 4ea5034 added round and rounde round to even tensor methods
new e8b3027 Merge pull request #749 from dcslin/round
new 5225395 Add EQ comparison operator
new d663ab6 Merge pull request #747 from chrishkchris/equal
new 3f5c97a fix pooling for same mode
new 4fa7ea0 Merge pull request #746 from joddiy/fix-pooling-same-mode
new a5ddfad Fix cuda_gpu and linting error
new 0625de8 reset the device when recreating it
new b0fb8b2 fix linting error
new 21be0ca remove unused code
new 9c80cbe Merge pull request #728 from chrishkchris/fix_create_cuda
new 04b2a7b issue 707 additional input check for autograd operator
new ea873fe Merge pull request #751 from dcslin/operator_input_check
new bffd9b7 Merge remote-tracking branch 'origin/dev' into imdb
new 65c81ae Merge pull request #733 from dcslin/imdb
new 6efd294 fix gpu conda build and update numpy version
new f9dd678 range of numpy version
new ddaa2e2 Merge pull request #754 from chrishkchris/fix_roundf
new 20c1174 update roundeven backend algorithm
new 9193218 update round even coding for backward comp
new 3d25b99 Merge pull request #758 from dcslin/jod/add-operators
new e96704e add round and rounde operators
new 67ca2a4 Merge pull request #748 from joddiy/add-operators
new 6b9c573 fix test operation Tensor class non hashable erorr
new b58384c Merge pull request #753 from dcslin/hf-test-operaitons-hashable
new 731b89b fix training loss error
new 8366813 Merge pull request #762 from chrishkchris/fix_loss_error
new 363cdd5 fix kint issue, cast kint to kfloat for computation and cast back
new a05ef9c Merge pull request #763 from dcslin/kint2
new b72d6b2 Fix softmax test case without DNNL
new 15f24f5 add license header
new 9ba6102 remove unused import
new b32700d fix situation tensor!=None
new 8eb952a fix test_operation.py
new 9e7c579 fix test_opt.py
new b79e1db Merge pull request #765 from chrishkchris/fix_softmaxtest
new 95aa1a9 fix onnx device
new becf6a9 Merge pull request #767 from joddiy/fix-onnx-device
new 7ac6804 update from deprecated functions
new 2c08b40 Merge pull request #766 from chrishkchris/testfix
new 850c8ba add where operator
new 61891a7 add onnx gpt2 example
new 6ac87ea Merge branch 'dev' of https://github.com/apache/singa into add-onnx-gpt-model
new 84c920c Merge pull request #755 from joddiy/add-onnx-gpt-model
new 12dfc9b C++ build with Github Actions
new 402ad8f Add Github Actions status badge for C++ build
new 857ddf4 check build on macOS with Github Actions
new 2044d9f Merge branch 'dev' of https://github.com/apache/singa into dev
new 58671eb Add C++ test
new 8b15f19 Update cpp.yaml
new c471c5a Check C++ build and test with oneDNN
new 5747a99 fix DNNL_ROOT
new 690547f turn off the C++ test without oneDNN
new e9e665f Merge pull request #761 from moazreyad/dev
new 54bbea6 test C++ on macOS with Github Actions
new 5f4b250 Merge pull request #768 from moazreyad/dev
new 223607f merge with origin/dev
new 387c76d done the test of the wheel package; updated the documentation of using pip to install singa
new 0456d1a add embedding layer
new e8ba5bb Add RMSProp, AdaGrad, Adam to opt.py
new fbb9f97 Update RBM example
new 81faccd delete old optimizer python interface
new 94ed62c delete old optimizer test cases
new a936833 update singa.i
new 93989b0 Add GAN examples and google formatting
new 7b7359d code cleaning and improve gan
new 12161b3 updated qabot training and data scripts, added max, mean, mlp qabot models, updated ranking loss fn in autograd, fix bug in cudnn rnn in autograd, added some utils tensor fn(random, zeros, ones), added cudnn rnn set param api, fixed and added test to autograd mse loss, cos sim, reduce mean
new 493f185 Merge pull request #772 from dcslin/qabot4
new 6abec93 Merge branch 'dev' into opt
new 67ae44d code cleaning and improve lsgan
new a43dfcc Merge pull request #773 from chrishkchris/opt
new c2c41cd Merge branch 'dev' into embedding
new 3014944 Merge pull request #724 from joddiy/embedding
new 675eae0 fix sonnx last layers
new ff0b239 Merge pull request #774 from joddiy/fix-sonnx-last-layers
new fff2014 Added ScatterElements operation.
new 8ad93fb Correction made to the function
new 87ba8eb reimplement scatter_elements
new c21c3eb Merge branch 'dev' into scatter
new b9b6b8d Merge pull request #1 from joddiy/scatter
new b30db36 Merge branch 'dev' into scatter
new 3f0997d Merge pull request #760 from Shashankwer/scatter
new 3504443 conda build and test with Github Actions
new 24a7d1d Merge pull request #781 from moazreyad/dev
new 79e7adc set training mode before training
new d54a1ca support parsing arguments, reorganize the code
new 5357df5 refactor rnn training example
new d76c855 move the copy operator to Tensor, fix bugs of buffering rnn operators
new 81b87e1 recursively remove creators for results
new 277ada9 optimize adding sync op and debug function
new cfd6eee resolve conflicts
new 2d838ee Merge pull request #1 from apache/dev
new f78d9b3 very minor debug
new bfc16b4 compatiable with utf-8 coding
new be7b413 turn the use_graph flag on
new da34a59 Merge pull request #730 from XJDKC/lstm-graph
new 0c86657 Add code coverage for C++
new a93aaee Merge pull request #782 from moazreyad/dev
new 210a216 merge upstream/dev into pr#770
new e5445a5 remove loss.py and metric.py and the tests for them
new d5cce78 remove native conv cpp impl in cpp operation/; rename github workflow stages
new 5d6c4b4 rename the workflow name for license check.
new fdef3c8 fix test cases error
new fdb8e93 Merge branch 'dev' of https://github.com/apache/singa into fix_test_cases
new 4d06729 Merge pull request #783 from joddiy/fix_test_cases
new 6319c8a Merge remote-tracking branch 'origin/dev' into dev
new 3c3b0cf fix linting error in sonnx.py
new af94f5e fix bugs in unittest
new 446c72c Merge pull request #770 from nudles/dev
new acbf930 added erf backend
new a171023 Merge pull request #777 from dcslin/erf
new 9d85403 Add dist_communicator.i, update rnn training code and instruction
new 7586a3b Add back the old Char RNN
new 178b107 Merge pull request #785 from chrishkchris/dist_interface
new 18c2af6 codecov for python
new 3e83809 Merge pull request #788 from moazreyad/dev
new 8a46d23 fixed imdb train script args
new 2cdb881 Merge pull request #789 from dcslin/imdb_script
new 076871f fix the bug in github workflow (Native-MacOS)
new 71a9fd3 Merge remote-tracking branch 'origin/dev' into dev
new 5114b3c Fix error in running cpp test in MacOS Github CI
new d9981fa macOS.yaml update
new dbb6b5d Merge pull request #786 from nudles/dev
new d693b6b keep the execution order of ops that read and write the same block
new 11894cc turn the sequential flag to False
new 2bc6456 format the added code
new 333c207 Add Test Case for Keep Order
new 2ad3552 Merge pull request #2 from apache/dev
new f4ee540 Merge branch 'keep-order' of https://github.com/XJDKC/singa into keep-order
new b1cd1a8 minor debug test case
new ded7ca8 Merge pull request #787 from XJDKC/keep-order
new 426963d Remove Travis CI
new ea17426 replace travis badge with github workflow bagdes
new 0d79d52 Merge pull request #790 from moazreyad/remove-travis
new cb84c93 added broadcasted() flag for tensor, added broadcast to mult, refactor gpu traverse unary
new e97acb2 fix bug for ignore last dim
new 95f7b39 Merge pull request #794 from dcslin/multbroadcast
new cc25f79 add erf operator
new 9f036b2 add ro_bert_a model
new 46f3339 Merge branch 'multbroadcast' into add-RoBERTa-model
new 87eabef Merge branch 'dev' into add-RoBERTa-model
new e6288fe fix bug for ignore last dim
new cc7c6b3 fix bug within sonnx for multi-dim gemm
new 1861484 Merge pull request #778 from joddiy/add-RoBERTa-model
new 2edf204 Add ShufflenetV2 model
new 043ef63 updated cudnn nd descriptor to allow 6d
new 84251c3 add SpaceToDepth && DepthToSpace operators
new c794398 Merge branch 'dev' into add-depth2space-space2depth
new d91a3e2 Merge pull request #750 from joddiy/add-depth2space-space2depth
new a0d7afb Test anaconda upload using github ci
new 5fb11fd Do conda packaging upload via github CI
new f9f901e Update Githut CI for conda build
new 7b5a0ab change the python version to 3.6 in conda build
new 814e2f7 Merge pull request #795 from nudles/dev
new fb9be0f Adding Operators (#738)
new f5a9cdf Merge branch 'dev' for V3.1-RC1
new 47017c1 fix the link issue in readme file
new 2b9c1f3 Updated activation.cc
new 8ff7e10 Merge pull request #745 from sourcepirate/patch-2
new 2ac89cc fix the license issue
new 2705c65 change the filename to be consistent with the docs
new b18a87e Merge remote-tracking branch 'nudles/master'
new 389bd56 Update LICENSE
new e1f346d Update pom.xml
new 4de1590 Merge pull request #798 from nudles/master
The 2225 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "add" were already present in the repository and have only
been added to this reference.
Summary of changes:
.asf.yaml | 2 +
.asf.yaml => .codecov.yml | 11 +-
.github/workflows/conda.yaml | 71 +
.github/workflows/macOS.yaml | 56 +
.github/workflows/rat.yaml | 2 +-
.github/workflows/ubuntu.yaml | 64 +
.travis.yml | 70 -
CMakeLists.txt | 1 +
LICENSE | 33 +
README.md | 9 +-
examples/cnn/README.md | 2 +-
examples/cnn/autograd/mnist_cnn.py | 27 +-
examples/cnn/autograd/xceptionnet.py | 119 +-
examples/cnn/benchmark.py | 26 +-
examples/cnn/model/alexnet.py | 70 +-
examples/cnn/model/cnn.py | 36 +-
examples/cnn/model/resnet.py | 129 +-
examples/cnn/model/xceptionnet.py | 133 +-
examples/cnn/{train.py => train_cnn.py} | 33 +-
examples/cnn/train_mpi.py | 29 +-
examples/cnn/train_multiprocess.py | 32 +-
examples/{rbm => gan}/README.md | 11 +-
examples/gan/lsgan.py | 335 +-
examples/gan/model/gan_mlp.py | 104 +
examples/gan/model/lsgan_mlp.py | 101 +
examples/gan/vanilla.py | 318 +-
examples/mlp/{module.py => model.py} | 54 +-
examples/mlp/native.py | 2 +-
examples/onnx/arcface.py | 59 +-
examples/onnx/bert/bert-squad.py | 50 +-
examples/onnx/bert/tokenization.py | 4 -
examples/onnx/{vgg16.py => densenet121.py} | 7 +-
examples/onnx/fer_emotion.py | 51 +-
examples/onnx/gpt2/gpt2.py | 110 +
examples/onnx/gpt2/requirements.txt | 1 +
examples/onnx/mnist.py | 320 -
examples/onnx/mobilenet.py | 52 +-
examples/onnx/resnet18.py | 49 +-
examples/onnx/ro_bert_a.py | 103 +
examples/onnx/{resnet18.py => shufflenetv1.py} | 63 +-
examples/onnx/{mobilenet.py => shufflenetv2.py} | 61 +-
examples/onnx/tiny_yolov2.py | 62 +-
examples/onnx/training/model.json | 84 +
examples/{cnn => onnx/training}/train.py | 140 +-
examples/onnx/utils.py | 13 +-
examples/onnx/vgg16.py | 50 +-
examples/onnx/{vgg16.py => vgg19.py} | 13 +-
CONTRIBUTING.md => examples/qabot/README.md | 12 +-
examples/qabot/qabot_data.py | 282 +
examples/qabot/qabot_model.py | 152 +
examples/qabot/qabot_train.py | 159 +
examples/rbm/train.py | 16 +-
examples/rnn/README.md | 21 +-
examples/rnn/{train.py => char_rnn.py} | 59 +-
examples/rnn/imdb_data.py | 283 +
examples/rnn/imdb_model.py | 58 +
examples/rnn/imdb_train.py | 176 +
include/singa/core/common.h | 12 +-
include/singa/core/device.h | 46 +-
include/singa/core/scheduler.h | 48 +-
include/singa/core/tensor.h | 45 +-
include/singa/io/communicator.h | 15 +-
java/pom.xml | 4 +
python/singa/autograd.py | 3018 ++++++----
python/singa/converter.py | 242 -
python/singa/device.py | 62 +-
python/singa/initializer.py | 264 +-
python/singa/layer.py | 2774 +++++----
python/singa/loss.py | 216 -
python/singa/metric.py | 218 -
python/singa/model.py | 354 ++
python/singa/module.py | 193 -
python/singa/net.py | 531 --
python/singa/opt.py | 656 +-
python/singa/optimizer.py | 472 --
python/singa/snapshot.py | 3 +
python/singa/sonnx.py | 1835 +++---
python/singa/tensor.py | 113 +-
python/singa/utils.py | 69 +-
setup.py | 440 ++
src/CMakeLists.txt | 5 +
src/api/core_device.i | 4 +
src/api/core_tensor.i | 36 +-
src/api/model_operation.i | 38 +
src/api/model_optimizer.i | 71 -
src/api/singa.i | 6 +-
src/core/device/cpp_cpu.cc | 13 +
src/core/device/cuda_gpu.cc | 77 +-
src/core/device/device.cc | 41 +-
src/core/device/opencl_device.cc | 2 +-
src/core/device/platform.cc | 2 +
src/core/scheduler/scheduler.cc | 197 +-
src/core/tensor/math_kernel.cu | 93 +-
src/core/tensor/math_kernel.h | 15 +-
src/core/tensor/tensor.cc | 374 +-
src/core/tensor/tensor_math.h | 30 +
src/core/tensor/tensor_math_cpp.h | 101 +
src/core/tensor/tensor_math_cuda.h | 165 +-
src/io/communicator.cc | 345 +-
src/model/layer/activation.cc | 2 +-
src/model/layer/cudnn_activation.cc | 4 +-
src/model/layer/cudnn_convolution.cc | 2 +-
src/model/layer/cudnn_dropout.cc | 8 +-
src/model/operation/batchnorm.cc | 12 +-
src/model/operation/convolution.cc | 34 +-
src/model/operation/pooling.cc | 8 +-
src/model/operation/rnn.cc | 808 +++
src/model/operation/rnn.h | 136 +
test/python/cuda_helper.py | 2 +-
test/python/run.py | 16 +-
test/python/test_api.py | 192 +-
test/python/test_dist.py | 2 +-
test/python/test_initializer.py | 123 +
test/python/test_layer.py | 276 -
test/python/test_loss.py | 71 -
test/python/test_metric.py | 74 -
test/python/test_model.py | 499 ++
test/python/test_module.py | 306 -
test/python/test_net.py | 115 -
test/python/test_onnx.py | 170 +-
test/python/test_onnx_backend.py | 3225 +---------
test/python/test_operation.py | 764 ++-
test/python/test_opt.py | 230 +
test/python/test_optimizer.py | 382 --
test/python/test_tensor.py | 149 +-
test/singa/test_cpp_cpu.cc | 4 +-
test/singa/test_operation_rnn.cc | 141 +
test/singa/test_scheduler.cc | 106 +-
test/singa/test_tensor_math.cc | 10 +
tool/conda/dist/meta.yaml | 8 +-
tool/conda/docker/{cuda10 => cuda10.2}/Dockerfile | 5 +-
tool/conda/gpu/meta.yaml | 2 +-
tool/conda/singa/conda_build_config.yaml | 25 +-
tool/conda/singa/meta.yaml | 13 +-
tool/cpplint.py | 6327 --------------------
tool/docker/README.md | 9 +-
.../devel/centos6/cuda10/Dockerfile.manylinux2014 | 132 +
tool/docker/devel/centos6/cuda10/cuda.repo | 6 +
tool/docker/devel/ubuntu/cuda10/Dockerfile | 2 +-
tool/docker/devel/ubuntu/cuda9/Dockerfile | 2 +-
tool/linting/py.sh | 4 +-
tool/travis/build.sh | 55 -
tool/travis/depends.sh | 43 -
tool/{debian-python2/postinst => wheel.sh} | 26 +-
144 files changed, 13680 insertions(+), 18485 deletions(-)
copy .asf.yaml => .codecov.yml (79%)
create mode 100644 .github/workflows/conda.yaml
create mode 100644 .github/workflows/macOS.yaml
create mode 100644 .github/workflows/ubuntu.yaml
delete mode 100644 .travis.yml
copy examples/cnn/{train.py => train_cnn.py} (93%)
copy examples/{rbm => gan}/README.md (76%)
create mode 100644 examples/gan/model/gan_mlp.py
create mode 100644 examples/gan/model/lsgan_mlp.py
rename examples/mlp/{module.py => model.py} (69%)
copy examples/onnx/{vgg16.py => densenet121.py} (92%)
create mode 100644 examples/onnx/gpt2/gpt2.py
create mode 100644 examples/onnx/gpt2/requirements.txt
delete mode 100644 examples/onnx/mnist.py
create mode 100644 examples/onnx/ro_bert_a.py
copy examples/onnx/{resnet18.py => shufflenetv1.py} (71%)
copy examples/onnx/{mobilenet.py => shufflenetv2.py} (71%)
create mode 100644 examples/onnx/training/model.json
rename examples/{cnn => onnx/training}/train.py (71%)
copy examples/onnx/{vgg16.py => vgg19.py} (91%)
copy CONTRIBUTING.md => examples/qabot/README.md (68%)
create mode 100644 examples/qabot/qabot_data.py
create mode 100644 examples/qabot/qabot_model.py
create mode 100644 examples/qabot/qabot_train.py
rename examples/rnn/{train.py => char_rnn.py} (84%)
create mode 100644 examples/rnn/imdb_data.py
create mode 100644 examples/rnn/imdb_model.py
create mode 100644 examples/rnn/imdb_train.py
delete mode 100644 python/singa/converter.py
delete mode 100644 python/singa/loss.py
delete mode 100644 python/singa/metric.py
create mode 100644 python/singa/model.py
delete mode 100644 python/singa/module.py
delete mode 100755 python/singa/net.py
delete mode 100644 python/singa/optimizer.py
create mode 100644 setup.py
delete mode 100644 src/api/model_optimizer.i
create mode 100644 src/model/operation/rnn.cc
create mode 100644 src/model/operation/rnn.h
create mode 100644 test/python/test_initializer.py
delete mode 100755 test/python/test_layer.py
delete mode 100644 test/python/test_loss.py
delete mode 100644 test/python/test_metric.py
create mode 100644 test/python/test_model.py
delete mode 100644 test/python/test_module.py
delete mode 100644 test/python/test_net.py
create mode 100644 test/python/test_opt.py
delete mode 100644 test/python/test_optimizer.py
create mode 100644 test/singa/test_operation_rnn.cc
copy tool/conda/docker/{cuda10 => cuda10.2}/Dockerfile (92%)
delete mode 100755 tool/cpplint.py
create mode 100644 tool/docker/devel/centos6/cuda10/Dockerfile.manylinux2014
create mode 100644 tool/docker/devel/centos6/cuda10/cuda.repo
delete mode 100644 tool/travis/build.sh
delete mode 100644 tool/travis/depends.sh
copy tool/{debian-python2/postinst => wheel.sh} (50%)