Skip to content

SeetaResearch / Dragon

Go to a project
Commit 218796ed by Ting PAN
Options

Remove the deprecated DALI API 

Summary:
This commit removes the deprecated API for DALI 0.24.
Besides, variable length keyword arguments are added for forward compatibility.
1 parent c40eaf7b
Inline Side-by-side
Showing with 1749 additions and 921 deletions
caffe/__init__.py
......@@ -21,6 +21,7 @@ from dragon.vm.caffe.core.solver import AdamSolver
from dragon.vm.caffe.core.solver import NesterovSolver
from dragon.vm.caffe.core.solver import RMSPropSolver
from dragon.vm.caffe.core.solver import SGDSolver
from dragon.vm.caffe.core.solver import Solver
# Functions
from dragon.vm.caffe.core.net_spec import to_proto
......
caffe/core/solver.py
......@@ -8,7 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""The solver to update parameters."""
"""The solver to optimize parameters."""
from __future__ import absolute_import
from __future__ import division
......@@ -28,7 +28,7 @@ from dragon.vm.caffe.core.proto import caffe_pb2
class Solver(object):
"""The abstraction ``caffe.Solver``."""
"""The base solver class to optimize parameters."""
def __init__(self, solver_file, is_root=True):
"""Create a ``Solver``.
......@@ -330,7 +330,7 @@ class AdamSolver(Solver):
momentum=0.9,
momentum2=0.999,
delta=1e-8,
)
}
```
"""
......@@ -397,7 +397,7 @@ class RMSPropSolver(Solver):
base_lr=0.01,
rms_decay=0.99,
delta=1e-8,
)
}
```
"""
......@@ -430,13 +430,13 @@ class SGDSolver(Solver):
solver {
base_lr=0.01,
momentum=0.9,
)
}
```
"""
def __init__(self, solver_file, is_root=True):
"""Create a `SGDSolver``.
"""Create a ``SGDSolver``.
Parameters
----------
......
cmake/MiscCheck.cmake
......@@ -12,7 +12,7 @@ if (MSVC)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /MP")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}
/wd4003 /wd4114
/wd4244 /wd4251 /wd4273 /wd4275
/wd4244 /wd4251 /wd4267 /wd4273 /wd4275
/wd4800 /wd4819 /wd4996")
string(REPLACE "/W3" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
string(REPLACE "/MD" "/MT" CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE}")
......
dali/__init__.py
......@@ -22,32 +22,32 @@ import sys as _sys
from dragon.vm.dali._api import ops
# Classes
from dragon.vm.dali.core.iterator import Iterator
from dragon.vm.dali.core.pipeline import Pipeline
from dragon.vm.dali.core.framework.iterator import Iterator
from dragon.vm.dali.core.framework.pipeline import Pipeline
# Functions
from dragon.vm.dali.core.context import device
from dragon.vm.dali.core.context import get_device_type
from dragon.vm.dali.core.context import get_distributed_info
from dragon.vm.dali.core.framework.context import device
from dragon.vm.dali.core.framework.context import get_device_type
from dragon.vm.dali.core.framework.context import get_distributed_info
# Enums
from dragon.vm.dali.core.types import BOOL
from dragon.vm.dali.core.types import BGR
from dragon.vm.dali.core.types import FLOAT
from dragon.vm.dali.core.types import FLOAT32
from dragon.vm.dali.core.types import FLOAT64
from dragon.vm.dali.core.types import INT8
from dragon.vm.dali.core.types import INT32
from dragon.vm.dali.core.types import INT64
from dragon.vm.dali.core.types import INTERP_TRIANGULAR
from dragon.vm.dali.core.types import NCHW
from dragon.vm.dali.core.types import NHWC
from dragon.vm.dali.core.types import RGB
from dragon.vm.dali.core.types import STRING
from dragon.vm.dali.core.types import UINT8
from dragon.vm.dali.core.types import UINT16
from dragon.vm.dali.core.types import UINT32
from dragon.vm.dali.core.types import UINT64
from dragon.vm.dali.core.framework.types import BOOL
from dragon.vm.dali.core.framework.types import BGR
from dragon.vm.dali.core.framework.types import FLOAT
from dragon.vm.dali.core.framework.types import FLOAT32
from dragon.vm.dali.core.framework.types import FLOAT64
from dragon.vm.dali.core.framework.types import INT8
from dragon.vm.dali.core.framework.types import INT32
from dragon.vm.dali.core.framework.types import INT64
from dragon.vm.dali.core.framework.types import INTERP_TRIANGULAR
from dragon.vm.dali.core.framework.types import NCHW
from dragon.vm.dali.core.framework.types import NHWC
from dragon.vm.dali.core.framework.types import RGB
from dragon.vm.dali.core.framework.types import STRING
from dragon.vm.dali.core.framework.types import UINT8
from dragon.vm.dali.core.framework.types import UINT16
from dragon.vm.dali.core.framework.types import UINT32
from dragon.vm.dali.core.framework.types import UINT64
# Attributes
_API_MODULE = ops
......
dali/_api/ops/__init__.py
......@@ -13,24 +13,24 @@ from __future__ import absolute_import as _absolute_import
from __future__ import division as _division
from __future__ import print_function as _print_function
from dragon.vm.dali.core.ops.array import Cast
from dragon.vm.dali.core.ops.array import Pad
from dragon.vm.dali.core.ops.array import Reshape
from dragon.vm.dali.core.ops.builtin import ExternalSource
from dragon.vm.dali.core.ops.color import BrightnessContrast
from dragon.vm.dali.core.ops.color import Hsv
from dragon.vm.dali.core.ops.crop import RandomBBoxCrop
from dragon.vm.dali.core.ops.crop import Slice
from dragon.vm.dali.core.ops.decoder import ImageDecoder
from dragon.vm.dali.core.ops.decoder import ImageDecoderRandomCrop
from dragon.vm.dali.core.ops.fused import CropMirrorNormalize
from dragon.vm.dali.core.ops.geometric import BbFlip
from dragon.vm.dali.core.ops.paste import BBoxPaste
from dragon.vm.dali.core.ops.paste import Paste
from dragon.vm.dali.core.ops.random import CoinFlip
from dragon.vm.dali.core.ops.random import Uniform
from dragon.vm.dali.core.ops.reader import KPLRecordReader
from dragon.vm.dali.core.ops.reader import TFRecordReader
from dragon.vm.dali.core.ops.resize import Resize
from dragon.vm.dali.core.ops.bbox_ops import BbFlip
from dragon.vm.dali.core.ops.bbox_ops import BBoxPaste
from dragon.vm.dali.core.ops.builtin_ops import ExternalSource
from dragon.vm.dali.core.ops.decoder_ops import ImageDecoder
from dragon.vm.dali.core.ops.decoder_ops import ImageDecoderRandomCrop
from dragon.vm.dali.core.ops.generic_ops import Cast
from dragon.vm.dali.core.ops.generic_ops import Pad
from dragon.vm.dali.core.ops.generic_ops import Reshape
from dragon.vm.dali.core.ops.generic_ops import Slice
from dragon.vm.dali.core.ops.image_ops import BrightnessContrast
from dragon.vm.dali.core.ops.image_ops import CropMirrorNormalize
from dragon.vm.dali.core.ops.image_ops import Hsv
from dragon.vm.dali.core.ops.image_ops import Paste
from dragon.vm.dali.core.ops.image_ops import RandomBBoxCrop
from dragon.vm.dali.core.ops.image_ops import Resize
from dragon.vm.dali.core.ops.random_ops import CoinFlip
from dragon.vm.dali.core.ops.random_ops import Uniform
from dragon.vm.dali.core.ops.reader_ops import KPLRecordReader
from dragon.vm.dali.core.ops.reader_ops import TFRecordReader
__all__ = [_s for _s in dir() if not _s.startswith('_')]
dali/core/framework/__init__.py 0 → 100644
# ------------------------------------------------------------
# Copyright (c) 2017-present, SeetaTech, Co.,Ltd.
#
# Licensed under the BSD 2-Clause License.
# You should have received a copy of the BSD 2-Clause License
# along with the software. If not, See,
#
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
dali/core/iterator.py → dali/core/framework/iterator.py
......@@ -26,7 +26,7 @@ from dragon.core.device import cuda
from dragon.core.eager.tensor import EagerTensor
from dragon.core.framework import device_spec
from dragon.core.framework import workspace
from dragon.vm.dali.core import types
from dragon.vm.dali.core.framework import types
class Iterator(object):
......
dali/core/pipeline.py → dali/core/framework/pipeline.py
......@@ -15,7 +15,7 @@ from __future__ import print_function
try:
from nvidia.dali import pipeline
from dragon.vm.dali.core import context
from dragon.vm.dali.core.framework import context
class Pipeline(pipeline.Pipeline):
"""The base pipeline class to define operations.
......@@ -151,6 +151,8 @@ except ImportError:
"""
self._batch_size = batch_size
self._num_threads = num_threads
self._seed = seed
self._prefetch_queue_depth = prefetch_queue_depth
@property
def batch_size(self):
......
dali/core/ops/paste.py → dali/core/ops/bbox_ops.py
......@@ -16,85 +16,65 @@ from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
ops = None
from dragon.core.util import deprecation
ops = deprecation.not_installed('nvidia.dali')
from dragon.vm.dali.core import context
from dragon.vm.dali.core.framework import context
class BBoxPaste(object):
"""Transform bounding boxes to match the ``Paste`` operator.
class BbFlip(object):
"""Flip the bounding boxes.
Examples:
```python
bbox_paste = dali.ops.BBoxPaste()
paste_pos = dali.ops.Uniform((0., 1.))
paste_ratio = dali.ops.Uniform((0., 3.))
paste_prob = dali.ops.CoinFlip(0.5)
bbox_flip = dali.ops.BbFlip()
flip_rng = dali.ops.CoinFlip(0.5)
bbox = bbox_flip(inputs['bbox'], horizontal=flip_rng())
```
bbox = bbox_paste(
inputs['bbox'],
# Expand ratio
ratio=paste_ratio() * paste_prob() + 1.,
# PosX, PosY
paste_x=paste_pos(),
paste_y=paste_pos(),
)
"""
def __new__(
cls,
ltrb=True,
ratio=None,
paste_x=None,
paste_y=None,
):
"""Create a ``BBoxPaste`` operator.
def __new__(cls, horizontal=None, vertical=None, ltrb=True, **kwargs):
"""Create a ``BbFlip`` operator.
Parameters
----------
horizontal : int, optional
Whether to apply the horizontal flip.
vertical : int, optional
Whether to apply the vertical flip.
ltrb : bool, optional, default=True
Indicate the bbox is ``ltrb`` or ``xywh`` format.
ratio : int, optional
The expand ratio.
paste_x : int, optional
The paste position at x-axis.
paste_y : int, optional
The paste position at y-axis.
Returns
-------
nvidia.dali.ops.BBoxPaste
nvidia.dali.ops.BbFlip
The operator.
"""
return ops.BBoxPaste(
return ops.BbFlip(
horizontal=horizontal,
vertical=vertical,
ltrb=ltrb,
ratio=ratio,
paste_x=paste_x,
paste_y=paste_y,
device='cpu',
device=context.get_device_type(),
**kwargs
)
class Paste(object):
"""Copy image into a larger canvas.
class BBoxPaste(object):
"""Transform bounding boxes to match the ``Paste`` operator.
Examples:
```python
paste = dali.ops.Paste(
# The image channels
n_channels=3,
# Historical values before mean subtraction
fill_value=(102., 115., 122.),
)
bbox_paste = dali.ops.BBoxPaste()
paste_pos = dali.ops.Uniform((0., 1.))
paste_ratio = dali.ops.Uniform((0., 3.))
paste_prob = dali.ops.CoinFlip(0.5)
y = paste(
inputs['x'],
bbox = bbox_paste(
inputs['bbox'],
# Expand ratio
ratio=paste_ratio() * paste_prob() + 1.,
# PosX, PosY
......@@ -107,20 +87,18 @@ class Paste(object):
def __new__(
cls,
n_channels=3,
fill_value=(0., 0., 0.),
ltrb=True,
ratio=None,
paste_x=None,
paste_y=None,
**kwargs
):
"""Create a ``Paste`` operator.
"""Create a ``BBoxPaste`` operator.
Parameters
----------
n_channels : int, optional, default=3
The image channels.
fill_value : Sequence[number], optional
The value(s) to fill for the canvas.
ltrb : bool, optional, default=True
Indicate the bbox is ``ltrb`` or ``xywh`` format.
ratio : int, optional
The expand ratio.
paste_x : int, optional
......@@ -130,15 +108,15 @@ class Paste(object):
Returns
-------
nvidia.dali.ops.Paste
nvidia.dali.ops.BBoxPaste
The operator.
"""
return ops.Paste(
n_channels=n_channels,
fill_value=fill_value,
return ops.BBoxPaste(
ltrb=ltrb,
ratio=ratio,
paste_x=paste_x,
paste_y=paste_y,
device=context.get_device_type(),
device='cpu',
**kwargs
)
dali/core/ops/builtin.py → dali/core/ops/builtin_ops.py
......@@ -16,9 +16,10 @@ from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
ops = None
from dragon.core.util import deprecation
ops = deprecation.not_installed('nvidia.dali')
from dragon.vm.dali.core import context
from dragon.vm.dali.core.framework import context
class ExternalSource(object):
......@@ -42,7 +43,7 @@ class ExternalSource(object):
"""
def __new__(cls):
def __new__(cls, **kwargs):
"""Create a ``ExternalSource`` operator.
Returns
......@@ -51,4 +52,4 @@ class ExternalSource(object):
The operator.
"""
return ops.ExternalSource(device=context.get_device_type())
return ops.ExternalSource(device=context.get_device_type(), **kwargs)
dali/core/ops/color.py deleted 100644 → 0
# ------------------------------------------------------------
# Copyright (c) 2017-present, SeetaTech, Co.,Ltd.
#
# Licensed under the BSD 2-Clause License.
# You should have received a copy of the BSD 2-Clause License
# along with the software. If not, See,
#
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
ops = None
from dragon.vm.dali.core import context
class BrightnessContrast(object):
"""Adjust the brightness and contrast.
Examples:
```python
# Historical jitter range for brightness and contrast
twist_rng = dali.ops.Uniform(range=[0.6, 1.4])
bc = dali.ops.BrightnessContrast()
y = bc(inputs['x'], brightness=twist_rng(), contrast=twist_rng())
```
"""
def __new__(cls):
"""Create a ``BrightnessContrastBrightnessContrast`` operator.
Returns
-------
nvidia.dali.ops.BrightnessContrast
The operator.
"""
return ops.BrightnessContrast(device=context.get_device_type())
class Hsv(object):
"""Adjust the hue and saturation.
Examples:
```python
# Historical jitter range for saturation
twist_rng = dali.ops.Uniform(range=[0.6, 1.4])
hsv = dali.ops.Hsv()
y = hsv(inputs['x'], saturation=twist_rng())
```
"""
def __new__(cls):
"""Create a ``Hsv`` operator.
Returns
-------
nvidia.dali.ops.Hsv
The operator.
"""
return ops.Hsv(device=context.get_device_type())
dali/core/ops/crop.py deleted 100644 → 0
# ------------------------------------------------------------
# Copyright (c) 2017-present, SeetaTech, Co.,Ltd.
#
# Licensed under the BSD 2-Clause License.
# You should have received a copy of the BSD 2-Clause License
# along with the software. If not, See,
#
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
ops = None
from dragon.vm.dali.core import context
class RandomBBoxCrop(object):
"""Return a valid crop restricted by bounding boxes.
Examples:
```python
bbox_crop = dali.ops.RandomBBoxCrop(
# Range of scale
scaling=[0.3, 1.0],
# Range of aspect ratio
aspect_ratio=[0.5, 2.0],
# Minimum IoUs to satisfy
thresholds=[0.0, 0.1, 0.3, 0.5, 0.7, 0.9],
)
crop_begin, crop_size, bbox, label = bbox_crop(inputs['bbox'], inputs['label'])
```
"""
def __new__(
cls,
scaling=(0.3, 1.0),
aspect_ratio=(0.5, 2.0),
thresholds=(0.0, 0.1, 0.3, 0.5, 0.7, 0.9),
allow_no_crop=True,
ltrb=True,
num_attempts=10,
):
"""Create a ``RandomBBoxCrop`` operator.
Parameters
----------
scaling : Sequence[float], optional, default=(0.3, 1.0)
The range of scale for sampling regions.
aspect_ratio : Sequence[float], optional, default=(0.5, 2.0)
The range of aspect ratio for sampling regions.
thresholds : Sequence[float], optional
The minimum IoU(s) to satisfy.
allow_no_crop : bool, optional, default=True
**True** to include the no-cropping as a option.
ltrb : bool, optional, default=True
Indicate the bbox is ``ltrb`` or ``xywh`` format.
num_attempts : int, optional, default=10
The max number of sampling trails.
Returns
-------
nvidia.dali.ops.RandomBBoxCrop
The operator.
"""
return ops.RandomBBoxCrop(
scaling=scaling,
aspect_ratio=aspect_ratio,
thresholds=thresholds,
allow_no_crop=allow_no_crop,
ltrb=ltrb,
num_attempts=num_attempts,
device='cpu',
)
class Slice(object):
"""Select an interval of elements from input.
Examples:
```python
slice = dali.ops.Slice(
# Axis of intervals
axes=[1, 0],
# Whether the begin of interval is normalized
# in a range of [0.0, 1.0]
normalized_anchor=True,
# Whether the size of interval is normalized
# in a range of [0.0, 1.0]
normalized_shape=True,
)
y = slice(inputs['x'], crop_begin, crop_size)
```
"""
def __new__(
cls,
axes=(1, 0),
normalized_anchor=True,
normalized_shape=True,
):
"""Create a ``Slice`` operator.
Parameters
----------
axes : Sequence[int], optional
The axis to select.
normalized_anchor : bool, optional, default=True
Whether the begin of interval is normalized.
normalized_shape : bool, optional, default=True
Whether the size of interval is normalized.
Returns
-------
nvidia.dali.ops.Slice
The operator.
"""
return ops.Slice(
axes=axes,
normalized_anchor=normalized_anchor,
device=context.get_device_type(),
)
dali/core/ops/decoder.py → dali/core/ops/decoder_ops.py
......@@ -16,11 +16,12 @@ from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
ops = None
from dragon.core.util import deprecation
ops = deprecation.not_installed('nvidia.dali')
from dragon.core.util import six
from dragon.vm.dali.core import context
from dragon.vm.dali.core import types
from dragon.vm.dali.core.framework import context
from dragon.vm.dali.core.framework import types
class ImageDecoder(object):
......@@ -40,6 +41,7 @@ class ImageDecoder(object):
output_type='BGR',
host_memory_padding=8388608,
device_memory_padding=16777216,
**kwargs
):
"""Create a ``ImageDecoder`` operator.
......@@ -65,6 +67,7 @@ class ImageDecoder(object):
host_memory_padding=host_memory_padding,
device_memory_padding=device_memory_padding,
device=context.get_device_type(mixed=True),
**kwargs
)
......@@ -93,6 +96,7 @@ class ImageDecoderRandomCrop(object):
random_area=(0.08, 1.),
random_aspect_ratio=(0.75, 1.33),
num_attempts=10,
**kwargs
):
"""Create a ``ImageDecoderRandomCrop`` operator.
......@@ -127,4 +131,5 @@ class ImageDecoderRandomCrop(object):
random_aspect_ratio=random_aspect_ratio,
num_attempts=num_attempts,
device=context.get_device_type(mixed=True),
**kwargs
)
dali/core/ops/fused.py deleted 100644 → 0
# ------------------------------------------------------------
# Copyright (c) 2017-present, SeetaTech, Co.,Ltd.
#
# Licensed under the BSD 2-Clause License.
# You should have received a copy of the BSD 2-Clause License
# along with the software. If not, See,
#
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
ops = None
from dragon.core.util import six
from dragon.vm.dali.core import context
from dragon.vm.dali.core import types
class CropMirrorNormalize(object):
"""Crop and normalize input with the horizontal flip.
Examples:
```python
flip_rng = dali.ops.CoinFlip(0.5)
cmn = dali.ops.CropMirrorNormalize(
# Match the number of spatial dims
# (H, W) for 2d input
# (D, H, W) for 3d input
crop=(224, 224),
# Historical values to normalize input
mean=(102., 115., 122.),
std=(1., 1., 1.),
# ``BGR``, ``RGB``, or ``GRAY``
image_type='BGR',
# Or ``float16`` for fp16 training
output_dtype='float32',
# Or ``NHWC``
output_layout='NCHW'
)
y = cmn(inputs['x'], mirror=flip_rng())
```
"""
def __new__(
cls,
crop=None,
mirror=None,
mean=0.,
std=1.,
image_type='BGR',
output_dtype='float32',
output_layout='NCHW',
):
"""Create a ``CropMirrorNormalize`` operator.
Parameters
----------
crop : Sequence[int], optional
The cropped spatial dimensions for output.
mirror : {0, 1}, optional
Whether to apply the horizontal flip.
mean : Union[float, Sequence[float]], optional
The values to subtract.
std : Union[float, Sequence[float]], optional
The values to divide after subtraction.
image_type : {'BGR', 'RGB'}, optional
The color space of input.
output_dtype : {'float16', 'float32'}, optional
The data type of output.
output_layout : {'NCHW', 'NHWC'}, optional
The data format of output.
Returns
-------
nvidia.dali.ops.CropMirrorNormalize
The operator.
"""
if isinstance(output_dtype, six.string_types):
output_dtype = getattr(types, output_dtype.upper())
if isinstance(output_layout, six.string_types):
output_layout = getattr(types, output_layout.upper())
if isinstance(image_type, six.string_types):
image_type = getattr(types, image_type.upper())
return ops.CropMirrorNormalize(
crop=crop,
mirror=mirror,
mean=mean,
std=std,
output_dtype=output_dtype,
output_layout=output_layout,
image_type=image_type,
device=context.get_device_type(),
)
dali/core/ops/array.py → dali/core/ops/generic_ops.py
......@@ -16,11 +16,12 @@ from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
ops = None
from dragon.core.util import deprecation
ops = deprecation.not_installed('nvidia.dali')
from dragon.core.util import six
from dragon.vm.dali.core import context
from dragon.vm.dali.core import types
from dragon.vm.dali.core.framework import context
from dragon.vm.dali.core.framework import types
class Cast(object):
......@@ -35,7 +36,7 @@ class Cast(object):
"""
def __new__(cls, dtype):
def __new__(cls, dtype, **kwargs):
"""Create a ``Cast`` operator.
Parameters
......@@ -54,6 +55,7 @@ class Cast(object):
return ops.Cast(
dtype=dtype,
device=context.get_device_type(),
**kwargs
)
......@@ -74,7 +76,7 @@ class Pad(object):
"""
def __new__(cls, axes=(0, 1), fill_value=0., align=None):
def __new__(cls, axes=(0, 1), fill_value=0., align=None, **kwargs):
"""Create a ``Pad`` operator.
Parameters
......@@ -97,6 +99,7 @@ class Pad(object):
fill_value=fill_value,
align=align,
device=context.get_device_type(),
**kwargs
)
......@@ -117,7 +120,7 @@ class Reshape(object):
"""
def __new__(cls, shape=None):
def __new__(cls, shape=None, **kwargs):
"""Create a ``Reshape`` operator.
Parameters
......@@ -134,4 +137,59 @@ class Reshape(object):
return ops.Reshape(
shape=shape,
device=context.get_device_type(),
**kwargs
)
class Slice(object):
"""Select an interval of elements from input.
Examples:
```python
slice = dali.ops.Slice(
# Axis of intervals
axes=[1, 0],
# Whether the begin of interval is normalized
# in a range of [0.0, 1.0]
normalized_anchor=True,
# Whether the size of interval is normalized
# in a range of [0.0, 1.0]
normalized_shape=True,
)
y = slice(inputs['x'], crop_begin, crop_size)
```
"""
def __new__(
cls,
axes=(1, 0),
normalized_anchor=True,
normalized_shape=True,
**kwargs
):
"""Create a ``Slice`` operator.
Parameters
----------
axes : Sequence[int], optional
The axis to select.
normalized_anchor : bool, optional, default=True
Whether the begin of interval is normalized.
normalized_shape : bool, optional, default=True
Whether the size of interval is normalized.
Returns
-------
nvidia.dali.ops.Slice
The operator.
"""
return ops.Slice(
axes=axes,
normalized_anchor=normalized_anchor,
device=context.get_device_type(),
**kwargs
)
dali/core/ops/geometric.py deleted 100644 → 0
# ------------------------------------------------------------
# Copyright (c) 2017-present, SeetaTech, Co.,Ltd.
#
# Licensed under the BSD 2-Clause License.
# You should have received a copy of the BSD 2-Clause License
# along with the software. If not, See,
#
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
ops = None
from dragon.vm.dali.core import context
class BbFlip(object):
"""Flip the bounding boxes.
Examples:
```python
bbox_flip = dali.ops.BbFlip()
flip_rng = dali.ops.CoinFlip(0.5)
bbox = bbox_flip(inputs['bbox'], horizontal=flip_rng())
```
"""
def __new__(cls, horizontal=None, vertical=None, ltrb=True):
"""Create a ``BbFlip`` operator.
Parameters
----------
horizontal : int, optional
Whether to apply the horizontal flip.
vertical : int, optional
Whether to apply the vertical flip.
ltrb : bool, optional, default=True
Indicate the bbox is ``ltrb`` or ``xywh`` format.
Returns
-------
nvidia.dali.ops.BbFlip
The operator.
"""
return ops.BbFlip(
horizontal=horizontal,
vertical=vertical,
ltrb=ltrb,
device=context.get_device_type(),
)
dali/core/ops/image_ops.py 0 → 100644
# ------------------------------------------------------------
# Copyright (c) 2017-present, SeetaTech, Co.,Ltd.
#
# Licensed under the BSD 2-Clause License.
# You should have received a copy of the BSD 2-Clause License
# along with the software. If not, See,
#
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
from dragon.core.util import deprecation
ops = deprecation.not_installed('nvidia.dali')
from dragon.core.util import six
from dragon.vm.dali.core.framework import context
from dragon.vm.dali.core.framework import types
class BrightnessContrast(object):
"""Adjust the brightness and contrast of image.
Examples:
```python
# Historical jitter range for brightness and contrast
twist_rng = dali.ops.Uniform(range=[0.6, 1.4])
bc = dali.ops.BrightnessContrast()
y = bc(inputs['x'], brightness=twist_rng(), contrast=twist_rng())
```
"""
def __new__(cls, **kwargs):
"""Create a ``BrightnessContrastBrightnessContrast`` operator.
Returns
-------
nvidia.dali.ops.BrightnessContrast
The operator.
"""
return ops.BrightnessContrast(device=context.get_device_type(), **kwargs)
class CropMirrorNormalize(object):
"""Crop and normalize image with the horizontal flip.
Examples:
```python
flip_rng = dali.ops.CoinFlip(0.5)
cmn = dali.ops.CropMirrorNormalize(
# Match the number of spatial dims
# (H, W) for 2d input
# (D, H, W) for 3d input
crop=(224, 224),
# Historical values to normalize input
mean=(102., 115., 122.),
std=(1., 1., 1.),
# Or ``float16`` for fp16 training
dtype='float32',
# Or ``NHWC``
output_layout='NCHW'
)
y = cmn(inputs['x'], mirror=flip_rng())
```
"""
def __new__(
cls,
crop=None,
mirror=None,
mean=0.,
std=1.,
dtype='float32',
output_layout='NCHW',
**kwargs
):
"""Create a ``CropMirrorNormalize`` operator.
Parameters
----------
crop : Sequence[int], optional
The cropped spatial dimensions for output.
mirror : {0, 1}, optional
Whether to apply the horizontal flip.
mean : Union[float, Sequence[float]], optional
The values to subtract.
std : Union[float, Sequence[float]], optional
The values to divide after subtraction.
dtype : {'float16', 'float32'}, optional
The data type of output.
output_layout : {'NCHW', 'NHWC'}, optional
The data format of output.
Returns
-------
nvidia.dali.ops.CropMirrorNormalize
The operator.
"""
if isinstance(dtype, six.string_types):
dtype = getattr(types, dtype.upper())
if isinstance(output_layout, six.string_types):
output_layout = getattr(types, output_layout.upper())
return ops.CropMirrorNormalize(
crop=crop,
mirror=mirror,
mean=mean,
std=std,
dtype=dtype,
output_layout=output_layout,
device=context.get_device_type(),
**kwargs
)
class Hsv(object):
"""Adjust the hue and saturation.
Examples:
```python
# Historical jitter range for saturation
twist_rng = dali.ops.Uniform(range=[0.6, 1.4])
hsv = dali.ops.Hsv()
y = hsv(inputs['x'], saturation=twist_rng())
```
"""
def __new__(cls, **kwargs):
"""Create a ``Hsv`` operator.
Returns
-------
nvidia.dali.ops.Hsv
The operator.
"""
return ops.Hsv(device=context.get_device_type(), **kwargs)
class Paste(object):
"""Copy image into a larger canvas.
Examples:
```python
paste = dali.ops.Paste(
# The image channels
n_channels=3,
# Historical values before mean subtraction
fill_value=(102., 115., 122.),
)
paste_pos = dali.ops.Uniform((0., 1.))
paste_ratio = dali.ops.Uniform((0., 3.))
paste_prob = dali.ops.CoinFlip(0.5)
y = paste(
inputs['x'],
# Expand ratio
ratio=paste_ratio() * paste_prob() + 1.,
# PosX, PosY
paste_x=paste_pos(),
paste_y=paste_pos(),
)
```
"""
def __new__(
cls,
n_channels=3,
fill_value=(0., 0., 0.),
ratio=None,
paste_x=None,
paste_y=None,
**kwargs
):
"""Create a ``Paste`` operator.
Parameters
----------
n_channels : int, optional, default=3
The image channels.
fill_value : Sequence[number], optional
The value(s) to fill for the canvas.
ratio : int, optional
The expand ratio.
paste_x : int, optional
The paste position at x-axis.
paste_y : int, optional
The paste position at y-axis.
Returns
-------
nvidia.dali.ops.Paste
The operator.
"""
return ops.Paste(
n_channels=n_channels,
fill_value=fill_value,
ratio=ratio,
paste_x=paste_x,
paste_y=paste_y,
device=context.get_device_type(),
**kwargs
)
class RandomBBoxCrop(object):
"""Return an valid image crop restricted by bounding boxes.
Examples:
```python
bbox_crop = dali.ops.RandomBBoxCrop(
# Range of scale
scaling=[0.3, 1.0],
# Range of aspect ratio
aspect_ratio=[0.5, 2.0],
# Minimum IoUs to satisfy
thresholds=[0.0, 0.1, 0.3, 0.5, 0.7, 0.9],
)
crop_begin, crop_size, bbox, label = bbox_crop(inputs['bbox'], inputs['label'])
```
"""
def __new__(
cls,
scaling=(0.3, 1.0),
aspect_ratio=(0.5, 2.0),
thresholds=(0.0, 0.1, 0.3, 0.5, 0.7, 0.9),
allow_no_crop=True,
ltrb=True,
num_attempts=10,
**kwargs
):
"""Create a ``RandomBBoxCrop`` operator.
Parameters
----------
scaling : Sequence[float], optional, default=(0.3, 1.0)
The range of scale for sampling regions.
aspect_ratio : Sequence[float], optional, default=(0.5, 2.0)
The range of aspect ratio for sampling regions.
thresholds : Sequence[float], optional
The minimum IoU(s) to satisfy.
allow_no_crop : bool, optional, default=True
**True** to include the no-cropping as a option.
ltrb : bool, optional, default=True
Indicate the bbox is ``ltrb`` or ``xywh`` format.
num_attempts : int, optional, default=10
The max number of sampling trails.
Returns
-------
nvidia.dali.ops.RandomBBoxCrop
The operator.
"""
return ops.RandomBBoxCrop(
scaling=scaling,
aspect_ratio=aspect_ratio,
thresholds=thresholds,
allow_no_crop=allow_no_crop,
ltrb=ltrb,
num_attempts=num_attempts,
device='cpu',
**kwargs
)
class Resize(object):
"""Resize the image.
Examples:
```python
# Resize to a fixed area
resize1 = dali.ops.Resize(resize_x=300, resize_y=300)
# Resize along the shorter side
resize2 = dali.ops.Resize(resize_shorter=600, max_size=1000)
# Resize along the longer side
resize3 = dali.ops.Resize(resize_longer=512)
```
"""
def __new__(
cls,
resize_x=None,
resize_y=None,
resize_shorter=None,
resize_longer=None,
max_size=None,
interp_type='TRIANGULAR',
):
"""Create a ``Resize`` operator.
Parameters
----------
resize_x : int, optional
The output image width.
resize_y : int, optional
The output image height.
resize_shorter : int, optional
Resize along the shorter side and limited by ``max_size``.
resize_longer : int, optional
Resize along the longer side.
max_size : int, optional, default=0
The limited size for ``resize_shorter``.
interp_type : {'NN', 'LINEAR', 'TRIANGULAR', 'CUBIC', 'GAUSSIAN', 'LANCZOS3'}, optional
The interpolation method.
"""
if isinstance(interp_type, six.string_types):
interp_type = getattr(types, 'INTERP_' + interp_type.upper())
return ops.Resize(
resize_x=resize_x,
resize_y=resize_y,
resize_shorter=resize_shorter,
resize_longer=resize_longer,
max_size=max_size,
interp_type=interp_type,
device=context.get_device_type(),
)
dali/core/ops/random.py → dali/core/ops/random_ops.py
......@@ -16,7 +16,8 @@ from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
ops = None
from dragon.core.util import deprecation
ops = deprecation.not_installed('nvidia.dali')
class CoinFlip(object):
......@@ -31,7 +32,7 @@ class CoinFlip(object):
"""
def __new__(cls, probability=0.5):
def __new__(cls, probability=0.5, **kwargs):
"""Create a ``CoinFlip`` operator.
Parameters
......@@ -45,11 +46,11 @@ class CoinFlip(object):
The operator.
"""
return ops.CoinFlip(probability=probability)
return ops.CoinFlip(probability=probability, **kwargs)
class Uniform(object):
"""Sample values from a uniform distribution.
"""Sample values from an uniform distribution.
Examples:
......@@ -60,7 +61,7 @@ class Uniform(object):
"""
def __new__(cls, range=(-1., 1.)):
def __new__(cls, range=(-1., 1.), **kwargs):
"""Create an ``Uniform`` operator.
Parameters
......@@ -74,4 +75,4 @@ class Uniform(object):
The operator.
"""
return ops.Uniform(range=range)
return ops.Uniform(range=range, **kwargs)
dali/core/ops/reader.py → dali/core/ops/reader_ops.py
......@@ -19,16 +19,16 @@ import os
try:
from nvidia.dali import ops
from nvidia.dali import tfrecord as tfrec
from nvidia.dali import tfrecord
except ImportError:
from dragon.core.util import deprecation
ops = deprecation.NotInstalled('nvidia.dali')
tfrec = deprecation.NotInstalled('nvidia.dali')
tfrecord = deprecation.NotInstalled('nvidia.dali')
from dragon.core.io import reader
from dragon.core.io import kpl_record
from dragon.vm.dali.core import context
from dragon.vm.dali.core.ops.builtin import ExternalSource
from dragon.vm.dali.core.framework import context
from dragon.vm.dali.core.ops.builtin_ops import ExternalSource
class KPLRecordReader(object):
......@@ -73,6 +73,7 @@ class KPLRecordReader(object):
num_shards=1,
shuffle_after_epoch=False,
shuffle_chunks=0,
**kwargs
):
"""Create a ``KPLRecordReader``.
......@@ -100,6 +101,7 @@ class KPLRecordReader(object):
num_parts=num_shards,
shuffle=shuffle_after_epoch,
num_chunks=shuffle_chunks,
**kwargs
)
self._buffer = self._reader.q_out = mp.Queue(
self._prefetch_depth * self._batch_size)
......@@ -186,6 +188,7 @@ class TFRecordReader(object):
num_shards=1,
random_shuffle=False,
initial_fill=1024,
**kwargs
):
"""Create a ``TFRecordReader``.
......@@ -217,6 +220,7 @@ class TFRecordReader(object):
features=features,
random_shuffle=random_shuffle,
initial_fill=initial_fill,
**kwargs
)
@staticmethod
......@@ -232,7 +236,10 @@ class TFRecordReader(object):
if features_file is None:
raise FileNotFoundError('File <FEATURES> is missing.')
with open(os.path.join(path, features_file), 'r') as f:
features = eval(f.read().replace('tf.', 'tfrec.'))
features = f.read()
features = features.replace('tf.', 'tfrecord.')
features = features.replace('tf.io.', 'tfrecord.')
features = eval(features)
data_files.sort()
index_files.sort()
data = [os.path.join(path, e) for e in data_files]
......
dali/core/ops/resize.py deleted 100644 → 0
# ------------------------------------------------------------
# Copyright (c) 2017-present, SeetaTech, Co.,Ltd.
#
# Licensed under the BSD 2-Clause License.
# You should have received a copy of the BSD 2-Clause License
# along with the software. If not, See,
#
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
try:
from nvidia.dali import ops
except ImportError:
ops = None
from dragon.core.util import six
from dragon.vm.dali.core import context
from dragon.vm.dali.core import types
class Resize(object):
"""Resize the image.
Examples:
```python
# Resize to a fixed area
resize1 = dali.ops.Resize(resize_x=300, resize_y=300)
# Resize along the shorter side
resize2 = dali.ops.Resize(resize_shorter=600, max_size=1000)
# Resize along the longer side
resize3 = dali.ops.Resize(resize_longer=512)
```
"""
def __new__(
cls,
resize_x=None,
resize_y=None,
resize_shorter=None,
resize_longer=None,
max_size=None,
interp_type='TRIANGULAR',
):
"""Create a ``Resize`` operator.
Parameters
----------
resize_x : int, optional
The output image width.
resize_y : int, optional
The output image height.
resize_shorter : int, optional
Resize along the shorter side and limited by ``max_size``.
resize_longer : int, optional
Resize along the longer side.
max_size : int, optional, default=0
The limited size for ``resize_shorter``.
interp_type : {'NN', 'LINEAR', 'TRIANGULAR', 'CUBIC', 'GAUSSIAN', 'LANCZOS3'}, optional
The interpolation method.
"""
if isinstance(interp_type, six.string_types):
interp_type = getattr(types, 'INTERP_' + interp_type.upper())
return ops.Resize(
resize_x=resize_x,
resize_y=resize_y,
resize_shorter=resize_shorter,
resize_longer=resize_longer,
max_size=max_size,
interp_type=interp_type,
device=context.get_device_type(),
)
docs/api/cc/conf.py
......@@ -35,14 +35,7 @@ master_doc = 'index'
source_suffix = '.rst'
# Extension
extensions = [
'sphinx.ext.autodoc',
'sphinx.ext.viewcode',
'sphinx.ext.napoleon',
'sphinxcontrib.katex',
'breathe',
]
napoleon_use_rtype = False
extensions = ['sphinx.ext.autodoc', 'sphinxcontrib.katex', 'breathe']
# Project
project = 'dragon'
......
docs/api/python/Makefile
......@@ -22,7 +22,6 @@ NUMBER_OF_PROCESSORS:=$(shell getconf _NPROCESSORS_ONLN)
help:
@echo "Please use \`make <target>' where <target> is one of"
@echo " html to make standalone HTML files"
@echo " debughtml to make debugging HTML files"
@echo " latex to make LaTeX files, you can set PAPER=a4 or PAPER=letter"
@echo " latexpdf to make LaTeX files and run them through pdflatex"
......@@ -30,11 +29,6 @@ clean:
rm -rf $(BUILDDIR)/*
html:
$(SPHINXBUILD) -b html $(ALLSPHINXOPTS) $(BUILDDIR)
@echo
@echo "Build finished. The HTML pages are in $(BUILDDIR)."
debughtml:
$(SPHINXBUILD) -b html -j ${NUMBER_OF_PROCESSORS} $(ALLSPHINXOPTS) $(BUILDDIR)
@echo
@echo "Build finished. The HTML pages are in $(BUILDDIR)."
......
docs/api/python/caffe.rst
......@@ -25,6 +25,9 @@ vm.caffe
: The Momentum-SGD solver.
`[Polyak, 1964] <https://doi.org/10.1016/0041-5553(64)90137-5>`_.
`class Solver <caffe/Solver.html>`_
: The base solver class to optimize parameters.
.. toctree::
:hidden:
......@@ -33,6 +36,7 @@ vm.caffe
caffe/Net
caffe/RMSPropSolver
caffe/SGDSolver
caffe/Solver
.. raw:: html
......
docs/api/python/caffe/AdamSolver.rst
......@@ -12,30 +12,36 @@ Properties
base_lr
#######
.. autoattribute:: dragon.vm.caffe.AdamSolver.base_lr
.. autoattribute:: dragon.vm.caffe.Solver.base_lr
:noindex:
iter
####
.. autoattribute:: dragon.vm.caffe.AdamSolver.iter
.. autoattribute:: dragon.vm.caffe.Solver.iter
:noindex:
net
###
.. autoattribute:: dragon.vm.caffe.AdamSolver.net
.. autoattribute:: dragon.vm.caffe.Solver.net
:noindex:
test_nets
#########
.. autoattribute:: dragon.vm.caffe.AdamSolver.test_nets
.. autoattribute:: dragon.vm.caffe.Solver.test_nets
:noindex:
Methods
-------
snapshot
########
.. automethod:: dragon.vm.caffe.AdamSolver.snapshot
.. automethod:: dragon.vm.caffe.Solver.snapshot
:noindex:
step
########
.. automethod:: dragon.vm.caffe.AdamSolver.step
.. automethod:: dragon.vm.caffe.Solver.step
:noindex:
.. raw:: html
......
docs/api/python/caffe/NesterovSolver.rst
......@@ -12,30 +12,36 @@ Properties
base_lr
#######
.. autoattribute:: dragon.vm.caffe.NesterovSolver.base_lr
.. autoattribute:: dragon.vm.caffe.Solver.base_lr
:noindex:
iter
####
.. autoattribute:: dragon.vm.caffe.NesterovSolver.iter
.. autoattribute:: dragon.vm.caffe.Solver.iter
:noindex:
net
###
.. autoattribute:: dragon.vm.caffe.NesterovSolver.net
.. autoattribute:: dragon.vm.caffe.Solver.net
:noindex:
test_nets
#########
.. autoattribute:: dragon.vm.caffe.NesterovSolver.test_nets
.. autoattribute:: dragon.vm.caffe.Solver.test_nets
:noindex:
Methods
-------
snapshot
########
.. automethod:: dragon.vm.caffe.NesterovSolver.snapshot
.. automethod:: dragon.vm.caffe.Solver.snapshot
:noindex:
step
########
.. automethod:: dragon.vm.caffe.NesterovSolver.step
.. automethod:: dragon.vm.caffe.Solver.step
:noindex:
.. raw:: html
......
docs/api/python/caffe/RMSPropSolver.rst
......@@ -12,30 +12,36 @@ Properties
base_lr
#######
.. autoattribute:: dragon.vm.caffe.RMSPropSolver.base_lr
.. autoattribute:: dragon.vm.caffe.Solver.base_lr
:noindex:
iter
####
.. autoattribute:: dragon.vm.caffe.RMSPropSolver.iter
.. autoattribute:: dragon.vm.caffe.Solver.iter
:noindex:
net
###
.. autoattribute:: dragon.vm.caffe.RMSPropSolver.net
.. autoattribute:: dragon.vm.caffe.Solver.net
:noindex:
test_nets
#########
.. autoattribute:: dragon.vm.caffe.RMSPropSolver.test_nets
.. autoattribute:: dragon.vm.caffe.Solver.test_nets
:noindex:
Methods
-------
snapshot
########
.. automethod:: dragon.vm.caffe.RMSPropSolver.snapshot
.. automethod:: dragon.vm.caffe.Solver.snapshot
:noindex:
step
########
.. automethod:: dragon.vm.caffe.RMSPropSolver.step
.. automethod:: dragon.vm.caffe.Solver.step
:noindex:
.. raw:: html
......
docs/api/python/caffe/SGDSolver.rst
......@@ -12,30 +12,36 @@ Properties
base_lr
#######
.. autoattribute:: dragon.vm.caffe.SGDSolver.base_lr
.. autoattribute:: dragon.vm.caffe.Solver.base_lr
:noindex:
iter
####
.. autoattribute:: dragon.vm.caffe.SGDSolver.iter
.. autoattribute:: dragon.vm.caffe.Solver.iter
:noindex:
net
###
.. autoattribute:: dragon.vm.caffe.SGDSolver.net
.. autoattribute:: dragon.vm.caffe.Solver.net
:noindex:
test_nets
#########
.. autoattribute:: dragon.vm.caffe.SGDSolver.test_nets
.. autoattribute:: dragon.vm.caffe.Solver.test_nets
:noindex:
Methods
-------
snapshot
########
.. automethod:: dragon.vm.caffe.SGDSolver.snapshot
.. automethod:: dragon.vm.caffe.Solver.snapshot
:noindex:
step
########
.. automethod:: dragon.vm.caffe.SGDSolver.step
.. automethod:: dragon.vm.caffe.Solver.step
:noindex:
.. raw:: html
......
docs/api/python/caffe/Solver.rst 0 → 100644
Solver
======
.. autoclass:: dragon.vm.caffe.Solver
__init__
--------
.. automethod:: dragon.vm.caffe.Solver.__init__
Properties
----------
base_lr
#######
.. autoattribute:: dragon.vm.caffe.Solver.base_lr
iter
####
.. autoattribute:: dragon.vm.caffe.Solver.iter
net
###
.. autoattribute:: dragon.vm.caffe.Solver.net
test_nets
#########
.. autoattribute:: dragon.vm.caffe.Solver.test_nets
Methods
-------
snapshot
########
.. automethod:: dragon.vm.caffe.Solver.snapshot
step
########
.. automethod:: dragon.vm.caffe.Solver.step
.. raw:: html
<style>
h1:before {
content: "caffe.";
color: #103d3e;
}
</style>
docs/api/python/conf.py
......@@ -33,9 +33,9 @@ source_suffix = '.rst'
# Extension
extensions = [
'sphinx.ext.autodoc',
'sphinx.ext.viewcode',
'sphinx.ext.napoleon',
'sphinxcontrib.katex',
'sphinx_seeta_theme.ext.viewcode',
]
napoleon_use_rtype = False
......
docs/api/python/dali/ops.rst
......@@ -13,19 +13,22 @@ vm.dali.ops
: Transform bounding boxes to match the ``Paste`` operator.
`class BrightnessContrast <ops/BrightnessContrast.html>`_
: Adjust the brightness and contrast.
: Adjust the brightness and contrast of image.
`class Cast <ops/Cast.html>`_
: Cast the data type of input.
`class CoinFlip <ops/CoinFlip.html>`_
: Sample values from a bernoulli distribution.
`class CropMirrorNormalize <ops/CropMirrorNormalize.html>`_
: Crop and normalize input with the horizontal flip.
: Crop and normalize image with the horizontal flip.
`class ExternalSource <ops/Cast.html>`_
: Create a placeholder providing data from feeding.
`class Hsv <ops/Hsv.html>`_
: Adjust the hue and saturation.
: Adjust the hue and saturation of image.
`class ImageDecoder <ops/ImageDecoder.html>`_
: Decode image from bytes.
......@@ -40,7 +43,7 @@ vm.dali.ops
: Copy image into a larger canvas.
`class RandomBBoxCrop <ops/RandomBBoxCrop.html>`_
: Return a valid crop restricted by bounding boxes.
: Return an valid image crop restricted by bounding boxes.
`class Reshape <ops/Reshape.html>`_
: Change the dimensions of input.
......@@ -58,7 +61,7 @@ vm.dali.ops
: Read examples from the tf-record file.
`class Uniform <ops/Uniform.html>`_
: Select an interval of elements from input.
: Sample values from an uniform distribution.
.. toctree::
:hidden:
......
docs/api/python/dragon/EagerTensor.rst
......@@ -175,8 +175,23 @@ __truediv__
.. _dragon.assign(...): assign.html
.. _dragon.cast(...): cast.html
.. _dragon.copy(...): copy.html
.. _dragon.fill(...): fill.html
.. _dragon.masked_assign(...): masked_assign.html
.. _dragon.masked_select(...): masked_select.html
.. _dragon.math.add(...): math/add.html
.. _dragon.math.div(...): math/div.html
.. _dragon.math.greater(...): math/greater.html
.. _dragon.math.greater_equal(...): math/greater_equal.html
.. _dragon.math.less(...): math/less.html
.. _dragon.math.less_equal(...): math/less_equal.html
.. _dragon.math.mul(...): math/mul.html
.. _dragon.math.negative(...): math/negative.html
.. _dragon.math.sub(...): math/sub.html
.. _dragon.random.glorot_normal(...): random/glorot_normal.html
.. _dragon.random.glorot_uniform(...): random/glorot_uniform.html
.. _dragon.random.normal(...): random/normal.html
.. _dragon.random.truncated_normal(...): random/truncated_normal.html
.. _dragon.random.uniform(...): random/uniform.html
.. _dragon.reshape(...): reshape.html
.. _dragon.slice(...): slice.html
......
docs/api/python/dragon/Tensor.rst
......@@ -155,8 +155,23 @@ __truediv__
.. _dragon.assign(...): assign.html
.. _dragon.cast(...): cast.html
.. _dragon.copy(...): copy.html
.. _dragon.fill(...): fill.html
.. _dragon.masked_assign(...): masked_assign.html
.. _dragon.masked_select(...): masked_select.html
.. _dragon.math.add(...): math/add.html
.. _dragon.math.div(...): math/div.html
.. _dragon.math.greater(...): math/greater.html
.. _dragon.math.greater_equal(...): math/greater_equal.html
.. _dragon.math.less(...): math/less.html
.. _dragon.math.less_equal(...): math/less_equal.html
.. _dragon.math.mul(...): math/mul.html
.. _dragon.math.negative(...): math/negative.html
.. _dragon.math.sub(...): math/sub.html
.. _dragon.random.glorot_normal(...): random/glorot_normal.html
.. _dragon.random.glorot_uniform(...): random/glorot_uniform.html
.. _dragon.random.normal(...): random/normal.html
.. _dragon.random.truncated_normal(...): random/truncated_normal.html
.. _dragon.random.uniform(...): random/uniform.html
.. _dragon.reshape(...): reshape.html
.. _dragon.slice(...): slice.html
......
docs/api/python/make.bat
......@@ -20,7 +20,6 @@ if "%1" == "help" (
:help
echo.Please use `make ^<target^>` where ^<target^> is one of
echo. html to make standalone HTML files
echo. debughtml to make debugging HTML files
echo. latex to make LaTeX files, you can set PAPER=a4 or PAPER=letter
echo. latexpdf to make LaTeX files and run them through pdflatex
goto end
......@@ -56,14 +55,6 @@ if errorlevel 9009 (
:sphinx_ok
if "%1" == "html" (
%SPHINXBUILD% -b html %ALLSPHINXOPTS% %BUILDDIR%
if errorlevel 1 exit /b 1
echo.
echo.Build finished. The HTML pages are in %BUILDDIR%.
goto end
)
if "%1" == "debughtml" (
%SPHINXBUILD% -b html -j %NUMBER_OF_PROCESSORS% %ALLSPHINXOPTS% %BUILDDIR%
if errorlevel 1 exit /b 1
echo.
......
docs/api/python/tensorflow/keras/initializers/GlorotNormal.rst
......@@ -12,7 +12,8 @@ Methods
__call__
########
.. automethod:: dragon.vm.tensorflow.keras.initializers.GlorotNormal.__call__
.. automethod:: dragon.vm.tensorflow.keras.initializers.VarianceScaling.__call__
:noindex:
.. raw:: html
......
docs/api/python/tensorflow/keras/initializers/GlorotUniform.rst
......@@ -12,7 +12,8 @@ Methods
__call__
########
.. automethod:: dragon.vm.tensorflow.keras.initializers.GlorotUniform.__call__
.. automethod:: dragon.vm.tensorflow.keras.initializers.VarianceScaling.__call__
:noindex:
.. raw:: html
......
docs/api/python/torch/Tensor_.rst
......@@ -479,6 +479,7 @@ zero\_
.. _torch.mul(...): mul.html
.. _torch.ne(...): ne.html
.. _torch.neg(...): neg.html
.. _torch.nonzero(...): nonzero.html
.. _torch.pow(...): pow.html
.. _torch.reciprocal(...): reciprocal.html
.. _torch.reshape(...): reshape.html
......@@ -487,8 +488,12 @@ zero\_
.. _torch.sign(...): sign.html
.. _torch.sin(...): sin.html
.. _torch.sqrt(...): sqrt.html
.. _torch.squeeze(...): squeeze.html
.. _torch.sub(...): sub.html
.. _torch.sum(...): sum.html
.. _torch.topk(...): topk.html
.. _torch.unsqueeze(...): unsqueeze.html
.. _torch.where(...): where.html
.. raw:: html
......
docs/api/python/torch/nn/Affine.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.Affine.__init__
.. _torch.nn.functional.affine(...): functional/affine.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/AvgPool2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.AvgPool2d.__init__
.. _torch.nn.functional.avg_pool2d(...): functional/avg_pool2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/BCEWithLogitsLoss.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.BCEWithLogitsLoss.__init__
.. _torch.nn.functional.binary_cross_entropy_with_logits(...): functional/binary_cross_entropy_with_logits.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/BatchNorm1d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.BatchNorm1d.__init__
.. _torch.nn.functional.batch_norm(...): functional/batch_norm.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/BatchNorm2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.BatchNorm2d.__init__
.. _torch.nn.functional.batch_norm(...): functional/batch_norm.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/BatchNorm3d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.BatchNorm3d.__init__
.. _torch.nn.functional.batch_norm(...): functional/batch_norm.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/CTCLoss.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.CTCLoss.__init__
.. _torch.nn.functional.ctc_loss(...): functional/ctc_loss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ConstantPad1d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ConstantPad1d.__init__
.. _torch.nn.functional.pad(...): functional/pad.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ConstantPad2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ConstantPad2d.__init__
.. _torch.nn.functional.pad(...): functional/pad.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ConstantPad3d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ConstantPad3d.__init__
.. _torch.nn.functional.pad(...): functional/pad.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/Conv2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.Conv2d.__init__
.. _torch.nn.functional.conv2d(...): functional/conv2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ConvTranspose2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ConvTranspose2d.__init__
.. _torch.nn.functional.conv_transpose2d(...): functional/conv_transpose2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/CrossEntropyLoss.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.CrossEntropyLoss.__init__
.. _torch.nn.functional.cross_entropy(...): functional/cross_entropy.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/DepthwiseConv2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.DepthwiseConv2d.__init__
.. _torch.nn.functional.depthwise_conv2d(...): functional/depthwise_conv2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/DropBlock2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.DropBlock2d.__init__
.. _torch.nn.functional.drop_block2d(...): functional/drop_block2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/DropPath.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.DropPath.__init__
.. _torch.nn.functional.drop_path(...): functional/drop_path.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/Dropout.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.Dropout.__init__
.. _torch.nn.functional.dropout(...): functional/dropout.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/GroupNorm.rst
......@@ -7,6 +7,7 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.GroupNorm.__init__
.. _torch.nn.functional.group_norm(...): functional/group_norm.html
.. raw:: html
......
docs/api/python/torch/nn/L1Loss.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.L1Loss.__init__
.. _torch.nn.functional.l1_loss(...): functional/l1_loss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/Linear.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.Linear.__init__
.. _torch.nn.functional.linear(...): functional/linear.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/LogSoftmax.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.LogSoftmax.__init__
.. _torch.nn.functional.log_softmax(...): functional/log_softmax.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/MSELoss.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.MSELoss.__init__
.. _torch.nn.functional.mse_loss(...): functional/mse_loss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/MaxPool2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.MaxPool2d.__init__
.. _torch.nn.functional.max_pool2d(...): functional/max_pool2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/NLLLoss.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.NLLLoss.__init__
.. _torch.nn.functional.nll_loss(...): functional/nll_loss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/PReLU.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.PReLU.__init__
.. _torch.nn.functional.prelu(...): functional/prelu.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ReLU.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ReLU.__init__
.. _torch.nn.functional.relu(...): functional/relu.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ReLU6.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ReLU6.__init__
.. _torch.nn.functional.relu6(...): functional/relu6.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ReflectionPad1d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ReflectionPad1d.__init__
.. _torch.nn.functional.pad(...): functional/pad.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ReflectionPad2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ReflectionPad2d.__init__
.. _torch.nn.functional.pad(...): functional/pad.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ReflectionPad3d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ReflectionPad3d.__init__
.. _torch.nn.functional.pad(...): functional/pad.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ReplicationPad1d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ReplicationPad1d.__init__
.. _torch.nn.functional.pad(...): functional/pad.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ReplicationPad2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ReplicationPad2d.__init__
.. _torch.nn.functional.pad(...): functional/pad.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ReplicationPad3d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ReplicationPad3d.__init__
.. _torch.nn.functional.pad(...): functional/pad.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/SELU.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.SELU.__init__
.. _torch.nn.functional.selu(...): functional/selu.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/Sigmoid.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.Sigmoid.__init__
.. _torch.nn.functional.sigmoid(...): functional/sigmoid.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/SigmoidFocalLoss.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.SigmoidFocalLoss.__init__
.. _torch.nn.functional.sigmoid_focal_loss(...): functional/sigmoid_focal_loss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/SmoothL1Loss.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.SmoothL1Loss.__init__
.. _torch.nn.functional.smooth_l1_loss(...): functional/smooth_l1_loss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/Softmax.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.Softmax.__init__
.. _torch.nn.functional.softmax(...): functional/softmax.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/SyncBatchNorm.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.SyncBatchNorm.__init__
.. _torch.nn.functional.batch_norm(...): functional/batch_norm.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/Tanh.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.Tanh.__init__
.. _torch.nn.functional.tanh(...): functional/tanh.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/Upsample.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.Upsample.__init__
.. _torch.nn.functional.interpolate(...): functional/interpolate.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/UpsamplingBilinear2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.UpsamplingBilinear2d.__init__
.. _torch.nn.functional.interpolate(...): functional/interpolate.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/UpsamplingNearest2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.UpsamplingNearest2d.__init__
.. _torch.nn.functional.interpolate(...): functional/interpolate.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/ZeroPad2d.rst
......@@ -7,6 +7,8 @@ __init__
--------
.. automethod:: dragon.vm.torch.nn.ZeroPad2d.__init__
.. _torch.nn.functional.pad(...): functional/pad.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/affine.rst
......@@ -3,6 +3,8 @@ affine
.. autofunction:: dragon.vm.torch.nn.functional.affine
.. _torch.nn.Affine(...): ../Affine.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/avg_pool2d.rst
......@@ -3,6 +3,8 @@ avg_pool2d
.. autofunction:: dragon.vm.torch.nn.functional.avg_pool2d
.. _torch.nn.AvgPool2d(...): ../AvgPool2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/batch_norm.rst
......@@ -3,6 +3,8 @@ batch_norm
.. autofunction:: dragon.vm.torch.nn.functional.batch_norm
.. _torch.nn.BatchNorm2d(...): ../BatchNorm2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/binary_cross_entropy_with_logits.rst
......@@ -3,6 +3,8 @@ binary_cross_entropy_with_logits
.. autofunction:: dragon.vm.torch.nn.functional.binary_cross_entropy_with_logits
.. _torch.nn.BCEWithLogitsLoss(...): ../BCEWithLogitsLoss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/conv2d.rst
......@@ -3,6 +3,8 @@ conv2d
.. autofunction:: dragon.vm.torch.nn.functional.conv2d
.. _torch.nn.Conv2d(...): ../Conv2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/conv_transpose2d.rst
......@@ -3,6 +3,8 @@ conv_transpose2d
.. autofunction:: dragon.vm.torch.nn.functional.conv_transpose2d
.. _torch.nn.ConvTranspose2d(...): ../ConvTranspose2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/cross_entropy.rst
......@@ -3,6 +3,8 @@ cross_entropy
.. autofunction:: dragon.vm.torch.nn.functional.cross_entropy
.. _torch.nn.CrossEntropyLoss(...): ../CrossEntropyLoss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/ctc_loss.rst
......@@ -3,6 +3,8 @@ ctc_loss
.. autofunction:: dragon.vm.torch.nn.functional.ctc_loss
.. _torch.nn.CTCLoss(...): ../CTCLoss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/depthwise_conv2d.rst
......@@ -3,6 +3,8 @@ depthwise_conv2d
.. autofunction:: dragon.vm.torch.nn.functional.depthwise_conv2d
.. _torch.nn.DepthwiseConv2d(...): ../DepthwiseConv2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/drop_block2d.rst
......@@ -3,6 +3,8 @@ drop_block2d
.. autofunction:: dragon.vm.torch.nn.functional.drop_block2d
.. _torch.nn.DropBlock2d(...): ../DropBlock2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/drop_path.rst
......@@ -3,6 +3,8 @@ drop_path
.. autofunction:: dragon.vm.torch.nn.functional.drop_path
.. _torch.nn.DropPath(...): ../DropPath.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/dropout.rst
......@@ -3,6 +3,8 @@ dropout
.. autofunction:: dragon.vm.torch.nn.functional.dropout
.. _torch.nn.Dropout(...): ../Dropout.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/elu.rst
......@@ -3,7 +3,7 @@ elu
.. autofunction:: dragon.vm.torch.nn.functional.elu
.. _torch.nn.ELU: ../ELU.html
.. _torch.nn.ELU(...): ../ELU.html
.. raw:: html
......
docs/api/python/torch/nn/functional/group_norm.rst
......@@ -3,6 +3,8 @@ group_norm
.. autofunction:: dragon.vm.torch.nn.functional.group_norm
.. _torch.nn.GroupNorm(...): ../GroupNorm.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/interpolate.rst
......@@ -3,6 +3,8 @@ interpolate
.. autofunction:: dragon.vm.torch.nn.functional.interpolate
.. _torch.nn.Upsample(...): ../Upsample.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/l1_loss.rst
......@@ -3,6 +3,8 @@ l1_loss
.. autofunction:: dragon.vm.torch.nn.functional.l1_loss
.. _torch.nn.L1Loss(...): ../L1Loss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/leaky_relu.rst
......@@ -3,7 +3,7 @@ leaky_relu
.. autofunction:: dragon.vm.torch.nn.functional.leaky_relu
.. _torch.nn.LeakyReLU: ../LeakyReLU.html
.. _torch.nn.LeakyReLU(...): ../LeakyReLU.html
.. raw:: html
......
docs/api/python/torch/nn/functional/linear.rst
......@@ -3,6 +3,8 @@ linear
.. autofunction:: dragon.vm.torch.nn.functional.linear
.. _torch.nn.Linear(...): ../Linear.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/local_response_norm.rst
......@@ -3,7 +3,7 @@ local_response_norm
.. autofunction:: dragon.vm.torch.nn.functional.local_response_norm
.. _torch.nn.LocalResponseNorm: ../LocalResponseNorm.html
.. _torch.nn.LocalResponseNorm(...): ../LocalResponseNorm.html
.. raw:: html
......
docs/api/python/torch/nn/functional/log_softmax.rst
......@@ -3,6 +3,8 @@ log_softmax
.. autofunction:: dragon.vm.torch.nn.functional.log_softmax
.. _torch.nn.LogSoftmax(...): ../LogSoftmax.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/max_pool2d.rst
......@@ -3,6 +3,8 @@ max_pool2d
.. autofunction:: dragon.vm.torch.nn.functional.max_pool2d
.. _torch.nn.MaxPool2d(...): ../MaxPool2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/mse_loss.rst
......@@ -3,6 +3,8 @@ mse_loss
.. autofunction:: dragon.vm.torch.nn.functional.mse_loss
.. _torch.nn.MSELoss(...): ../MSELoss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/nll_loss.rst
......@@ -3,6 +3,8 @@ nll_loss
.. autofunction:: dragon.vm.torch.nn.functional.nll_loss
.. _torch.nn.NLLLoss(...): ../NLLLoss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/pad.rst
......@@ -3,6 +3,10 @@ pad
.. autofunction:: dragon.vm.torch.nn.functional.pad
.. _torch.nn.ConstantPad2d(...): ../ConstantPad2d.html
.. _torch.nn.ReflectionPad2d(...): ../ReflectionPad2d.html
.. _torch.nn.ReplicationPad2d(...): ../ReplicationPad2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/prelu.rst
......@@ -3,6 +3,8 @@ prelu
.. autofunction:: dragon.vm.torch.nn.functional.prelu
.. _torch.nn.PReLU(...): ../PReLU.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/relu.rst
......@@ -3,6 +3,8 @@ relu
.. autofunction:: dragon.vm.torch.nn.functional.relu
.. _torch.nn.ReLU(...): ../ReLU.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/relu6.rst
......@@ -3,6 +3,8 @@ relu6
.. autofunction:: dragon.vm.torch.nn.functional.relu6
.. _torch.nn.ReLU6(...): ../ReLU6.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/selu.rst
......@@ -3,6 +3,8 @@ selu
.. autofunction:: dragon.vm.torch.nn.functional.selu
.. _torch.nn.SELU(...): ../SELU.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/sigmoid.rst
......@@ -3,6 +3,8 @@ sigmoid
.. autofunction:: dragon.vm.torch.nn.functional.sigmoid
.. _torch.nn.Sigmoid(...): ../Sigmoid.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/sigmoid_focal_loss.rst
......@@ -3,6 +3,8 @@ sigmoid_focal_loss
.. autofunction:: dragon.vm.torch.nn.functional.sigmoid_focal_loss
.. _torch.nn.SigmoidFocalLoss(...): ../SigmoidFocalLoss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/smooth_l1_loss.rst
......@@ -3,6 +3,8 @@ smooth_l1_loss
.. autofunction:: dragon.vm.torch.nn.functional.smooth_l1_loss
.. _torch.nn.SmoothL1Loss(...): ../SmoothL1Loss.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/softmax.rst
......@@ -3,6 +3,8 @@ softmax
.. autofunction:: dragon.vm.torch.nn.functional.softmax
.. _torch.nn.Softmax(...): ../Softmax.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/sync_batch_norm.rst
......@@ -3,6 +3,8 @@ sync_batch_norm
.. autofunction:: dragon.vm.torch.nn.functional.sync_batch_norm
.. _torch.nn.SyncBatchNorm(...): ../SyncBatchNorm.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/tanh.rst
......@@ -3,6 +3,8 @@ tanh
.. autofunction:: dragon.vm.torch.nn.functional.tanh
.. _torch.nn.Tanh(...): ../Tanh.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/upsample.rst
......@@ -3,6 +3,8 @@ upsample
.. autofunction:: dragon.vm.torch.nn.functional.upsample
.. _torch.nn.Upsample(...): ../Upsample.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/upsample_bilinear.rst
......@@ -3,6 +3,8 @@ upsample_bilinear
.. autofunction:: dragon.vm.torch.nn.functional.upsample_bilinear
.. _torch.nn.UpsamplingBilinear2d(...): ../UpsamplingBilinear2d.html
.. raw:: html
<style>
......
docs/api/python/torch/nn/functional/upsample_nearest.rst
......@@ -3,6 +3,8 @@ upsample_nearest
.. autofunction:: dragon.vm.torch.nn.functional.upsample_nearest
.. _torch.nn.UpsamplingNearest2d(...): ../UpsamplingNearest2d.html
.. raw:: html
<style>
......
dragon/core/context_cuda.h
......@@ -40,32 +40,30 @@ class CUDAObjects {
/*! \brief Destructor */
~CUDAObjects() {
for (int i = 0; i < CUDA_MAX_DEVICES; i++) {
for (int j = 0; j < cuda_streams_[i].size(); j++) {
auto& stream = cuda_streams_[i][j];
#ifdef USE_NCCL
for (auto& comm : nccl_comms_[i]) {
/*!
* Do not check the stream destroying,
* error code 29 (driver shutting down) is inevitable.
* Temporarily disable the comm destroying,
* to avoid an unhandled error.
*/
if (stream) cudaStreamDestroy(stream);
}
for (auto& handle : cublas_handles_[i])
if (handle) {
CUBLAS_CHECK(cublasDestroy(handle));
}
#endif
#ifdef USE_CUDNN
for (auto& handle : cudnn_handles_[i])
if (handle) {
CUDNN_CHECK(cudnnDestroy(handle));
for (auto& handle : cudnn_handles_[i]) {
if (handle) CUDNN_CHECK(cudnnDestroy(handle));
}
#endif
#ifdef USE_NCCL
for (auto& comm : nccl_comms_[i]) {
for (auto& handle : cublas_handles_[i]) {
if (handle) CUBLAS_CHECK(cublasDestroy(handle));
}
for (int j = 0; j < cuda_streams_[i].size(); j++) {
auto& stream = cuda_streams_[i][j];
/*!
* Temporarily disable the comm destroying,
* to avoid an unhandled error.
* Do not check the stream destroying,
* error code 29 (driver shutting down) is inevitable.
*/
if (stream) cudaStreamDestroy(stream);
}
#endif
}
}
......
dragon/core/operator.cc
......@@ -179,10 +179,10 @@ OperatorBase* NewOperator(const OperatorDef& def, Workspace* ws) {
}
OperatorDef mutable_def(def);
// Heuristically make each random seed slightly different
static unsigned int op_seed_uuid = 0;
static unsigned int seed_offset = 0;
mutable_def.mutable_device_option()->set_random_seed(
op_seed_uuid + def.device_option().random_seed());
op_seed_uuid = (op_seed_uuid + 1) % UINT32_MAX;
seed_offset + def.device_option().random_seed());
seed_offset = (seed_offset + 1) % UINT32_MAX;
return TryCreateOperator(def.type(), mutable_def, ws);
}
......
dragon/core/workspace.h
......@@ -47,29 +47,29 @@ class DRAGON_API Workspace {
}
/*! \brief Create the tensor */
Tensor* CreateTensor(const string&, FillerInfo* = nullptr);
Tensor* CreateTensor(const string& name, FillerInfo* filler = nullptr);
/*! \brief Try to return the tensor */
Tensor* TryGetTensor(const string&, bool = true) const;
Tensor* TryGetTensor(const string& name, bool external = true) const;
/*! \brief Return the tensor */
Tensor* GetTensor(const string&, bool = true) const;
Tensor* GetTensor(const string& name, bool external = true) const;
/*! \brief Reset the tensor */
void ResetTensor(const string&);
void ResetTensor(const string& name);
/*! \brief Return the filler info */
FillerInfo* GetFillerInfo(const string&);
FillerInfo* GetFillerInfo(const string& name);
/*! \brief Run the operator */
void RunOperator(const OperatorDef&);
void RunOperator(const OperatorDef& def);
/*! \brief Create the graph */
GraphBase* CreateGraph(const GraphDef&);
GraphBase* CreateGraph(const GraphDef& def);
/*! \brief Run the graph */
void RunGraph(
const string& graph_name,
const string& name,
const string& include = "",
const string& exclude = "",
const int stream = 0);
......@@ -88,28 +88,30 @@ class DRAGON_API Workspace {
/*! \brief Return a group of the shared raw data */
template <class Context>
vector<void*> data(const vector<size_t>& segments) {
int64_t nbytes = 0;
vector<void*> ret(segments.size());
for (auto& segment : segments)
nbytes += (int64_t)segment;
auto* T = CreateTensor("/share/data")->Reshape({nbytes});
ret[0] = T->template mutable_data<uint8_t, Context>();
for (int i = 1; i < segments.size(); i++)
ret[i] = (uint8_t*)ret[i - 1] + segments[i - 1];
return ret;
vector<void*> group(segments.size());
auto total_bytes = std::accumulate(segments.begin(), segments.end(), 0);
group[0] = CreateTensor("/share/data")
->Reshape({(int64_t)total_bytes})
->template mutable_data<uint8_t, Context>();
for (int i = 1; i < segments.size(); ++i) {
group[i] = (uint8_t*)group[i - 1] + segments[i - 1];
}
return group;
}
/*! \brief Return a group of shared typed data */
template <typename T, class Context>
vector<T*> data(const vector<int64_t>& segments) {
vector<size_t> segments_in_byte;
vector<T*> ret(segments.size());
for (const auto& e : segments)
segments_in_byte.emplace_back(e * sizeof(T));
auto ret_in_byte = data<Context>(segments_in_byte);
for (int i = 0; i < segments.size(); i++)
ret[i] = (T*)ret_in_byte[i];
return ret;
vector<T*> group(segments.size());
vector<size_t> segments_v2;
for (const auto size : segments) {
segments_v2.push_back(size * sizeof(T));
}
auto group_v2 = data<Context>(segments_v2);
for (int i = 0; i < segments.size(); ++i) {
group[i] = (T*)group_v2[i];
}
return group;
}
private:
......
dragon/onnx/onnx_backend.h
......@@ -71,7 +71,7 @@ class ConversionContext {
const int opset_version_;
};
typedef struct {
struct ONNXImporterReturns {
vector<OperatorDef> ops;
OperatorDef* AddOp() {
......@@ -83,7 +83,7 @@ typedef struct {
CHECK_LT(index, ops.size());
return &ops[index];
}
} ONNXImporterReturns;
};
class ONNXAttributes {
public:
......
dragon/python/core/autograph/tensor.py
......@@ -189,7 +189,7 @@ class Tensor(types.TensorMetaclass):
See Also
--------
`dragon.cast(...)`_ : Cast the data type of input.
`dragon.cast(...)`_
"""
......@@ -208,6 +208,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The self.
See Also
--------
`dragon.fill(...)`_
"""
return self._register_as('constant', value=value)
......@@ -221,7 +225,7 @@ class Tensor(types.TensorMetaclass):
See Also
--------
`dragon.copy(...)`_ : Copy the input.
`dragon.copy(...)`_
"""
......@@ -252,6 +256,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The self.
See Also
--------
`dragon.random.glorot_normal(...)`_
"""
return self._register_as('glorot_normal', mode=mode, scale=scale)
......@@ -273,6 +281,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The self.
See Also
--------
`dragon.random.glorot_uniform(...)`_
"""
return self._register_as('glorot_uniform', mode=mode, scale=scale)
......@@ -293,6 +305,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The self.
See Also
--------
`dragon.random.normal(...)`_
"""
return self._register_as('normal', mean=mean, std=std)
......@@ -311,7 +327,7 @@ class Tensor(types.TensorMetaclass):
See Also
--------
`dragon.reshape(...)`_ : Change the dimensions of input.
`dragon.reshape(...)`_
"""
......@@ -347,6 +363,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The self.
See Also
--------
`dragon.random.truncated_normal(...)`_
"""
return self._register_as('truncated_normal', mean=mean, std=std)
......@@ -367,6 +387,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The self.
See Also
--------
`dragon.random.uniform(...)`_
"""
return self._register_as('uniform', low=low, high=high)
......@@ -441,6 +465,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.add(...)`_
"""
def __float__(self):
......@@ -467,6 +495,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.greater_equal(...)`_
"""
def __getitem__(self, item):
......@@ -492,6 +524,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.greater(...)`_
"""
def __hash__(self):
......@@ -521,6 +557,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.less_equal(...)`_
"""
def __lt__(self, other):
......@@ -536,6 +576,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.less(...)`_
"""
def __mul__(self, other):
......@@ -551,6 +595,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.mul(...)`_
"""
def __neg__(self):
......@@ -561,6 +609,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.negative(...)`_
"""
def __radd__(self, other):
......@@ -576,6 +628,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.add(...)`_
"""
def __repr__(self):
......@@ -600,6 +656,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.div(...)`_
"""
def __rmul__(self, other):
......@@ -615,6 +675,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.mul(...)`_
"""
def __rsub__(self, other):
......@@ -630,6 +694,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.sub(...)`_
"""
def __setitem__(self, key, value):
......@@ -648,6 +716,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.sub(...)`_
"""
def __truediv__(self, other):
......@@ -663,6 +735,10 @@ class Tensor(types.TensorMetaclass):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.div(...)`_
"""
......
dragon/python/core/eager/tensor.py
......@@ -188,7 +188,7 @@ class EagerTensor(Tensor):
See Also
--------
`dragon.cast(...)`_ : Cast the data type of input.
`dragon.cast(...)`_
"""
......@@ -207,6 +207,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The self.
See Also
--------
`dragon.fill(...)`_
"""
def copy(self):
......@@ -219,7 +223,7 @@ class EagerTensor(Tensor):
See Also
--------
`dragon.copy(...)`_ : Copy the input.
`dragon.copy(...)`_
"""
......@@ -251,6 +255,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The self.
See Also
--------
`dragon.random.glorot_normal(...)`_
"""
def glorot_uniform(self, mode='fan_in', scale=3.0):
......@@ -271,6 +279,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The self.
See Also
--------
`dragon.random.glorot_uniform(...)`_
"""
def numpy(self, readonly=True):
......@@ -306,6 +318,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The self.
See Also
--------
`dragon.random.normal(...)`_
"""
def reshape(self, shape):
......@@ -323,7 +339,7 @@ class EagerTensor(Tensor):
See Also
--------
`dragon.reshape(...)`_ : Change the dimensions of input.
`dragon.reshape(...)`_
"""
......@@ -363,6 +379,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The self.
See Also
--------
`dragon.random.truncated_normal(...)`_
"""
def uniform(self, low=0, high=1):
......@@ -382,6 +402,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The self.
See Also
--------
`dragon.random.uniform(...)`_
"""
def _from_numpy(self, array, copy):
......@@ -413,6 +437,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.add(...)`_
"""
def __del__(self):
......@@ -445,6 +473,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.greater_equal(...)`_
"""
def __getitem__(self, item):
......@@ -470,6 +502,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.greater(...)`_
"""
def __hash__(self):
......@@ -488,6 +524,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The self.
See Also
--------
`dragon.math.add(...)`_
"""
def __imul__(self, other):
......@@ -503,6 +543,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The self.
See Also
--------
`dragon.math.mul(...)`_
"""
def __int__(self):
......@@ -529,6 +573,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The self.
See Also
--------
`dragon.math.sub(...)`_
"""
def __le__(self, other):
......@@ -544,6 +592,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.less_equal(...)`_
"""
def __lt__(self, other):
......@@ -559,6 +611,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.less(...)`_
"""
def __mul__(self, other):
......@@ -574,6 +630,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.mul(...)`_
"""
def __neg__(self):
......@@ -584,6 +644,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.negative(...)`_
"""
def __radd__(self, other):
......@@ -599,6 +663,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.add(...)`_
"""
def __repr__(self):
......@@ -629,6 +697,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.div(...)`_
"""
def __rmul__(self, other):
......@@ -644,6 +716,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.mul(...)`_
"""
def __rsub__(self, other):
......@@ -677,6 +753,10 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.sub(...)`_
"""
def __truediv__(self, other):
......@@ -692,4 +772,8 @@ class EagerTensor(Tensor):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.div(...)`_
"""
dragon/python/core/ops/array_ops.py
......@@ -1554,7 +1554,7 @@ def where(inputs, **kwargs):
See Also
--------
`dragon.nonzero(...)`_ : Return the index of non-zero elements.
dragon.nonzero()
"""
if types.is_tensor(inputs) or len(inputs) == 1:
......
dragon/python/core/ops/tensorbind_eager.py
......@@ -58,7 +58,7 @@ def astype(self, dtype, inplace=False):
See Also
--------
`dragon.cast(...)`_ : Cast the data type of input.
`dragon.cast(...)`_
"""
return array_ops_lib.Cast \
......@@ -80,6 +80,10 @@ def constant(self, value=0):
dragon.EagerTensor
The self.
See Also
--------
`dragon.fill(...)`_
"""
shape = self.shape
return init_ops_lib.Fill \
......@@ -100,7 +104,7 @@ def copy(self):
See Also
--------
`dragon.copy(...)`_ : Copy the value to ref.
`dragon.copy(...)`_
"""
return control_flow_ops_lib.Copy \
......@@ -120,6 +124,10 @@ def div(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.div(...)`_
"""
return _binary_op(self, other, 'Div')
......@@ -137,6 +145,10 @@ def ge(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.greater_equal(...)`_
"""
return _binary_op(self, other, 'GreaterEqual')
......@@ -179,6 +191,10 @@ def glorot_normal(self, mode='fan_in', scale=2.0):
dragon.EagerTensor
The self.
See Also
--------
`dragon.random.glorot_normal(...)`_
"""
shape = self.shape
return init_ops_lib.GlorotNormal \
......@@ -208,6 +224,10 @@ def glorot_uniform(self, mode='fan_in', scale=3.0):
dragon.EagerTensor
The self.
See Also
--------
`dragon.random.glorot_uniform(...)`_
"""
shape = self.shape
return init_ops_lib.GlorotUniform \
......@@ -232,6 +252,10 @@ def gt(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.greater(...)`_
"""
return _binary_op(self, other, 'Greater')
......@@ -249,6 +273,10 @@ def iadd(self, other):
dragon.EagerTensor
The self.
See Also
--------
`dragon.math.add(...)`_
"""
return _binary_op(self, other, 'Add', [self])
......@@ -266,6 +294,10 @@ def idiv(self, other):
dragon.EagerTensor
The self.
See Also
--------
`dragon.math.div(...)`_
"""
return _binary_op(self, other, 'Div', [self])
......@@ -283,6 +315,10 @@ def imul(self, other):
dragon.EagerTensor
The self.
See Also
--------
`dragon.math.mul(...)`_
"""
return _binary_op(self, other, 'Mul', [self])
......@@ -300,6 +336,10 @@ def isub(self, other):
dragon.EagerTensor
The self.
See Also
--------
`dragon.math.sub(...)`_
"""
return _binary_op(self, other, 'Sub', [self])
......@@ -317,6 +357,10 @@ def le(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.less_equal(...)`_
"""
return _binary_op(self, other, 'LessEqual')
......@@ -334,6 +378,10 @@ def lt(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.less(...)`_
"""
return _binary_op(self, other, 'Less')
......@@ -351,6 +399,10 @@ def mul(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.mul(...)`_
"""
return _binary_op(self, other, 'Mul')
......@@ -363,6 +415,10 @@ def neg(self):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.negative(...)`_
"""
return _unary_op(self, 'Neg')
......@@ -384,6 +440,10 @@ def normal(self, mean=0, std=1):
dragon.EagerTensor
The self.
See Also
--------
`dragon.random.normal(...)`_
"""
shape = self.shape
return init_ops_lib.RandomNormal \
......@@ -408,6 +468,10 @@ def radd(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.add(...)`_
"""
return _binary_op(other, self, 'Add')
......@@ -425,6 +489,10 @@ def rdiv(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.div(...)`_
"""
return _binary_op(other, self, 'Div')
......@@ -444,7 +512,7 @@ def reshape(self, shape):
See Also
--------
`dragon.reshape(...)`_ : Change the dimensions of input.
`dragon.reshape(...)`_
"""
with context.eager_mode():
......@@ -464,6 +532,10 @@ def rmul(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.mul(...)`_
"""
return _binary_op(other, self, 'Mul')
......@@ -481,6 +553,10 @@ def rsub(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.sub(...)`_
"""
return _binary_op(other, self, 'Sub')
......@@ -516,6 +592,10 @@ def sub(self, other):
dragon.EagerTensor
The output tensor.
See Also
--------
`dragon.math.sub(...)`_
"""
return _binary_op(self, other, 'Sub')
......@@ -537,6 +617,10 @@ def truncated_normal(self, mean=0, std=1):
dragon.EagerTensor
The self.
See Also
--------
`dragon.random.truncated_normal(...)`_
"""
shape = self.shape
return init_ops_lib.TruncatedNormal \
......@@ -565,6 +649,10 @@ def uniform(self, low=0, high=1):
dragon.EagerTensor
The self.
See Also
--------
`dragon.random.uniform(...)`_
"""
shape = self.shape
return init_ops_lib.RandomUniform \
......
dragon/python/core/ops/tensorbind_symbol.py
......@@ -35,6 +35,10 @@ def add(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.add(...)`_
"""
return _binary_op(self, other, 'Add')
......@@ -56,7 +60,7 @@ def astype(self, dtype, inplace=False):
See Also
--------
`dragon.cast(...)`_ : Cast the data type of input.
`dragon.cast(...)`_
"""
if self.dtype == dtype:
......@@ -75,7 +79,7 @@ def copy(self):
See Also
--------
`dragon.copy(...)`_ : Copy the value to ref.
`dragon.copy(...)`_
"""
outputs = [Tensor(shape=self.shape, dtype=self.dtype)]
......@@ -95,6 +99,10 @@ def div(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.div(...)`_
"""
return _binary_op(self, other, 'Div')
......@@ -112,6 +120,10 @@ def ge(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.greater_equal(...)`_
"""
return _binary_op(self, other, 'GreaterEqual')
......@@ -162,6 +174,10 @@ def gt(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.greater(...)`_
"""
return _binary_op(self, other, 'Greater')
......@@ -179,6 +195,10 @@ def le(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.less_equal(...)`_
"""
return _binary_op(self, other, 'LessEqual')
......@@ -196,6 +216,10 @@ def lt(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.less(...)`_
"""
return _binary_op(self, other, 'Less')
......@@ -213,6 +237,10 @@ def mul(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.mul(...)`_
"""
return _binary_op(self, other, 'Mul')
......@@ -225,6 +253,10 @@ def neg(self):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.negative(...)`_
"""
return _unary_op(self, 'Neg')
......@@ -242,6 +274,10 @@ def radd(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.add(...)`_
"""
return _binary_op(other, self, 'Add')
......@@ -259,6 +295,10 @@ def rdiv(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.div(...)`_
"""
return _binary_op(other, self, 'Div')
......@@ -278,7 +318,7 @@ def reshape(self, shape):
See Also
--------
`dragon.reshape(...)`_ : Change the dimensions of input.
`dragon.reshape(...)`_
"""
with context.graph_mode():
......@@ -298,6 +338,10 @@ def rmul(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.mul(...)`_
"""
return _binary_op(other, self, 'Mul')
......@@ -315,6 +359,10 @@ def rsub(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.sub(...)`_
"""
return _binary_op(other, self, 'Sub')
......@@ -368,6 +416,10 @@ def sub(self, other):
dragon.Tensor
The output tensor.
See Also
--------
`dragon.math.sub(...)`_
"""
return _binary_op(self, other, 'Sub')
......
torch/core/autograd/execute.py
......@@ -8,7 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""Execute tensor operations. """
"""Execute tensor operations."""
from __future__ import absolute_import
from __future__ import division
......
torch/core/nn/functional.py
......@@ -39,11 +39,12 @@ def affine(input, weight, bias=None):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.Affine(...)`_
"""
return _functions.Affine \
.instantiate(
input.device,
).apply(input, weight, bias)
return _functions.Affine.instantiate(input.device).apply(input, weight, bias)
def avg_pool2d(
......@@ -82,6 +83,10 @@ def avg_pool2d(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.AvgPool2d(...)`_
"""
return _pool(
_pool_mode='AVG',
......@@ -138,6 +143,10 @@ def batch_norm(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.BatchNorm2d(...)`_
"""
return _functions.BatchNorm \
.instantiate(
......@@ -181,6 +190,10 @@ def binary_cross_entropy_with_logits(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.BCEWithLogitsLoss(...)`_
"""
if size_average is not None or reduce is not None:
reduction = _reduction.legacy_get_string(size_average, reduce)
......@@ -236,6 +249,10 @@ def conv2d(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.Conv2d(...)`_
"""
return _conv(
_nd_util=utils._pair,
......@@ -290,6 +307,10 @@ def conv_transpose2d(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.ConvTranspose2d(...)`_
"""
return _conv_transpose(
_nd_util=utils._pair,
......@@ -335,6 +356,10 @@ def cross_entropy(
dragon.vm.torch.Tensor
The loss.
See Also
--------
`torch.nn.CrossEntropyLoss(...)`_
"""
if size_average is not None or reduce is not None:
reduction = _reduction.legacy_get_string(size_average, reduce)
......@@ -368,6 +393,10 @@ def ctc_loss(input, target, padding_mask=-1, reduction='mean'):
dragon.vm.torch.Tensor
The loss.
See Also
--------
`torch.nn.CTCLoss(...)`_
"""
prob = softmax(input, 2)
return _functions.CTCLoss \
......@@ -408,6 +437,10 @@ def depthwise_conv2d(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.DepthwiseConv2d(...)`_
"""
return _conv(
_nd_util=utils._pair,
......@@ -440,6 +473,10 @@ def dropout(input, p=0.5, training=True, inplace=False):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.Dropout(...)`_
"""
if not training:
return input
......@@ -494,6 +531,10 @@ def drop_block2d(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.DropBlock2d(...)`_
"""
if not training:
return input
......@@ -543,6 +584,10 @@ def drop_path(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.DropPath(...)`_
"""
if not training:
return input
......@@ -570,7 +615,7 @@ def elu(input, alpha=1., inplace=False):
See Also
--------
`torch.nn.ELU`_ - Apply the exponential linear unit.
`torch.nn.ELU(...)`_
Parameters
----------
......@@ -588,10 +633,8 @@ def elu(input, alpha=1., inplace=False):
"""
return _functions.Elu \
.instantiate(
input.device,
alpha=alpha,
).apply(input, inplace=inplace)
.instantiate(input.device, alpha=alpha) \
.apply(input, inplace=inplace)
def group_norm(input, weight, bias, groups=32, eps=1e-5):
......@@ -621,6 +664,10 @@ def group_norm(input, weight, bias, groups=32, eps=1e-5):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.GroupNorm(...)`_
"""
return _functions.GroupNorm \
.instantiate(
......@@ -679,6 +726,10 @@ def interpolate(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.Upsample(...)`_
"""
if size is not None:
size = nest.flatten(size)
......@@ -725,16 +776,18 @@ def l1_loss(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.L1Loss(...)`_
"""
if size_average is not None or reduce is not None:
reduction = _reduction.legacy_get_string(size_average, reduce)
else:
reduction = reduction
return _functions.L1Loss \
.instantiate(
input.device,
reduction=reduction,
).apply([input, target])
.instantiate(input.device, reduction=reduction) \
.apply([input, target])
def leaky_relu(input, negative_slope=0.01, inplace=False):
......@@ -749,10 +802,6 @@ def leaky_relu(input, negative_slope=0.01, inplace=False):
slope * x, & \text{ otherwise }
\end{cases}
See Also
--------
`torch.nn.LeakyReLU`_ - Apply the leaky rectified linear unit.
Parameters
----------
input : dragon.vm.torch.Tensor
......@@ -767,12 +816,14 @@ def leaky_relu(input, negative_slope=0.01, inplace=False):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.LeakyReLU(...)`_
"""
return _functions.Relu \
.instantiate(
input.device,
alpha=float(negative_slope),
).apply(input, inplace=inplace)
.instantiate(input.device, alpha=float(negative_slope)) \
.apply(input, inplace=inplace)
def linear(input, weight, bias=None):
......@@ -794,11 +845,12 @@ def linear(input, weight, bias=None):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.Linear(...)`_
"""
return _functions.Linear \
.instantiate(
input.device,
).apply(input, weight, bias)
return _functions.Linear.instantiate(input.device).apply(input, weight, bias)
def local_response_norm(input, size, alpha=1e-4, beta=0.75, k=1.):
......@@ -812,10 +864,6 @@ def local_response_norm(input, size, alpha=1e-4, beta=0.75, k=1.):
\sum_{j=\max(0, i-n/2)}^{\min(N-1,i+n/2)}x_{j}^2
\right)^{-\beta}
See Also
--------
`torch.nn.LocalResponseNorm`_ - Apply the local response normalization.
Parameters
----------
input : dragon.vm.torch.Tensor
......@@ -834,6 +882,10 @@ def local_response_norm(input, size, alpha=1e-4, beta=0.75, k=1.):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.LocalResponseNorm(...)`_
"""
return _functions.LocalResponseNorm \
.instantiate(
......@@ -864,6 +916,10 @@ def log_softmax(input, dim):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.LogSoftmax(...)`_
"""
return input - input.logsumexp(dim, keepdim=True)
......@@ -883,11 +939,12 @@ def lstm_cell(input, cx):
sequence of dragon.vm.torch.Tensor
The **h** and **c**.
See Also
--------
`torch.nn.LSTMCell(...)`_
"""
return _functions.LSTMCell \
.instantiate(
input.device,
).apply(input, cx)
return _functions.LSTMCell.instantiate(input.device).apply(input, cx)
def max_pool2d(
......@@ -926,6 +983,10 @@ def max_pool2d(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.MaxPool2d(...)`_
"""
return _pool(
_pool_mode='MAX',
......@@ -966,16 +1027,18 @@ def mse_loss(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.MSELoss(...)`_
"""
if size_average is not None or reduce is not None:
reduction = _reduction.legacy_get_string(size_average, reduce)
else:
reduction = reduction
return _functions.L2Loss \
.instantiate(
input.device,
reduction=reduction,
).apply([input, target])
.instantiate(input.device, reduction=reduction) \
.apply([input, target])
def nll_loss(
......@@ -1015,6 +1078,10 @@ def nll_loss(
dragon.vm.torch.Tensor
The loss.
See Also
--------
`torch.nn.NLLLoss(...)`_
"""
if size_average is not None or reduce is not None:
reduction = _reduction.legacy_get_string(size_average, reduce)
......@@ -1089,6 +1156,12 @@ def pad(input, pad, mode='constant', value=0):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.ConstantPad2d(...)`_,
`torch.nn.ReflectionPad2d(...)`_,
`torch.nn.ReplicationPad2d(...)`_
"""
ndim = input.ndimension()
pads_begin, pads_end = [0] * ndim, [0] * ndim
......@@ -1132,11 +1205,12 @@ def prelu(input, weight):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.PReLU(...)`_
"""
return _functions.PRelu \
.instantiate(
input.device,
).apply(input, weight)
return _functions.PRelu.instantiate(input.device).apply(input, weight)
def relu(input, inplace=False):
......@@ -1164,6 +1238,10 @@ def relu(input, inplace=False):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.ReLU(...)`_
"""
return leaky_relu(input, 0., inplace=inplace)
......@@ -1193,11 +1271,14 @@ def relu6(input, inplace=False):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.ReLU6(...)`_
"""
return _functions.Relu6 \
.instantiate(
input.device,
).apply(input, inplace=inplace)
.instantiate(input.device) \
.apply(input, inplace=inplace)
def selu(input, inplace=False):
......@@ -1225,6 +1306,10 @@ def selu(input, inplace=False):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.SELU(...)`_
"""
return _activation(input, inplace, 'Selu')
......@@ -1248,6 +1333,10 @@ def sigmoid(input, inplace=False):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.Sigmoid(...)`_
"""
return _activation(input, inplace, 'Sigmoid')
......@@ -1296,6 +1385,10 @@ def sigmoid_focal_loss(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.SigmoidFocalLoss(...)`_
"""
if size_average is not None or reduce is not None:
reduction = _reduction.legacy_get_string(size_average, reduce)
......@@ -1351,6 +1444,10 @@ def smooth_l1_loss(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.SmoothL1Loss(...)`_
"""
if size_average is not None or reduce is not None:
reduction = _reduction.legacy_get_string(size_average, reduce)
......@@ -1385,12 +1482,14 @@ def softmax(input, dim, inplace=False):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.Softmax(...)`_
"""
return _functions.Softmax \
.instantiate(
input.device,
axis=dim,
).apply(input, inplace=inplace)
.instantiate(input.device, axis=dim) \
.apply(input, inplace=inplace)
def sync_batch_norm(
......@@ -1447,6 +1546,10 @@ def sync_batch_norm(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.SyncBatchNorm(...)`_
"""
if process_group is None:
raise ValueError('<process_group> is required.')
......@@ -1479,6 +1582,10 @@ def tanh(input, inplace=False):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.Tanh(...)`_
"""
return _activation(input, inplace, 'Tanh')
......@@ -1496,8 +1603,8 @@ def upsample(
```python
x = torch.ones((1, 2, 3, 4))
y = F.interpolate(x, size=6) # Shape: (1, 2, 6, 6)
z = F.interpolate(x, scale_factor=2) # Shape: (1, 2, 6, 8)
y = F.upsample(x, size=6) # Shape: (1, 2, 6, 6)
z = F.upsample(x, scale_factor=2) # Shape: (1, 2, 6, 8)
```
Set ``align_corners`` to determine the input coordinates in linear ``mode``:
......@@ -1532,6 +1639,10 @@ def upsample(
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.Upsample(...)`_
"""
return interpolate(input, size, scale_factor, mode, align_corners)
......@@ -1561,6 +1672,10 @@ def upsample_bilinear(input, size=None, scale_factor=None):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.UpsamplingBilinear2d(...)`_
"""
return interpolate(input, size, scale_factor, 'linear', align_corners=True)
......@@ -1590,16 +1705,18 @@ def upsample_nearest(input, size=None, scale_factor=None):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nn.UpsamplingNearest2d(...)`_
"""
return interpolate(input, size, scale_factor, 'nearest')
def _activation(input, inplace=False, _op_type=''):
return _functions._Activation \
.instantiate(
input.device,
op_type=_op_type,
).apply(input, inplace=inplace)
.instantiate(input.device, op_type=_op_type) \
.apply(input, inplace=inplace)
def _conv(
......
torch/core/nn/modules/activation.py
......@@ -8,6 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""Activation modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -43,7 +44,7 @@ class ELU(Module):
See Also
--------
`torch.nn.functional.elu(...)`_ - Apply the exponential linear unit to input.
`torch.nn.functional.elu(...)`_
"""
......@@ -145,7 +146,7 @@ class LeakyReLU(Module):
See Also
--------
`torch.nn.functional.leaky_relu(...)`_ - Apply the leaky rectified linear unit.
`torch.nn.functional.leaky_relu(...)`_
"""
......@@ -187,6 +188,10 @@ class LogSoftmax(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.log_softmax(...)`_
"""
def __init__(self, dim):
......@@ -234,6 +239,11 @@ class PReLU(Module):
mm = torch.nn.PReLU(num_parameters=3)
z = mm(x)
```
See Also
--------
`torch.nn.functional.prelu(...)`_
"""
def __init__(self, num_parameters=1, init=0.25):
......@@ -279,6 +289,10 @@ class ReLU(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.relu(...)`_
"""
def __init__(self, inplace=False):
......@@ -322,6 +336,10 @@ class ReLU6(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.relu6(...)`_
"""
def __init__(self, inplace=False):
......@@ -365,6 +383,10 @@ class SELU(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.selu(...)`_
"""
def __init__(self, inplace=False):
......@@ -402,6 +424,10 @@ class Sigmoid(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.sigmoid(...)`_
"""
def __init__(self, inplace=False):
......@@ -439,6 +465,10 @@ class Softmax(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.softmax(...)`_
"""
def __init__(self, dim, inplace=False):
......@@ -479,6 +509,10 @@ class Tanh(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.tanh(...)`_
"""
def __init__(self, inplace=False):
......
torch/core/nn/modules/affine.py
......@@ -8,6 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""Affine modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -25,8 +26,6 @@ class Affine(Module):
.. math:: y = Ax + b
This transform is often taken as a post-processing of normalization.
Specially, a trained ``BatchNorm`` can be fused to this under some
fine-tune settings, such as detection and segmentation.
Examples:
......@@ -46,6 +45,10 @@ class Affine(Module):
y4d = m(x4d)
```
See Also
--------
`torch.nn.functional.affine(...)`_
"""
def __init__(
......
torch/core/nn/modules/batchnorm.py
......@@ -8,6 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""BatchNorm modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -89,10 +90,14 @@ class BatchNorm1d(_BatchNorm):
.. math::
y = \frac{x - \mathrm{E}[x]}{\sqrt{\mathrm{Var}[x] + \epsilon}} * \gamma + \beta
The moving average of stats are calculated as:
The running average of statistics are calculated as:
.. math::
x_{moving} \leftarrow (1 - momentum) * x_{moving} + momentum * x_{stat}
x_{\text{running}} = (1 - \text{momentum}) * x_{\text{running}} + \text{momentum} * x_{\text{stat}}
See Also
--------
`torch.nn.functional.batch_norm(...)`_
"""
......@@ -136,10 +141,14 @@ class BatchNorm2d(_BatchNorm):
.. math::
y = \frac{x - \mathrm{E}[x]}{\sqrt{\mathrm{Var}[x] + \epsilon}} * \gamma + \beta
The moving average of stats are calculated as:
The running average of statistics are calculated as:
.. math::
x_{moving} \leftarrow (1 - momentum) * x_{moving} + momentum * x_{stat}
x_{\text{running}} = (1 - \text{momentum}) * x_{\text{running}} + \text{momentum} * x_{\text{stat}}
See Also
--------
`torch.nn.functional.batch_norm(...)`_
"""
......@@ -183,10 +192,14 @@ class BatchNorm3d(_BatchNorm):
.. math::
y = \frac{x - \mathrm{E}[x]}{\sqrt{\mathrm{Var}[x] + \epsilon}} * \gamma + \beta
The moving average of stats are calculated as:
The running average of statistics are calculated as:
.. math::
x_{moving} \leftarrow (1 - momentum) * x_{moving} + momentum * x_{stat}
x_{\text{running}} = (1 - \text{momentum}) * x_{\text{running}} + \text{momentum} * x_{\text{stat}}
See Also
--------
`torch.nn.functional.batch_norm(...)`_
"""
......@@ -230,15 +243,19 @@ class SyncBatchNorm(_BatchNorm):
.. math::
y = \frac{x - \mathrm{E}[x]}{\sqrt{\mathrm{Var}[x] + \epsilon}} * \gamma + \beta
The moving average of stats are calculated as:
The running average of statistics are calculated as:
.. math::
x_{moving} \leftarrow (1 - momentum) * x_{moving} + momentum * x_{stat}
x_{\text{running}} = (1 - \text{momentum}) * x_{\text{running}} + \text{momentum} * x_{\text{stat}}
Additionally, you can specify ``process_group`` to perform synchronization.
Additionally, specify ``process_group`` to perform synchronization.
If not, value returning from ``dragon.distributed.get_group(...)`` will be used.
See Also
--------
`torch.nn.functional.batch_norm(...)`_
"""
def __init__(
......
torch/core/nn/modules/container.py
......@@ -12,6 +12,7 @@
# <https://github.com/pytorch/pytorch/blob/master/torch/nn/modules/container.py>
#
# ------------------------------------------------------------
"""Container modules."""
from __future__ import absolute_import
from __future__ import division
......
torch/core/nn/modules/conv.py
......@@ -8,6 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""Convolution modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -90,17 +91,7 @@ class _ConvNd(Module):
class Conv2d(_ConvNd):
r"""Apply the 2d convolution.
The spatial output dimension is computed as:
.. math::
\begin{cases}
\text{DK}_{size} = dilation *
(\text{K}_{size} - 1) + 1 \\
\text{Dim}_{out} = (\text{Dim}_{in} +
2 * pad - \text{DK}_{size}) / stride + 1
\end{cases}
"""Apply the 2d convolution.
Examples:
......@@ -110,6 +101,10 @@ class Conv2d(_ConvNd):
y = m(x)
```
See Also
--------
`torch.nn.functional.conv2d(...)`_
"""
def __init__(
......@@ -171,17 +166,7 @@ class Conv2d(_ConvNd):
class ConvTranspose2d(_ConvNd):
r"""Apply the 2d deconvolution.
The spatial output dimension is computed as:
.. math::
\begin{cases}
\text{DK}_{size} = dilation *
(\text{K}_{size} - 1) + 1 \\
\text{Dim}_{out} = (\text{Dim}_{in} - 1) *
stride + \text{DK}_{size} - 2 * pad
\end{cases}
"""Apply the 2d deconvolution.
Examples:
......@@ -191,6 +176,10 @@ class ConvTranspose2d(_ConvNd):
y = m(x)
```
See Also
--------
`torch.nn.functional.conv_transpose2d(...)`_
"""
def __init__(
......@@ -256,17 +245,7 @@ class ConvTranspose2d(_ConvNd):
class DepthwiseConv2d(Conv2d):
r"""Apply the 2d depthwise convolution.
The spatial output dimension is computed as:
.. math::
\begin{cases}
\text{DK}_{size} = dilation *
(\text{K}_{size} - 1) + 1 \\
\text{Dim}_{out} = (\text{Dim}_{in} +
2 * pad - \text{DK}_{size}) / stride + 1
\end{cases}
"""Apply the 2d depthwise convolution.
Examples:
......@@ -276,6 +255,10 @@ class DepthwiseConv2d(Conv2d):
y = m(x)
```
See Also
--------
`torch.nn.functional.depthwise_conv2d(...)`_
"""
def __init__(
......
torch/core/nn/modules/dropout.py
......@@ -8,6 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""Dropout modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -38,6 +39,10 @@ class DropBlock2d(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.drop_block2d(...)`_
"""
# Store the global unique slot index
......@@ -114,6 +119,10 @@ class Dropout(Module):
z = m(x)
```
See Also
--------
`torch.nn.functional.dropout(...)`_
"""
def __init__(self, p=0.5, inplace=False):
......@@ -155,6 +164,10 @@ class DropPath(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.drop_path(...)`_
"""
# Store the global unique slot index
......
torch/core/nn/modules/linear.py
......@@ -8,6 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""Linear modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -34,6 +35,10 @@ class Linear(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.linear(...)`_
"""
def __init__(self, in_features, out_features, bias=True):
......
torch/core/nn/modules/loss.py
......@@ -7,11 +7,8 @@
#
# <https://opensource.org/licenses/BSD-2-Clause>
#
# Codes are based on:
#
# <https://github.com/pytorch/pytorch/blob/master/torch/nn/modules/loss.py>
#
# ------------------------------------------------------------
"""Loss modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -68,6 +65,10 @@ class CTCLoss(_Loss):
loss = m(logits, labels)
```
See Also
--------
`torch.nn.functional.ctc_loss(...)`_
"""
def __init__(self, padding_mask=-1, reduction='mean'):
......@@ -107,6 +108,10 @@ class NLLLoss(_WeightedLoss):
loss = m2(m1(torch.randn(2, 2)), torch.tensor([0, 1]))
```
See Also
--------
`torch.nn.functional.nll_loss(...)`_
"""
def __init__(
......@@ -155,6 +160,10 @@ class BCEWithLogitsLoss(_WeightedLoss):
loss = m(torch.randn(2, 1), torch.tensor([0., 1.], 'float32'))
```
See Also
--------
`torch.nn.functional.binary_cross_entropy_with_logits(...)`_
"""
def __init__(
......@@ -205,6 +214,10 @@ class CrossEntropyLoss(_WeightedLoss):
loss = m(logits, targets)
```
See Also
--------
`torch.nn.functional.cross_entropy(...)`_
"""
def __init__(
......@@ -257,6 +270,10 @@ class L1Loss(_Loss):
loss = m(torch.ones(2, 3), torch.zeros(2, 3))
```
See Also
--------
`torch.nn.functional.l1_loss(...)`_
"""
def __init__(self, size_average=None, reduce=None, reduction='mean'):
......@@ -292,6 +309,10 @@ class MSELoss(_Loss):
loss = m(torch.ones(2, 3) * 2, torch.zeros(2, 3))
```
See Also
--------
`torch.nn.functional.mse_loss(...)`_
"""
def __init__(self, size_average=None, reduce=None, reduction='mean'):
......@@ -333,6 +354,10 @@ class SmoothL1Loss(_Loss):
loss = m(torch.ones(2, 3), torch.zeros(2, 3))
```
See Also
--------
`torch.nn.functional.smooth_l1_loss(...)`_
"""
def __init__(
......@@ -384,6 +409,10 @@ class SigmoidFocalLoss(_WeightedLoss):
loss = m(logits, targets)
```
See Also
--------
`torch.nn.functional.sigmoid_focal_loss(...)`_
"""
def __init__(
......
torch/core/nn/modules/module.py
......@@ -7,11 +7,8 @@
#
# <https://opensource.org/licenses/BSD-2-Clause>
#
# Codes are based on:
#
# <https://github.com/pytorch/pytorch/blob/master/torch/nn/modules/module.py>
#
# ------------------------------------------------------------
"""Base module class."""
from __future__ import absolute_import
from __future__ import division
......
torch/core/nn/modules/normalization.py
......@@ -8,6 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""Normalization modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -39,6 +40,10 @@ class GroupNorm(Module):
y = m(x)
```
See Also
--------
`torch.nn.functional.group_norm(...)`_
"""
def __init__(
......@@ -123,7 +128,7 @@ class LocalResponseNorm(Module):
See Also
--------
`torch.nn.functional.local_response_norm(...)`_ - Apply the local response normalization to input.
`torch.nn.functional.local_response_norm(...)`_
"""
......
torch/core/nn/modules/padding.py
......@@ -8,6 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""Padding modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -69,7 +70,12 @@ class ConstantPad1d(_ConstantPadNd):
y = m(x) # (1, 2) -> (1, 4)
```
See Also
--------
`torch.nn.functional.pad(...)`_
"""
def __init__(self, padding, value):
"""Create a ``ConstantPad1d`` module.
......@@ -100,6 +106,10 @@ class ConstantPad2d(_ConstantPadNd):
y = m(x) # (1, 2, 2) -> (1, 4, 4)
```
See Also
--------
`torch.nn.functional.pad(...)`_
"""
def __init__(self, padding, value):
......@@ -132,6 +142,10 @@ class ConstantPad3d(_ConstantPadNd):
y = m(x) # (1, 2, 2, 2) -> (1, 4, 4, 4)
```
See Also
--------
`torch.nn.functional.pad(...)`_
"""
def __init__(self, padding, value):
......@@ -164,6 +178,10 @@ class ReflectionPad1d(_ReflectionPadNd):
y = m(x) # (1, 4) -> (1, 6)
```
See Also
--------
`torch.nn.functional.pad(...)`_
"""
def __init__(self, padding):
......@@ -194,6 +212,10 @@ class ReflectionPad2d(_ReflectionPadNd):
y = m(x) # (1, 4, 4) -> (1, 6, 6)
```
See Also
--------
`torch.nn.functional.pad(...)`_
"""
def __init__(self, padding):
......@@ -224,6 +246,10 @@ class ReflectionPad3d(_ReflectionPadNd):
y = m(x) # (1, 4, 4, 4) -> (1, 6, 6, 6)
```
See Also
--------
`torch.nn.functional.pad(...)`_
"""
def __init__(self, padding):
......@@ -254,6 +280,10 @@ class ReplicationPad1d(_ReplicationPadNd):
y = m(x) # (1, 4) -> (1, 6)
```
See Also
--------
`torch.nn.functional.pad(...)`_
"""
def __init__(self, padding):
......@@ -284,6 +314,10 @@ class ReplicationPad2d(_ReplicationPadNd):
y = m(x) # (1, 4, 4) -> (1, 6, 6)
```
See Also
--------
`torch.nn.functional.pad(...)`_
"""
def __init__(self, padding):
......@@ -314,6 +348,10 @@ class ReplicationPad3d(_ReplicationPadNd):
y = m(x) # (1, 4, 4, 4) -> (1, 6, 6, 6)
```
See Also
--------
`torch.nn.functional.pad(...)`_
"""
def __init__(self, padding):
......@@ -344,6 +382,10 @@ class ZeroPad2d(ConstantPad2d):
y = m(x) # (1, 2, 2) -> (1, 4, 4)
```
See Also
--------
`torch.nn.functional.pad(...)`_
"""
def __init__(self, padding):
......
torch/core/nn/modules/pooling.py
......@@ -8,6 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""Pooling modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -59,6 +60,10 @@ class AvgPool2d(_PoolNd):
y = m(x)
```
See Also
--------
`torch.nn.functional.avg_pool2d(...)`_
"""
def __init__(
......@@ -121,6 +126,10 @@ class MaxPool2d(_PoolNd):
y = m(x)
```
See Also
--------
`torch.nn.functional.max_pool2d(...)`_
"""
def __init__(
......
torch/core/nn/modules/rnn.py
......@@ -7,11 +7,8 @@
#
# <https://opensource.org/licenses/BSD-2-Clause>
#
# Codes are based on:
#
# <https://github.com/pytorch/pytorch/blob/master/torch/nn/modules/rnn.py>
#
# ------------------------------------------------------------
"""RNN modules."""
from __future__ import absolute_import
from __future__ import division
......
torch/core/nn/modules/upsampling.py
......@@ -8,6 +8,7 @@
# <https://opensource.org/licenses/BSD-2-Clause>
#
# ------------------------------------------------------------
"""Upsampling modules."""
from __future__ import absolute_import
from __future__ import division
......@@ -28,6 +29,10 @@ class Upsample(Module):
z = torch.nn.UpSample(scale_factor=2)(x) # Shape: (1, 2, 6, 8)
```
See Also
--------
`torch.nn.functional.interpolate(...)`_
"""
def __init__(
......@@ -89,6 +94,10 @@ class UpsamplingBilinear2d(Upsample):
z = torch.nn.UpsamplingBilinear2d(scale_factor=2)(x) # Shape: (1, 2, 6, 8)
```
See Also
--------
`torch.nn.functional.interpolate(...)`_
"""
def __init__(self, size=None, scale_factor=None):
......@@ -117,6 +126,10 @@ class UpsamplingNearest2d(Upsample):
z = torch.nn.UpsamplingNearest2d(scale_factor=2)(x) # Shape: (1, 2, 6, 8)
```
See Also
--------
`torch.nn.functional.interpolate(...)`_
"""
def __init__(self, size=None, scale_factor=None):
......
torch/core/nn/modules/utils.py
......@@ -12,6 +12,7 @@
# <https://github.com/pytorch/pytorch/blob/master/torch/nn/modules/utils.py>
#
# ------------------------------------------------------------
"""Module utilities."""
from __future__ import absolute_import
from __future__ import division
......
torch/core/ops/tensorbind.py
......@@ -33,7 +33,7 @@ def abs(self):
See Also
--------
`torch.abs(...)`_ : Compute the absolute value of input.
`torch.abs(...)`_
"""
return math_funcs.abs(self)
......@@ -56,7 +56,7 @@ def add(self, other):
See Also
--------
`torch.add(...)`_ : Compute the element-wise addition.
`torch.add(...)`_
"""
return math_funcs.add(self, other)
......@@ -79,7 +79,7 @@ def add_(self, other):
See Also
--------
`torch.add(...)`_ : Compute the element-wise addition.
`torch.add(...)`_
"""
return math_funcs.add(self, other, self)
......@@ -102,7 +102,7 @@ def argmax(self, dim=None, keepdim=False):
See Also
--------
`torch.argmax(...)`_ : Return the index of maximum elements along the given dimension.
`torch.argmax(...)`_
"""
return array_funcs.argmax(self, dim, keepdim)
......@@ -125,7 +125,7 @@ def argmin(self, dim=None, keepdim=False):
See Also
--------
`torch.argmin(...)`_ : Return the index of minimum elements along the given dimension.
`torch.argmin(...)`_
"""
return array_funcs.argmin(self, dim, keepdim)
......@@ -161,7 +161,7 @@ def bitwise_not(self):
See Also
--------
`torch.bitwise_not(...)`_ : Compute the element-wise NOT bitwise operation.
`torch.bitwise_not(...)`_
"""
return math_funcs.bitwise_not(self)
......@@ -179,7 +179,7 @@ def bitwise_not_(self):
See Also
--------
`torch.bitwise_not(...)`_ : Compute the element-wise NOT bitwise operation.
`torch.bitwise_not(...)`_
"""
return math_funcs.bitwise_not(self, self)
......@@ -202,7 +202,7 @@ def bitwise_xor(self, other):
See Also
--------
`torch.bitwise_xor(...)`_ : Compute the element-wise XOR bitwise operation.
`torch.bitwise_xor(...)`_
"""
return math_funcs.bitwise_xor(self, other)
......@@ -225,7 +225,7 @@ def bitwise_xor_(self, other):
See Also
--------
`torch.bitwise_xor(...)`_ : Compute the element-wise XOR bitwise operation.
`torch.bitwise_xor(...)`_
"""
return math_funcs.bitwise_xor(self, other, self)
......@@ -291,7 +291,7 @@ def ceil(self):
See Also
--------
`torch.ceil(...)`_ : Compute the smallest integer not less than input.
`torch.ceil(...)`_
"""
return math_funcs.ceil(self)
......@@ -309,7 +309,7 @@ def ceil_(self):
See Also
--------
`torch.ceil(...)`_ : Compute the smallest integer not less than input.
`torch.ceil(...)`_
"""
return math_funcs.ceil(self, self)
......@@ -375,7 +375,7 @@ def clamp(self, min=None, max=None):
See Also
--------
`torch.clamp(...)`_ : Compute the clipped input according to the given bounds.
`torch.clamp(...)`_
"""
return math_funcs.clamp(self, min, max)
......@@ -398,7 +398,7 @@ def clamp_(self, min=None, max=None):
See Also
--------
`torch.clamp(...)`_ : Compute the clipped input according to the given bounds.
`torch.clamp(...)`_
"""
return math_funcs.clamp(self, min, max, self)
......@@ -416,7 +416,7 @@ def cos(self):
See Also
--------
`torch.cos(...)`_ : Compute the cos of input.
`torch.cos(...)`_
"""
return math_funcs.cos(self)
......@@ -437,7 +437,7 @@ def cumsum(self, dim):
See Also
--------
`torch.cumsum(...)`_ : Compute the cumulative sum of elements along the given dimension.
`torch.cumsum(...)`_
"""
return array_funcs.cumsum(self, dim)
......@@ -460,7 +460,7 @@ def div(self, other):
See Also
--------
`torch.div(...)`_ : Compute the element-wise division.
`torch.div(...)`_
"""
return math_funcs.div(self, other)
......@@ -483,7 +483,7 @@ def div_(self, other):
See Also
--------
`torch.div(...)`_ : Compute the element-wise division.
`torch.div(...)`_
"""
return math_funcs.div(self, other, self)
......@@ -530,7 +530,7 @@ def eq(self, other):
See Also
--------
`torch.eq(...)`_ : Compute the element-wise equal comparison.
`torch.eq(...)`_
"""
return math_funcs.eq(self, other)
......@@ -548,7 +548,7 @@ def exp(self):
See Also
--------
`torch.exp(...)`_ : Compute the exponential of input.
`torch.exp(...)`_
"""
return math_funcs.exp(self)
......@@ -569,7 +569,7 @@ def expand(self, *sizes):
See Also
--------
`torch.expand(...)`_ : Broadcast input according to given sizes.
`torch.expand(...)`_
"""
return array_funcs.expand(self, sizes)
......@@ -630,7 +630,7 @@ def floor(self):
See Also
--------
`torch.floor(...)`_ : Compute the largest integer not greater than input.
`torch.floor(...)`_
"""
return math_funcs.floor(self)
......@@ -648,7 +648,7 @@ def floor_(self):
See Also
--------
`torch.floor(...)`_ : Compute the largest integer not greater than input.
`torch.floor(...)`_
"""
return math_funcs.floor(self, self)
......@@ -671,7 +671,7 @@ def ge(self, other):
See Also
--------
`torch.ge(...)`_ : Compute the element-wise greater-equal comparison.
`torch.ge(...)`_
"""
return math_funcs.ge(self, other)
......@@ -715,7 +715,7 @@ def gt(self, other):
See Also
--------
`torch.gt(...)`_ : Compute the element-wise greater comparison.
`torch.gt(...)`_
"""
return math_funcs.gt(self, other)
......@@ -805,7 +805,7 @@ def le(self, other):
See Also
--------
`torch.le(...)`_ : Compute the element-wise less-equal comparison.
`torch.le(...)`_
"""
return math_funcs.le(self, other)
......@@ -887,7 +887,7 @@ def lt(self, other):
See Also
--------
`torch.lt(...)`_ : Compute the element-wise less comparison.
`torch.lt(...)`_
"""
return math_funcs.lt(self, other)
......@@ -1010,7 +1010,7 @@ def mul(self, other):
See Also
--------
`torch.mul(...)`_ : Compute the element-wise multiplication.
`torch.mul(...)`_
"""
return math_funcs.mul(self, other)
......@@ -1033,7 +1033,7 @@ def mul_(self, other):
See Also
--------
`torch.mul(...)`_ : Compute the element-wise multiplication.
`torch.mul(...)`_
"""
return math_funcs.mul(self, other, self)
......@@ -1097,7 +1097,7 @@ def ne(self, other):
See Also
--------
`torch.ne(...)`_ : Compute the element-wise not-equal comparison.
`torch.ne(...)`_
"""
return math_funcs.ne(self, other)
......@@ -1115,20 +1115,26 @@ def neg(self):
See Also
--------
`torch.neg(...)`_ : Compute the element-wise negative.
`torch.neg(...)`_
"""
return math_funcs.neg(self)
def nonzero(self):
"""Return the index of non-zero elements.
r"""Return the index of non-zero elements.
.. math:: \text{out} = \{i\}, \text{ if } \text{self}_{i} \neq 0
Returns
-------
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nonzero(...)`_
"""
return array_funcs.nonzero(self)
......@@ -1186,7 +1192,7 @@ def pow(self, exponent):
See Also
--------
`torch.pow(...)`_ : Compute the power of input.
`torch.pow(...)`_
"""
return math_funcs.pow(self, exponent)
......@@ -1204,7 +1210,7 @@ def reciprocal(self):
See Also
--------
`torch.reciprocal(...)`_ : Compute the reciprocal of input.
`torch.reciprocal(...)`_
"""
return math_funcs.reciprocal(self)
......@@ -1222,7 +1228,7 @@ def reciprocal_(self):
See Also
--------
`torch.reciprocal(...)`_ : Compute the reciprocal of input.
`torch.reciprocal(...)`_
"""
return math_funcs.reciprocal(self, self)
......@@ -1260,7 +1266,7 @@ def reshape(self, shape):
See Also
--------
`torch.reshape(...)`_ : Change the shape of input.
`torch.reshape(...)`_
"""
return array_funcs.reshape(self, shape)
......@@ -1281,7 +1287,7 @@ def reshape_(self, shape):
See Also
--------
`torch.reshape(...)`_ : Change the shape of input.
`torch.reshape(...)`_
"""
return array_funcs.reshape(self, shape, self)
......@@ -1299,7 +1305,7 @@ def round(self):
See Also
--------
`torch.round(...)`_ : Compute the nearest integer of input.
`torch.round(...)`_
"""
return math_funcs.round(self)
......@@ -1317,7 +1323,7 @@ def round_(self):
See Also
--------
`torch.round(...)`_ : Compute the nearest integer of input.
`torch.round(...)`_
"""
return math_funcs.round(self, self)
......@@ -1335,7 +1341,7 @@ def rsqrt(self):
See Also
--------
`torch.rsqrt(...)`_ : Compute the square root of input.
`torch.rsqrt(...)`_
"""
return math_funcs.rsqrt(self)
......@@ -1353,7 +1359,7 @@ def rsqrt_(self):
See Also
--------
`torch.rsqrt(...)`_ : Compute the square root of input.
`torch.rsqrt(...)`_
"""
return math_funcs.rsqrt(self, self)
......@@ -1395,7 +1401,7 @@ def sign(self):
See Also
--------
`torch.sign(...)`_ : Compute the sign indication of input.
`torch.sign(...)`_
"""
return math_funcs.sign(self)
......@@ -1419,7 +1425,7 @@ def sign_(self):
See Also
--------
`torch.sign(...)`_ : Compute the sign indication of input.
`torch.sign(...)`_
"""
return math_funcs.sign(self, self)
......@@ -1437,7 +1443,7 @@ def sin(self):
See Also
--------
`torch.sin(...)`_ : Compute the sin of input.
`torch.sin(...)`_
"""
return math_funcs.sin(self)
......@@ -1455,7 +1461,7 @@ def sqrt(self):
See Also
--------
`torch.sqrt(...)`_ : Compute the square root of input.
`torch.sqrt(...)`_
"""
return math_funcs.sqrt(self)
......@@ -1473,7 +1479,7 @@ def sqrt_(self):
See Also
--------
`torch.sqrt(...)`_ : Compute the square root of input.
`torch.sqrt(...)`_
"""
return math_funcs.sqrt(self, self)
......@@ -1492,12 +1498,16 @@ def squeeze(self, dim=None):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.squeeze(...)`_
"""
return array_funcs.squeeze(self, dim)
def squeeze_(self, dim=None):
"""Inplace version of ``Tensor.squeeze()``.
"""Remove the dimensions with size 1.
Parameters
----------
......@@ -1509,6 +1519,10 @@ def squeeze_(self, dim=None):
dragon.vm.torch.Tensor
The self.
See Also
--------
`torch.squeeze(...)`_
"""
return array_funcs.squeeze(self, dim, self)
......@@ -1528,6 +1542,10 @@ def sum(self, dim=None, keepdim=False):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.sum(...)`_
"""
return array_funcs.sum(self, dim, keepdim)
......@@ -1549,7 +1567,7 @@ def sub(self, other):
See Also
--------
`torch.sub(...)`_ : Compute the element-wise subtraction.
`torch.sub(...)`_
"""
return math_funcs.sub(self, other)
......@@ -1572,7 +1590,7 @@ def sub_(self, other):
See Also
--------
`torch.sub(...)`_ : Compute the element-wise subtraction.
`torch.sub(...)`_
"""
return math_funcs.sub(self, other, self)
......@@ -1599,7 +1617,7 @@ def topk(self, k, dim=None, largest=True, sorted=True):
See Also
--------
`torch.topk(...)`_ : Return the top-K largest or smallest elements along the given dimension.
`torch.topk(...)`_
"""
return array_funcs.topk(self, k, dim, largest, sorted)
......@@ -1660,12 +1678,16 @@ def unsqueeze(self, dim):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.unsqueeze(...)`_
"""
return array_funcs.unsqueeze(self, dim)
def unsqueeze_(self, dim):
"""In-place version of ``Tensor.unsqueeze()``.
"""Insert the dimensions of size 1.
Parameters
----------
......@@ -1677,6 +1699,10 @@ def unsqueeze_(self, dim):
dragon.vm.torch.Tensor
The self.
See Also
--------
`torch.unsqueeze(...)`_
"""
return array_funcs.unsqueeze(self, dim, self)
......@@ -1703,6 +1729,10 @@ def where(self, condition, y):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.where(...)`_
"""
return array_funcs.where(condition, self, y)
......
torch/core/tensor.py
......@@ -221,7 +221,7 @@ class Tensor(object):
See Also
--------
`torch.abs(...)`_ : Compute the absolute value of input.
`torch.abs(...)`_
"""
......@@ -242,7 +242,7 @@ class Tensor(object):
See Also
--------
`torch.add(...)`_ : Compute the element-wise addition.
`torch.add(...)`_
"""
......@@ -263,7 +263,7 @@ class Tensor(object):
See Also
--------
`torch.add(...)`_ : Compute the element-wise addition.
`torch.add(...)`_
"""
......@@ -284,7 +284,7 @@ class Tensor(object):
See Also
--------
`torch.argmax(...)`_ : Return the index of maximum elements along the given dimension.
`torch.argmax(...)`_
"""
......@@ -305,7 +305,7 @@ class Tensor(object):
See Also
--------
`torch.argmin(...)`_ : Return the index of minimum elements along the given dimension.
`torch.argmin(...)`_
"""
......@@ -333,7 +333,7 @@ class Tensor(object):
See Also
--------
`torch.bitwise_not(...)`_ : Compute the element-wise NOT bitwise operation.
`torch.bitwise_not(...)`_
"""
......@@ -349,7 +349,7 @@ class Tensor(object):
See Also
--------
`torch.bitwise_not(...)`_ : Compute the element-wise NOT bitwise operation.
`torch.bitwise_not(...)`_
"""
......@@ -370,7 +370,7 @@ class Tensor(object):
See Also
--------
`torch.bitwise_xor(...)`_ : Compute the element-wise XOR bitwise operation.
`torch.bitwise_xor(...)`_
"""
......@@ -391,7 +391,7 @@ class Tensor(object):
See Also
--------
`torch.bitwise_xor(...)`_ : Compute the element-wise XOR bitwise operation.
`torch.bitwise_xor(...)`_
"""
......@@ -447,7 +447,7 @@ class Tensor(object):
See Also
--------
`torch.ceil(...)`_ : Compute the smallest integer not less than input.
`torch.ceil(...)`_
"""
......@@ -463,7 +463,7 @@ class Tensor(object):
See Also
--------
`torch.ceil(...)`_ : Compute the smallest integer not less than input.
`torch.ceil(...)`_
"""
......@@ -521,7 +521,7 @@ class Tensor(object):
See Also
--------
`torch.clamp(...)`_ : Compute the clipped input according to the given bounds.
`torch.clamp(...)`_
"""
......@@ -542,7 +542,7 @@ class Tensor(object):
See Also
--------
`torch.clamp(...)`_ : Compute the clipped input according to the given bounds.
`torch.clamp(...)`_
"""
......@@ -585,7 +585,7 @@ class Tensor(object):
See Also
--------
`torch.cos(...)`_ : Compute the cos of input.
`torch.cos(...)`_
"""
......@@ -638,7 +638,7 @@ class Tensor(object):
See Also
--------
`torch.cumsum(...)`_ : Compute the cumulative sum of elements along the given dimension.
`torch.cumsum(...)`_
"""
......@@ -681,7 +681,7 @@ class Tensor(object):
See Also
--------
`torch.div(...)`_ : Compute the element-wise division.
`torch.div(...)`_
"""
......@@ -702,7 +702,7 @@ class Tensor(object):
See Also
--------
`torch.div(...)`_ : Compute the element-wise division.
`torch.div(...)`_
"""
......@@ -743,7 +743,7 @@ class Tensor(object):
See Also
--------
`torch.eq(...)`_ : Compute the element-wise equal comparison.
`torch.eq(...)`_
"""
......@@ -759,7 +759,7 @@ class Tensor(object):
See Also
--------
`torch.exp(...)`_ : Compute the exponential of input.
`torch.exp(...)`_
"""
......@@ -778,7 +778,7 @@ class Tensor(object):
See Also
--------
`torch.expand(...)`_ : Broadcast input according to given sizes.
`torch.expand(...)`_
"""
......@@ -797,7 +797,7 @@ class Tensor(object):
See Also
--------
`torch.expand(...)`_ : Broadcast input according to given sizes.
`torch.expand(...)`_
"""
return self.expand(*other.size())
......@@ -851,7 +851,7 @@ class Tensor(object):
See Also
--------
`torch.floor(...)`_ : Compute the largest integer not greater than input.
`torch.floor(...)`_
"""
......@@ -867,7 +867,7 @@ class Tensor(object):
See Also
--------
`torch.floor(...)`_ : Compute the largest integer not greater than input.
`torch.floor(...)`_
"""
......@@ -888,7 +888,7 @@ class Tensor(object):
See Also
--------
`torch.ge(...)`_ : Compute the element-wise greater-equal comparison.
`torch.ge(...)`_
"""
......@@ -909,7 +909,7 @@ class Tensor(object):
See Also
--------
`torch.gt(...)`_ : Compute the element-wise greater comparison.
`torch.gt(...)`_
"""
......@@ -1000,7 +1000,7 @@ class Tensor(object):
See Also
--------
`torch.le(...)`_ : Compute the element-wise less-equal comparison.
`torch.le(...)`_
"""
......@@ -1072,7 +1072,7 @@ class Tensor(object):
See Also
--------
`torch.lt(...)`_ : Compute the element-wise less comparison.
`torch.lt(...)`_
"""
......@@ -1183,7 +1183,7 @@ class Tensor(object):
See Also
--------
`torch.mul(...)`_ : Compute the element-wise multiplication.
`torch.mul(...)`_
"""
......@@ -1204,7 +1204,7 @@ class Tensor(object):
See Also
--------
`torch.mul(...)`_ : Compute the element-wise multiplication.
`torch.mul(...)`_
"""
......@@ -1273,7 +1273,7 @@ class Tensor(object):
See Also
--------
`torch.ne(...)`_ : Compute the element-wise not-equal comparison.
`torch.ne(...)`_
"""
......@@ -1289,18 +1289,24 @@ class Tensor(object):
See Also
--------
`torch.neg(...)`_ : Compute the element-wise negative.
`torch.neg(...)`_
"""
def nonzero(self):
"""Return the index of non-zero elements.
r"""Return the index of non-zero elements.
.. math:: \text{out} = \{i\}, \text{ if } \text{self}_{i} \neq 0
Returns
-------
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.nonzero(...)`_
"""
def normal_(self, mean=0, std=1):
......@@ -1392,7 +1398,7 @@ class Tensor(object):
See Also
--------
`torch.pow(...)`_ : Compute the power of input.
`torch.pow(...)`_
"""
......@@ -1408,7 +1414,7 @@ class Tensor(object):
See Also
--------
`torch.reciprocal(...)`_ : Compute the reciprocal of input.
`torch.reciprocal(...)`_
"""
......@@ -1424,7 +1430,7 @@ class Tensor(object):
See Also
--------
`torch.reciprocal(...)`_ : Compute the reciprocal of input.
`torch.reciprocal(...)`_
"""
......@@ -1458,7 +1464,7 @@ class Tensor(object):
See Also
--------
`torch.reshape(...)`_ : Change the shape of input.
`torch.reshape(...)`_
"""
......@@ -1477,7 +1483,7 @@ class Tensor(object):
See Also
--------
`torch.reshape(...)`_ : Change the shape of input.
`torch.reshape(...)`_
"""
......@@ -1498,7 +1504,7 @@ class Tensor(object):
See Also
--------
`torch.round(...)`_ : Compute the nearest integer of input.
`torch.round(...)`_
"""
......@@ -1514,7 +1520,7 @@ class Tensor(object):
See Also
--------
`torch.round(...)`_ : Compute the nearest integer of input.
`torch.round(...)`_
"""
......@@ -1530,7 +1536,7 @@ class Tensor(object):
See Also
--------
`torch.rsqrt(...)`_ : Compute the square root of input.
`torch.rsqrt(...)`_
"""
......@@ -1546,7 +1552,7 @@ class Tensor(object):
See Also
--------
`torch.rsqrt(...)`_ : Compute the square root of input.
`torch.rsqrt(...)`_
"""
......@@ -1568,7 +1574,7 @@ class Tensor(object):
See Also
--------
`torch.sign(...)`_ : Compute the sign indication of input.
`torch.sign(...)`_
"""
......@@ -1590,7 +1596,7 @@ class Tensor(object):
See Also
--------
`torch.sign(...)`_ : Compute the sign indication of input.
`torch.sign(...)`_
"""
......@@ -1606,7 +1612,7 @@ class Tensor(object):
See Also
--------
`torch.sin(...)`_ : Compute the sin of input.
`torch.sin(...)`_
"""
......@@ -1639,7 +1645,7 @@ class Tensor(object):
See Also
--------
`torch.sqrt(...)`_ : Compute the square root of input.
`torch.sqrt(...)`_
"""
......@@ -1655,7 +1661,7 @@ class Tensor(object):
See Also
--------
`torch.sqrt(...)`_ : Compute the square root of input.
`torch.sqrt(...)`_
"""
......@@ -1672,10 +1678,14 @@ class Tensor(object):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.squeeze(...)`_
"""
def squeeze_(self, dim=None):
"""Inplace version of ``Tensor.squeeze()``.
"""Remove the dimensions with size 1.
Parameters
----------
......@@ -1687,6 +1697,10 @@ class Tensor(object):
dragon.vm.torch.Tensor
The self.
See Also
--------
`torch.squeeze(...)`_
"""
def sum(self, dim=None, keepdim=False):
......@@ -1704,6 +1718,10 @@ class Tensor(object):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.sum(...)`_
"""
def sub(self, other):
......@@ -1723,7 +1741,7 @@ class Tensor(object):
See Also
--------
`torch.sub(...)`_ : Compute the element-wise subtraction.
`torch.sub(...)`_
"""
......@@ -1744,7 +1762,7 @@ class Tensor(object):
See Also
--------
`torch.sub(...)`_ : Compute the element-wise subtraction.
`torch.sub(...)`_
"""
......@@ -1769,7 +1787,7 @@ class Tensor(object):
See Also
--------
`torch.topk(...)`_ : Return the top-K largest or smallest elements along the given dimension.
`torch.topk(...)`_
"""
......@@ -1822,10 +1840,14 @@ class Tensor(object):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.unsqueeze(...)`_
"""
def unsqueeze_(self, dim):
"""In-place version of ``Tensor.unsqueeze()``.
"""Insert the dimensions of size 1.
Parameters
----------
......@@ -1837,6 +1859,10 @@ class Tensor(object):
dragon.vm.torch.Tensor
The self.
See Also
--------
`torch.unsqueeze(...)`_
"""
def view(self, *shape):
......@@ -1854,7 +1880,7 @@ class Tensor(object):
See Also
--------
`torch.reshape(...)`_ : Change the shape of input.
`torch.reshape(...)`_
"""
return self.reshape(shape)
......@@ -1874,7 +1900,7 @@ class Tensor(object):
See Also
--------
`torch.reshape(...)`_ : Change the shape of input.
`torch.reshape(...)`_
"""
return self.reshape_(shape)
......@@ -1918,6 +1944,10 @@ class Tensor(object):
dragon.vm.torch.Tensor
The output tensor.
See Also
--------
`torch.where(...)`_
"""
def zero_(self):
......
Markdown is supported
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!