Merge pull request #813 from chrishkchris/resolve_conflict

Update the dev branch before the development

CMakeLists.txt

@@ -22,17 +22,22 @@ PROJECT(singa)
 
 LIST(APPEND CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake/Thirdparty)
 
-include(GetGitRevisionDescription)
-git_describe(VERSION --tags --dirty=-d)
-string(REGEX REPLACE "^([0-9]+)\\..*" "\\1" VERSION_MAJOR "${VERSION}")
-string(REGEX REPLACE "^[0-9]+\\.([0-9]+).*" "\\1" VERSION_MINOR "${VERSION}")
-string(REGEX REPLACE "^[0-9]+\\.[0-9]+\\.([0-9]+).*" "\\1" VERSION_PATCH "${VERSION}")
+#include(GetGitRevisionDescription)
+#git_describe(VERSION --tags --dirty=-d)
+#string(REGEX REPLACE "^([0-9]+)\\..*" "\\1" VERSION_MAJOR "${VERSION}")
+#string(REGEX REPLACE "^[0-9]+\\.([0-9]+).*" "\\1" VERSION_MINOR "${VERSION}")
+#string(REGEX REPLACE "^[0-9]+\\.[0-9]+\\.([0-9]+).*" "\\1" VERSION_PATCH "${VERSION}")
+
+
+SET(PACKAGE_VERSION 3.1.0) # ${VERSION})
+SET(VERSION 3.1.0)
+SET(SINGA_MAJOR_VERSION 3)
+SET(SINGA_MINOR_VERSION 1)
+SET(SINGA_PATCH_VERSION 0)
+#SET(SINGA_MAJOR_VERSION ${VERSION_MAJOR})  # 0 -
+#SET(SINGA_MINOR_VERSION ${VERSION_MINOR})  # 0 - 9
+#SET(SINGA_PATCH_VERSION ${VERSION_PATCH})  # 0 - 99
 
-
-SET(PACKAGE_VERSION ${VERSION})
-SET(SINGA_MAJOR_VERSION ${VERSION_MAJOR})  # 0 -
-SET(SINGA_MINOR_VERSION ${VERSION_MINOR})  # 0 - 9
-SET(SINGA_PATCH_VERSION ${VERSION_PATCH})  # 0 - 99
 MATH(EXPR SINGA_VERSION "${SINGA_MAJOR_VERSION} * 1000 + ${SINGA_MINOR_VERSION} * 100 + ${SINGA_PATCH_VERSION}")
 
 message(STATUS "SINGA git Version ${VERSION}")

LICENSE

@@ -512,6 +512,8 @@ SINGA bundles the following under Apache License Version 2.0:
 examples/onnx/tokenization.py
 examples/onnx/run_onnx_squad.py
 examples/onnx/inputs.json
+examples/onnx/gpt2/requirements.txt
+examples/onnx/ro_bert_a/requirements.txt
 
 # Copyright 2018 The Google AI Language Team Authors.
 #
@@ -526,3 +528,34 @@ examples/onnx/inputs.json
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+
+
+===============================================================================
+SINGA bundles the following under new BSD 3-clause license:
+tool/docker/devel/centos6/cuda10/cuda.repo
+
+Copyright (c) 2019,2020 NVIDIA CORPORATION. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions
+are met:
+ * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+   notice, this list of conditions and the following disclaimer in the
+   documentation and/or other materials provided with the distribution.
+ * Neither the name of NVIDIA CORPORATION nor the names of its
+   contributors may be used to endorse or promote products derived
+   from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

README.md

@@ -21,9 +21,9 @@
 
 # Apache SINGA
 
-[![Native Ubuntu build status](https://github.com/apache/singa/workflows/Native-Ubuntu/badge.svg)
-[![Native Mac build status](https://github.com/apache/singa/workflows/Native-MacOS/badge.svg)
-[![conda build status](https://github.com/apache/singa/workflows/conda/badge.svg)
+![Native Ubuntu build status](https://github.com/apache/singa/workflows/Native-Ubuntu/badge.svg)
+![Native Mac build status](https://github.com/apache/singa/workflows/Native-MacOS/badge.svg)
+![conda build status](https://github.com/apache/singa/workflows/conda/badge.svg)
 [![Documentation Status](https://readthedocs.org/projects/apache-singa/badge/?version=latest)](https://apache-singa.readthedocs.io/en/latest/?badge=latest)
 ![License](http://img.shields.io/:license-Apache%202.0-blue.svg)
 [![Follow Apache SINGA on Twitter](https://img.shields.io/twitter/follow/apachesinga.svg?style=social&label=Follow)](https://twitter.com/ApacheSinga)
@@ -35,7 +35,7 @@ Distributed deep learning system
 
 ## Quick Start
 
-* [Installation](doc/en/docs/installation.md)
+* [Installation](http://singa.apache.org/docs/installation/)
 * [Examples](examples)
 
 ## Issues

RELEASE_NOTES

@@ -1,3 +1,44 @@
+Release Notes - SINGA - Version singa-3.1.0
+
+SINGA is a distributed deep learning library.
+
+This release includes following changes:
+
+  * Tensor core:
+    * Support tensor transformation (reshape, transpose) for tensors up to 6 dimensions.
+    * Implement traverse_unary_transform in Cuda backend, which is similar to CPP backend one.
+
+  * Add new tensor operators into the autograd module, including
+    CosSim, DepthToSpace, Embedding, Erf, Expand, Floor, Pad, Round, Rounde, SpaceToDepth, UpSample, Where. 
+    The corresponding ONNX operators are thus supported by SINGA.
+
+  * Add Embedding and Gemm into the layer module.
+
+  * Add SGD operators to opt module, including RMSProp, Adam, and AdaGrad.
+
+  * Extend the sonnx module to support 
+    DenseNet121, ShuffleNetv1, ShuffleNetv2, SqueezeNet, VGG19, GPT2, and RoBERTa,
+
+  * Reconstruct sonnx to 
+    * Support creating operators from both layer and autograd.
+    * Re-write SingaRep to provide a more powerful intermediate representation of SINGA.
+    * Add a SONNXModel which implements from Model to provide uniform API and features.
+
+  * Add one example that trains a BiLSTM model over the InsuranceQA data.
+
+  * Replace the Travis CI with Github workflow. Add quality and coverage management.
+
+  * Add compiling and packaging scripts to creat wheel packages for distribution.
+
+  * Fix bugs
+    * Fix IMDB LSTM model example training script.
+    * Fix Tensor operation Mult on Broadcasting use cases.
+    * Gaussian function on Tensor now can run on Tensor with odd size.
+    * Updated a testing helper function gradients() in autograd to lookup param gradient by param python object id for testing purpose.
+
+
+----------------------------------------------------------------------------------------------
+
 Release Notes - SINGA - Version singa-3.0.0
 
 SINGA is a distributed deep learning library.

examples/cnn/train_cnn.py

@@ -150,8 +150,8 @@ def run(global_rank,
             os.path.abspath(inspect.getfile(inspect.currentframe())))
         parent = os.path.dirname(current)
         sys.path.insert(0, parent)
-        from mlp import module
-        model = module.create_model(data_size=data_size,
+        from mlp import model
+        model = model.create_model(data_size=data_size,
                                     num_classes=num_classes)
 
     # For distributed training, sequential gives better performance

examples/onnx/densenet121.py

@@ -0,0 +1,116 @@
+#
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under th
+
+import os
+import numpy as np
+from PIL import Image
+
+from singa import device
+from singa import tensor
+from singa import autograd
+from singa import sonnx
+import onnx
+from utils import download_model, update_batch_size, check_exist_or_download
+
+import logging
+logging.basicConfig(level=logging.INFO, format='%(asctime)-15s %(message)s')
+
+
+def preprocess(img):
+    img = img.resize((256, 256))
+    img = img.crop((16, 16, 240, 240))
+    img = np.array(img).astype(np.float32) / 255.
+    img = np.rollaxis(img, 2, 0)
+    for channel, mean, std in zip(range(3), [0.485, 0.456, 0.406],
+                                  [0.229, 0.224, 0.225]):
+        img[channel, :, :] -= mean
+        img[channel, :, :] /= std
+    img = np.expand_dims(img, axis=0)
+    return img
+
+
+def get_image_labe():
+    # download label
+    label_url = 'https://s3.amazonaws.com/onnx-model-zoo/synset.txt'
+    with open(check_exist_or_download(label_url), 'r') as f:
+        labels = [l.rstrip() for l in f]
+
+    # download image
+    image_url = 'https://s3.amazonaws.com/model-server/inputs/kitten.jpg'
+    img = Image.open(check_exist_or_download(image_url))
+    return img, labels
+
+
+class Infer:
+
+    def __init__(self, sg_ir):
+        self.sg_ir = sg_ir
+        for idx, tens in sg_ir.tensor_map.items():
+            # allow the tensors to be updated
+            tens.requires_grad = True
+            tens.stores_grad = True
+            sg_ir.tensor_map[idx] = tens
+
+    def forward(self, x):
+        return sg_ir.run([x])[0]
+
+
+if __name__ == "__main__":
+    url = 'https://s3.amazonaws.com/download.onnx/models/opset_9/densenet121.tar.gz'
+    download_dir = '/tmp/'
+    model_path = os.path.join(download_dir, 'densenet121', 'model.onnx')
+
+    logging.info("onnx load model...")
+    download_model(url)
+    onnx_model = onnx.load(model_path)
+
+    # set batch size
+    onnx_model = update_batch_size(onnx_model, 1)
+
+    # prepare the model
+    logging.info("prepare model...")
+    dev = device.create_cuda_gpu()
+    sg_ir = sonnx.prepare(onnx_model, device=dev)
+    autograd.training = False
+    model = Infer(sg_ir)
+
+    # verifty the test
+    # from utils import load_dataset
+    # inputs, ref_outputs = load_dataset(
+    #     os.path.join('/tmp', 'densenet121', 'test_data_set_0'))
+    # x_batch = tensor.Tensor(device=dev, data=inputs[0])
+    # outputs = model.forward(x_batch)
+    # for ref_o, o in zip(ref_outputs, outputs):
+    #     np.testing.assert_almost_equal(ref_o, tensor.to_numpy(o), 4)
+
+    # inference
+    logging.info("preprocessing...")
+    img, labels = get_image_labe()
+    img = preprocess(img)
+
+    logging.info("model running...")
+    x_batch = tensor.Tensor(device=dev, data=img)
+    y = model.forward(x_batch)
+
+    logging.info("postprocessing...")
+    y = tensor.softmax(y)
+    scores = tensor.to_numpy(y)
+    scores = np.squeeze(scores)
+    a = np.argsort(scores)[::-1]
+    for i in a[0:5]:
+        logging.info('class=%s ; probability=%f' % (labels[i], scores[i]))

examples/onnx/gpt2/gpt2.py

@@ -106,5 +106,5 @@ def train_one_batch(self, x, y):
         y = tensor.Tensor(device=dev, data=y)
         x = tensor.concatenate([x, y], 2)
 
-    text = tokenizer.decode(output)
+    text = postprocess(output)
     print(text)

examples/onnx/ro_bert_a/requirements.txt

@@ -0,0 +1 @@
+transformers==2.5.1

examples/onnx/ro_bert_a.py → examples/onnx/ro_bert_a/ro_bert_a.py

@@ -17,10 +17,6 @@
 # under the License.
 #
 
-
-import sys
-sys.path.append('/singa/build/python/')
-
 import os
 import numpy as np
 
@@ -30,6 +26,8 @@
 from singa import autograd
 import onnx
 
+import sys
+sys.path.append(os.path.dirname(__file__) + '/..')
 from utils import download_model, check_exist_or_download
 
 import logging
@@ -45,12 +43,6 @@ def preprocess():
     tokens = np.array(tokens)
     return tokens.reshape([1, -1]).astype(np.float32)
 
-
-def postprocess(out):
-    text = tokenizer.decode(out)
-    return text
-
-
 class MyModel(sonnx.SONNXModel):
 
     def __init__(self, onnx_model):