tensorpack.tfv1

docs/modules/index.rst

@@ -4,7 +4,6 @@ API Documentation
 .. toctree::
   :maxdepth: 1
 
-
   dataflow
   dataflow.dataset
   dataflow.imgaug
@@ -17,3 +16,4 @@ API Documentation
   tfutils
   utils
   contrib
+  tfv1

examples/ConvolutionalPoseMachines/README.md

-Moved to [../CaffeModels](../CaffeModels).

examples/DisturbLabel/README.md

-
-## DisturbLabel
-
-I ran into the paper [DisturbLabel: Regularizing CNN on the Loss Layer](https://arxiv.org/abs/1605.00055) on CVPR16,
-which basically said that noisy data gives you better performance.
-As many, I didn't believe the method and the results. This code exists to
-disprove the results in the paper.
-
-This is a simple mnist training script with DisturbLabel. It uses the simple architecture in the paper, and
-hyperparameters in my original [mnist example](../mnist-convnet.py).
-The results surprised me, clean labels give the worst accuracy:
-
-![mnist](mnist.png)
-
-Experiements were repeated 15 times for p=0, 10 times for p=0.02 & 0.05, and 5 times for other values
-of p. All experiements run for 100 epochs, with lr decay, which are enough for them to converge.
-
-I suppose the disturb method works as a random noise that could prevent SGD from getting stuck, if
-training data are too easy to fit or too few.
-The method didn't work for slightly harder problems such as SVHN:
-
-![svhn](svhn.png)
-
-The SVHN experiements used the model & hyperparemeters as my original [svhn example](../svhn-digit-convnet.py).
-Experiements were all repeated 10 times to get the error bar. It apparently does not work.
-
-It will not work for ImageNet either. There is indeed a terribly weak
-ImageNet experiment in this paper, and that's a CVPR paper.

examples/DisturbLabel/disturb.py

-# -*- coding: utf-8 -*-
-# File: disturb.py
-# Author: Yuxin Wu
-
-from tensorpack.dataflow import ProxyDataFlow, RNGDataFlow
-
-
-class DisturbLabel(ProxyDataFlow, RNGDataFlow):
-    def __init__(self, ds, prob):
-        super(DisturbLabel, self).__init__(ds)
-        self.prob = prob
-
-    def reset_state(self):
-        RNGDataFlow.reset_state(self)
-        ProxyDataFlow.reset_state(self)
-
-    def __iter__(self):
-        for dp in self.ds:
-            img, l = dp
-            if self.rng.rand() < self.prob:
-                l = self.rng.choice(10)
-            yield [img, l]

examples/DisturbLabel/mnist-disturb.py

-#!/usr/bin/env python
-# -*- coding: utf-8 -*-
-# File: mnist-disturb.py
-
-import argparse
-import imp
-import os
-import tensorflow as tf
-
-from tensorpack import *
-from tensorpack.dataflow import dataset
-from tensorpack.utils import logger
-
-from disturb import DisturbLabel
-
-mnist_example = imp.load_source('mnist_example',
-                                os.path.join(os.path.dirname(__file__), '..', 'basics', 'mnist-convnet.py'))
-get_config = mnist_example.get_config
-
-
-def get_data():
-    dataset_train = BatchData(DisturbLabel(dataset.Mnist('train'), args.prob), 128)
-    dataset_test = BatchData(dataset.Mnist('test'), 256, remainder=True)
-    return dataset_train, dataset_test
-
-
-mnist_example.get_data = get_data
-
-
-class Model(mnist_example.Model):
-    def build_graph(self, image, label):
-        image = tf.expand_dims(image, 3)
-
-        logits = (LinearWrap(image)  # the starting brace is oactivationy for line-breaking
-                  .Conv2D('conv0', 32, 5, padding='VALID', activation=tf.nn.relu)
-                  .MaxPooling('pool0', 2)
-                  .Conv2D('conv1', 64, 5, padding='VALID', activation=tf.nn.relu)
-                  .MaxPooling('pool1', 2)
-                  .FullyConnected('fc0', 512, activation=tf.nn.relu)
-                  .FullyConnected('fc1', out_dim=10, activation=tf.identity)())
-        tf.nn.softmax(logits, name='prob')
-
-        wrong = tf.cast(tf.logical_not(tf.nn.in_top_k(logits, label, 1)), tf.float32, name='incorrect_vector')
-        add_moving_summary(tf.reduce_mean(wrong, name='train_error'))
-
-        cost = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=logits, labels=label)
-        cost = tf.reduce_mean(cost, name='cross_entropy_loss')
-        wd_cost = tf.multiply(1e-5, regularize_cost('fc.*/W', tf.nn.l2_loss),
-                              name='regularize_loss')
-
-        return tf.add_n([wd_cost, cost], name='cost')
-
-
-if __name__ == '__main__':
-    parser = argparse.ArgumentParser()
-    parser.add_argument('--gpu', help='comma separated list of GPU(s) to use.')
-    parser.add_argument('--prob', help='disturb prob', type=float, required=True)
-    args = parser.parse_args()
-    if args.gpu:
-        os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
-
-    logger.auto_set_dir()
-    config = get_config()
-    launch_train_with_config(config, SimpleTrainer())

examples/DisturbLabel/svhn-disturb.py

-#!/usr/bin/env python
-# -*- coding: utf-8 -*-
-# File: svhn-disturb.py
-
-import argparse
-import imp
-import os
-
-from tensorpack import *
-from tensorpack.dataflow import dataset
-from tensorpack.utils import logger
-
-from disturb import DisturbLabel
-
-svhn_example = imp.load_source('svhn_example',
-                               os.path.join(os.path.dirname(__file__), '..',
-                                            'basics', 'svhn-digit-convnet.py'))
-Model = svhn_example.Model
-
-
-def get_data():
-    d1 = dataset.SVHNDigit('train')
-    d2 = dataset.SVHNDigit('extra')
-    data_train = RandomMixData([d1, d2])
-    data_train = DisturbLabel(data_train, args.prob)
-    data_test = dataset.SVHNDigit('test')
-
-    augmentors = [
-        imgaug.Resize((40, 40)),
-        imgaug.Brightness(30),
-        imgaug.Contrast((0.5, 1.5)),
-    ]
-    data_train = AugmentImageComponent(data_train, augmentors)
-    data_train = BatchData(data_train, 128)
-    data_train = MultiProcessRunner(data_train, 5, 5)
-
-    augmentors = [imgaug.Resize((40, 40))]
-    data_test = AugmentImageComponent(data_test, augmentors)
-    data_test = BatchData(data_test, 128, remainder=True)
-    return data_train, data_test
-
-
-if __name__ == '__main__':
-    parser = argparse.ArgumentParser()
-    parser.add_argument('--prob', help='disturb prob', type=float, required=True)
-    args = parser.parse_args()
-
-    logger.auto_set_dir()
-    data_train, data_test = get_data()
-    config = TrainConfig(
-        model=Model(),
-        data=QueueInput(data_train),
-        callbacks=[
-            ModelSaver(),
-            InferenceRunner(data_test,
-                            ScalarStats(['cost', 'accuracy']))
-        ],
-        max_epoch=350,
-    )
-    launch_train_with_config(config, SimpleTrainer())

examples/FasterRCNN/modeling/generalized_rcnn.py

 # -*- coding: utf-8 -*-
 # File:
 
-from tensorpack.compat import tfv1 as tf
+from tensorpack import tfv1 as tf
 from tensorpack import ModelDesc
 from tensorpack.models import GlobalAvgPooling, l2_regularizer, regularize_cost
 from tensorpack.tfutils import optimizer

examples/ImageNetModels/alexnet.py

 #!/usr/bin/env python
-# -*- coding: utf-8 -*-
 # File: alexnet.py
 
 import argparse
 import numpy as np
 import os
 import cv2
-import tensorflow as tf
 
+from tensorpack import tfv1 as tf
 from tensorpack import *
 from tensorpack.dataflow import imgaug
 from tensorpack.tfutils import argscope

examples/ImageNetModels/imagenet_utils.py

-# -*- coding: utf-8 -*-
 # File: imagenet_utils.py
 
-
 import multiprocessing
 import numpy as np
 import os
 from abc import abstractmethod
 import cv2
-import tensorflow as tf
 import tqdm
 
+from tensorpack import tfv1 as tf
 from tensorpack import ModelDesc
 from tensorpack.dataflow import (
     AugmentImageComponent, BatchData, MultiThreadMapData,

examples/ImageNetModels/inception-bn.py

 #!/usr/bin/env python
-# -*- coding: utf-8 -*-
 # File: inception-bn.py
 # Author: Yuxin Wu
 
 import argparse
 import os
-import tensorflow as tf
+from tensorpack import tfv1 as tf
 
 from tensorpack import *
 from tensorpack.dataflow import dataset

examples/ImageNetModels/shufflenet.py

@@ -7,7 +7,7 @@ import math
 import numpy as np
 import os
 import cv2
-import tensorflow as tf
+from tensorpack import tfv1 as tf
 
 from tensorpack import *
 from tensorpack.dataflow import imgaug

examples/ImageNetModels/vgg16.py

@@ -4,7 +4,7 @@
 
 import argparse
 import os
-import tensorflow as tf
+from tensorpack import tfv1 as tf
 
 from tensorpack import *
 from tensorpack.tfutils import argscope

examples/SpatialTransformer/mnist-addition.py

 #!/usr/bin/env python
-# -*- coding: utf-8 -*-
 # File: mnist-addition.py
 # Author: Yuxin Wu
 
@@ -7,7 +6,7 @@ import argparse
 import numpy as np
 import os
 import cv2
-import tensorflow as tf
+from tensorpack import tfv1 as tf
 
 from tensorpack import *
 from tensorpack.dataflow import dataset
@@ -194,7 +193,10 @@ def get_data(isTrain):
 
     ds = JoinData([ds, ds])
     # stack the two digits into two channels, and label it with the sum
-    ds = MapData(ds, lambda dp: [np.stack([dp[0], dp[2]], axis=2), dp[1] + dp[3]])
+
+    def mapper(dp):
+        return [np.stack([dp[0], dp[2]], axis=2), dp[1] + dp[3]]
+    ds = MapData(ds, dp)
     ds = BatchData(ds, 128)
     return ds
 

examples/basics/cifar-convnet.py

@@ -3,7 +3,7 @@
 # Author: Yuxin Wu
 import argparse
 import os
-from tensorpack.compat import tfv1 as tf
+from tensorpack import tfv1 as tf
 
 from tensorpack import *
 from tensorpack.dataflow import dataset

examples/basics/export-model.py

@@ -3,7 +3,7 @@
 
 import argparse
 import cv2
-from tensorpack.compat import tfv1 as tf
+from tensorpack import tfv1 as tf
 from tensorpack import *
 from tensorpack.tfutils.export import ModelExporter
 

examples/basics/mnist-convnet.py

@@ -2,7 +2,7 @@
 # -*- coding: utf-8 -*-
 # File: mnist-convnet.py
 
-from tensorpack.compat import tfv1 as tf
+from tensorpack import tfv1 as tf
 from tensorpack import *
 from tensorpack.dataflow import dataset
 from tensorpack.tfutils import summary

examples/basics/mnist-tflayers.py

@@ -2,7 +2,7 @@
 # -*- coding: utf-8 -*-
 # File: mnist-tflayers.py
 
-from tensorpack.compat import tfv1 as tf
+from tensorpack import tfv1 as tf
 from tensorpack import *
 from tensorpack.dataflow import dataset
 from tensorpack.tfutils import summary

examples/basics/mnist-visualizations.py

@@ -6,7 +6,7 @@
 The same MNIST ConvNet example, but with weights/activations visualization.
 """
 
-from tensorpack.compat import tfv1 as tf
+from tensorpack import tfv1 as tf
 from tensorpack import *
 from tensorpack.dataflow import dataset
 

examples/basics/svhn-digit-convnet.py

@@ -3,7 +3,7 @@
 
 import argparse
 import os
-from tensorpack.compat import tfv1 as tf
+from tensorpack import tfv1 as tf
 from tensorpack import *
 from tensorpack.dataflow import dataset
 from tensorpack.tfutils.summary import *

tensorpack/__init__.py

@@ -21,3 +21,5 @@ if STATICA_HACK:
     from tensorpack.train import *
     from tensorpack.input_source import *
     from tensorpack.predict import *
+
+from tensorpack.compat import tfv1

tensorpack/compat/__init__.py

 import tensorflow as tf
 
 
-def backport_tensor_spec():
+def _backport_tensor_spec():
     if hasattr(tf, 'TensorSpec'):
         return tf.TensorSpec
     try:
@@ -19,6 +19,9 @@ def backport_tensor_spec():
 
 
 def is_tfv2():
+    """
+    Returns whether tensorflow is operating in V2 mode.
+    """
     try:
         from tensorflow.python import tf2
         return tf2.enabled()

tensorpack/dataflow/imgaug/geometry.py

@@ -80,7 +80,7 @@ class Rotation(ImageAugmentor):
             arr = cv2.warpAffine(arr, mat, arr.shape)
         assert np.all(arr == orig)
         """
-        mat = cv2.getRotationMatrix2D(tuple(center - 0.5), deg, 1)
+        mat = cv2.getRotationMatrix2D(tuple(center - 0.5), float(deg), 1)
         return WarpAffineTransform(
             mat, img.shape[1::-1], interp=self.interp,
             borderMode=self.border, borderValue=self.border_value)
@@ -104,7 +104,7 @@ class RotationAndCropValid(ImageAugmentor):
         deg = self._rand_range(-self.max_deg, self.max_deg)
         if self.step_deg:
             deg = deg // self.step_deg * self.step_deg
-        return deg
+        return float(deg)
 
     def get_transform(self, img):
         deg = self._get_deg(img)

tensorpack/models/conv2d.py

@@ -87,17 +87,17 @@ def Conv2D(
         channel_axis = 3 if data_format == 'NHWC' else 1
         in_channel = in_shape[channel_axis]
         assert in_channel is not None, "[Conv2D] Input cannot have unknown channel!"
-        assert in_channel % split == 0
+        assert in_channel % split == 0, in_channel
 
         assert kernel_regularizer is None and bias_regularizer is None and activity_regularizer is None, \
             "Not supported by group conv or dilated conv!"
 
         out_channel = filters
-        assert out_channel % split == 0
+        assert out_channel % split == 0, out_channel
         assert dilation_rate == [1, 1] or get_tf_version_tuple() >= (1, 5), 'TF>=1.5 required for dilated conv.'
 
         kernel_shape = shape2d(kernel_size)
-        filter_shape = kernel_shape + [in_channel / split, out_channel]
+        filter_shape = kernel_shape + [in_channel // split, out_channel]
         stride = shape4d(strides, data_format=data_format)
 
         kwargs = {"data_format": data_format}

tensorpack/train/model_desc.py

@@ -7,9 +7,9 @@ import tensorflow as tf
 from ..utils.argtools import memoized_method
 from ..tfutils.common import get_op_tensor_name
 from ..tfutils.tower import get_current_tower_context
-from ..compat import backport_tensor_spec, tfv1
+from ..compat import _backport_tensor_spec, tfv1
 
-TensorSpec = backport_tensor_spec()
+TensorSpec = _backport_tensor_spec()
 
 
 __all__ = ['ModelDesc', 'ModelDescBase']