mirror of
https://github.com/open-mmlab/mmcv.git
synced 2025-06-03 21:54:52 +08:00
Fix typos (#1041)
* Fix typos Signed-off-by: lizz <lizz@sensetime.com> * Add deprecation warning Signed-off-by: lizz <lizz@sensetime.com>
This commit is contained in:
lizz
GitHub
parent
9d1436fb6c
commit
13888df2aa
@ -370,9 +370,9 @@ Let us introduce the usage of `initialize` in detail.
|
||||
|
||||
`BaseModule` is inherited from `torch.nn.Module`, and the only different between them is that `BaseModule` implements `init_weight`.
|
||||
|
||||
`Sequential` is inhertied from `BaseModule` and `torch.nn.Sequential`.
|
||||
`Sequential` is inherited from `BaseModule` and `torch.nn.Sequential`.
|
||||
|
||||
`ModuleList` is inhertied from `BaseModule` and `torch.nn.ModuleList`.
|
||||
`ModuleList` is inherited from `BaseModule` and `torch.nn.ModuleList`.
|
||||
|
||||
`````python
|
||||
import torch.nn as nn
|
||||
@ -534,5 +534,5 @@ The following types are supported for `filename` argument of `mmcv.load_checkpoi
|
||||
|
||||
- filepath: The filepath of the checkpoint.
|
||||
- `http://xxx` and `https://xxx`: The link to download the checkpoint. The `SHA256` postfix should be contained in the filename.
|
||||
- `torchvison://xxx`: The model links in `torchvision.models`.Please refer to [torchvision](https://pytorch.org/docs/stable/torchvision/models.html) for details.
|
||||
- `torchvision://xxx`: The model links in `torchvision.models`.Please refer to [torchvision](https://pytorch.org/docs/stable/torchvision/models.html) for details.
|
||||
- `open-mmlab://xxx`: The model links or filepath provided in default and additional json files.
|
||||
|
||||
@ -88,7 +88,7 @@ Here is an example.
|
||||
import torch
|
||||
import onnx
|
||||
|
||||
from mmcv.tensorrt import (TRTWraper, onnx2trt, save_trt_engine,
|
||||
from mmcv.tensorrt import (TRTWrapper, onnx2trt, save_trt_engine,
|
||||
is_tensorrt_plugin_loaded)
|
||||
|
||||
assert is_tensorrt_plugin_loaded(), 'Requires to complie TensorRT plugins in mmcv'
|
||||
@ -117,7 +117,7 @@ trt_engine = onnx2trt(
|
||||
save_trt_engine(trt_engine, trt_file)
|
||||
|
||||
# Run inference with TensorRT
|
||||
trt_model = TRTWraper(trt_file, ['input'], ['output'])
|
||||
trt_model = TRTWrapper(trt_file, ['input'], ['output'])
|
||||
|
||||
with torch.no_grad():
|
||||
trt_outputs = trt_model({'input': inputs})
|
||||
@ -161,7 +161,7 @@ Below are the main steps:
|
||||
|
||||
### Reminders
|
||||
|
||||
- Some of the [custom ops](https://mmcv.readthedocs.io/en/latest/ops.html) in `mmcv` have their cuda implementations, which could be refered.
|
||||
- Some of the [custom ops](https://mmcv.readthedocs.io/en/latest/ops.html) in `mmcv` have their cuda implementations, which could be referred.
|
||||
|
||||
## Known Issues
|
||||
|
||||
|
||||
@ -69,7 +69,7 @@ a = 1
|
||||
b = dict(b1=[0, 1, 2], b2=None)
|
||||
```
|
||||
|
||||
#### Inherit from base config without overlaped keys
|
||||
#### Inherit from base config without overlapped keys
|
||||
|
||||
`config_b.py`
|
||||
|
||||
@ -90,7 +90,7 @@ d = 'string'
|
||||
|
||||
New fields in `config_b.py` are combined with old fields in `config_a.py`
|
||||
|
||||
#### Inherit from base config with overlaped keys
|
||||
#### Inherit from base config with overlapped keys
|
||||
|
||||
`config_c.py`
|
||||
|
||||
@ -203,7 +203,7 @@ for i, task in enumerate(mmcv.track_iter_progress(tasks)):
|
||||
|
||||
### Timer
|
||||
|
||||
It is convinient to compute the runtime of a code block with `Timer`.
|
||||
It is convenient to compute the runtime of a code block with `Timer`.
|
||||
|
||||
```python
|
||||
import time
|
||||
|
||||
@ -237,7 +237,7 @@ def print_model_with_flops(model,
|
||||
|
||||
>>> model = ExampleModel()
|
||||
>>> x = (3, 16, 16)
|
||||
to print the complexity inforamtion state for each layer, you can use
|
||||
to print the complexity information state for each layer, you can use
|
||||
>>> get_model_complexity_info(model, x)
|
||||
or directly use
|
||||
>>> print_model_with_flops(model, 4579784.0, 37361)
|
||||
|
||||
@ -98,7 +98,7 @@ class BaseInit(object):
|
||||
def __init__(self, *, bias=0, bias_prob=None, layer=None):
|
||||
self.wholemodule = False
|
||||
if not isinstance(bias, (int, float)):
|
||||
raise TypeError(f'bias must be a numbel, but got a {type(bias)}')
|
||||
raise TypeError(f'bias must be a number, but got a {type(bias)}')
|
||||
|
||||
if bias_prob is not None:
|
||||
if not isinstance(bias_prob, float):
|
||||
|
||||
@ -28,13 +28,13 @@ def list_from_file(filename, prefix='', offset=0, max_num=0):
|
||||
def dict_from_file(filename, key_type=str):
|
||||
"""Load a text file and parse the content as a dict.
|
||||
|
||||
Each line of the text file will be two or more columns splited by
|
||||
Each line of the text file will be two or more columns split by
|
||||
whitespaces or tabs. The first column will be parsed as dict keys, and
|
||||
the following columns will be parsed as dict values.
|
||||
|
||||
Args:
|
||||
filename(str): Filename.
|
||||
key_type(type): Type of the dict's keys. str is user by default and
|
||||
key_type(type): Type of the dict keys. str is user by default and
|
||||
type conversion will be performed if specified.
|
||||
|
||||
Returns:
|
||||
|
||||
@ -528,7 +528,7 @@ def _get_shear_matrix(magnitude, direction='horizontal'):
|
||||
|
||||
Args:
|
||||
magnitude (int | float): The magnitude used for shear.
|
||||
direction (str): Thie flip direction, either "horizontal"
|
||||
direction (str): The flip direction, either "horizontal"
|
||||
or "vertical".
|
||||
|
||||
Returns:
|
||||
@ -552,7 +552,7 @@ def imshear(img,
|
||||
img (ndarray): Image to be sheared with format (h, w)
|
||||
or (h, w, c).
|
||||
magnitude (int | float): The magnitude used for shear.
|
||||
direction (str): Thie flip direction, either "horizontal"
|
||||
direction (str): The flip direction, either "horizontal"
|
||||
or "vertical".
|
||||
border_value (int | tuple[int]): Value used in case of a
|
||||
constant border.
|
||||
|
||||
@ -140,7 +140,7 @@ def imread(img_or_path, flag='color', channel_order='bgr', backend=None):
|
||||
it will be returned as is.
|
||||
flag (str): Flags specifying the color type of a loaded image,
|
||||
candidates are `color`, `grayscale` and `unchanged`.
|
||||
Note that the `turbojpeg` backened does not support `unchanged`.
|
||||
Note that the `turbojpeg` backend does not support `unchanged`.
|
||||
channel_order (str): Order of channel, candidates are `bgr` and `rgb`.
|
||||
backend (str | None): The image decoding backend type. Options are
|
||||
`cv2`, `pillow`, `turbojpeg`, `tifffile`, `None`.
|
||||
@ -234,7 +234,7 @@ def imwrite(img, file_path, params=None, auto_mkdir=True):
|
||||
Args:
|
||||
img (ndarray): Image array to be written.
|
||||
file_path (str): Image file path.
|
||||
params (None or list): Same as opencv's :func:`imwrite` interface.
|
||||
params (None or list): Same as opencv :func:`imwrite` interface.
|
||||
auto_mkdir (bool): If the parent folder of `file_path` does not exist,
|
||||
whether to create it automatically.
|
||||
|
||||
|
||||
@ -119,7 +119,7 @@ def adjust_color(img, alpha=1, beta=None, gamma=0):
|
||||
beta = 1 - alpha
|
||||
colored_img = cv2.addWeighted(img, alpha, gray_img, beta, gamma)
|
||||
if not colored_img.dtype == np.uint8:
|
||||
# Note when the dtype of `img` is not defaultly `np.uint8`
|
||||
# Note when the dtype of `img` is not the default `np.uint8`
|
||||
# (e.g. np.float32), the value in `colored_img` got from cv2
|
||||
# is not guaranteed to be in range [0, 255], so here clip
|
||||
# is needed.
|
||||
@ -320,9 +320,9 @@ def adjust_sharpness(img, factor=1., kernel=None):
|
||||
# adopted from PIL.ImageFilter.SMOOTH
|
||||
kernel = np.array([[1., 1., 1.], [1., 5., 1.], [1., 1., 1.]]) / 13
|
||||
assert isinstance(kernel, np.ndarray), \
|
||||
f'kernel must be of type np.ndarrray, but got {type(kernel)} instead.'
|
||||
f'kernel must be of type np.ndarray, but got {type(kernel)} instead.'
|
||||
assert kernel.ndim == 2, \
|
||||
f'kernel must have a dimention of 2, but got {kernel.ndim} instead.'
|
||||
f'kernel must have a dimension of 2, but got {kernel.ndim} instead.'
|
||||
|
||||
degenerated = cv2.filter2D(img, -1, kernel)
|
||||
sharpened_img = cv2.addWeighted(
|
||||
@ -340,13 +340,13 @@ def adjust_lighting(img, eigval, eigvec, alphastd=0.1, to_rgb=True):
|
||||
<https://dl.acm.org/doi/pdf/10.1145/3065386>`_.
|
||||
|
||||
Args:
|
||||
img (ndarray): Image to be ajusted lighting. BGR order.
|
||||
img (ndarray): Image to be adjusted lighting. BGR order.
|
||||
eigval (ndarray): the eigenvalue of the convariance matrix of pixel
|
||||
values, respectively.
|
||||
eigvec (ndarray): the eigenvector of the convariance matrix of pixel
|
||||
values, respectively.
|
||||
alphastd (float): The standard deviation for distribution of alpha.
|
||||
Dafaults to 0.1
|
||||
Defaults to 0.1
|
||||
to_rgb (bool): Whether to convert img to rgb.
|
||||
|
||||
Returns:
|
||||
|
||||
@ -49,7 +49,7 @@ def bbox_overlaps(bboxes1, bboxes2, mode='iou', aligned=False, offset=0):
|
||||
mode_dict = {'iou': 0, 'iof': 1}
|
||||
assert mode in mode_dict.keys()
|
||||
mode_flag = mode_dict[mode]
|
||||
# Either the boxes are empty or the length of boxes's last dimenstion is 4
|
||||
# Either the boxes are empty or the length of boxes' last dimension is 4
|
||||
assert (bboxes1.size(-1) == 4 or bboxes1.size(0) == 0)
|
||||
assert (bboxes2.size(-1) == 4 or bboxes2.size(0) == 0)
|
||||
assert offset == 1 or offset == 0
|
||||
|
||||
@ -195,15 +195,15 @@ class FusedBiasLeakyReLU(nn.Module):
|
||||
|
||||
The bias term comes from the convolution operation. In addition, to keep
|
||||
the variance of the feature map or gradients unchanged, they also adopt a
|
||||
scale similarly with Kaiming initalization. However, since the
|
||||
scale similarly with Kaiming initialization. However, since the
|
||||
:math:`1 + \alpha^2` : is too small, we can just ignore it. Therefore, the
|
||||
final sacle is just :math:`\sqrt{2}`:. Of course, you may change it with # noqa: W605, E501
|
||||
final scale is just :math:`\sqrt{2}`:. Of course, you may change it with # noqa: W605, E501
|
||||
your own scale.
|
||||
|
||||
TODO: Implement the CPU version.
|
||||
|
||||
Args:
|
||||
channel (int): The channnel number of the feature map.
|
||||
channel (int): The channel number of the feature map.
|
||||
negative_slope (float, optional): Same as nn.LeakyRelu.
|
||||
Defaults to 0.2.
|
||||
scale (float, optional): A scalar to adjust the variance of the feature
|
||||
@ -230,9 +230,9 @@ def fused_bias_leakyrelu(input, bias, negative_slope=0.2, scale=2**0.5):
|
||||
|
||||
The bias term comes from the convolution operation. In addition, to keep
|
||||
the variance of the feature map or gradients unchanged, they also adopt a
|
||||
scale similarly with Kaiming initalization. However, since the
|
||||
scale similarly with Kaiming initialization. However, since the
|
||||
:math:`1 + \alpha^2` : is too small, we can just ignore it. Therefore, the
|
||||
final sacle is just :math:`\sqrt{2}`:. Of course, you may change it with # noqa: W605, E501
|
||||
final scale is just :math:`\sqrt{2}`:. Of course, you may change it with # noqa: W605, E501
|
||||
your own scale.
|
||||
|
||||
Args:
|
||||
|
||||
@ -10,7 +10,7 @@ from ..cnn import ConvModule
|
||||
class BaseMergeCell(nn.Module):
|
||||
"""The basic class for cells used in NAS-FPN and NAS-FCOS.
|
||||
|
||||
BaseMergeCell takes 2 inputs. After applying concolution
|
||||
BaseMergeCell takes 2 inputs. After applying convolution
|
||||
on them, they are resized to the target size. Then,
|
||||
they go through binary_op, which depends on the type of cell.
|
||||
If with_out_conv is True, the result of output will go through
|
||||
|
||||
@ -350,7 +350,7 @@ def nms_rotated(dets, scores, iou_threshold, labels=None):
|
||||
be in (x_ctr, y_ctr, width, height, angle_radian) format.
|
||||
scores (Tensor): scores in shape (N, ).
|
||||
iou_threshold (float): IoU thresh for NMS.
|
||||
labels (Tensor): boxes's label in shape (N,).
|
||||
labels (Tensor): boxes' label in shape (N,).
|
||||
|
||||
Returns:
|
||||
tuple: kept dets(boxes and scores) and indice, which is always the \
|
||||
|
||||
@ -14,7 +14,7 @@ def pixel_group(score, mask, embedding, kernel_label, kernel_contour,
|
||||
Arguments:
|
||||
score (np.array or Tensor): The foreground score with size hxw.
|
||||
mask (np.array or Tensor): The foreground mask with size hxw.
|
||||
embedding (np.array or Tensor): The emdedding with size hxwxc to
|
||||
embedding (np.array or Tensor): The embedding with size hxwxc to
|
||||
distinguish instances.
|
||||
kernel_label (np.array or Tensor): The instance kernel index with
|
||||
size hxw.
|
||||
|
||||
@ -23,7 +23,7 @@ def scatter(input, devices, streams=None):
|
||||
with torch.cuda.device(devices[0]), torch.cuda.stream(stream):
|
||||
output = output.cuda(devices[0], non_blocking=True)
|
||||
else:
|
||||
# unsquzee the first dimension thus the tensor's shape is the
|
||||
# unsqueeze the first dimension thus the tensor's shape is the
|
||||
# same as those scattered with GPU.
|
||||
output = output.unsqueeze(0)
|
||||
return output
|
||||
|
||||
@ -22,7 +22,7 @@ class BaseModule(nn.Module, metaclass=ABCMeta):
|
||||
|
||||
super(BaseModule, self).__init__()
|
||||
# define default value of init_cfg instead of hard code
|
||||
# in init_weigt() function
|
||||
# in init_weight() function
|
||||
self._is_init = False
|
||||
self.init_cfg = init_cfg
|
||||
|
||||
|
||||
@ -361,7 +361,7 @@ class CyclicLrUpdaterHook(LrUpdaterHook):
|
||||
Implement the cyclical learning rate policy (CLR) described in
|
||||
https://arxiv.org/pdf/1506.01186.pdf
|
||||
|
||||
Different from the original paper, we use cosine anealing rather than
|
||||
Different from the original paper, we use cosine annealing rather than
|
||||
triangular policy inside a cycle. This improves the performance in the
|
||||
3D detection area.
|
||||
|
||||
|
||||
@ -59,7 +59,7 @@ if TORCH_VERSION != 'parrots' and TORCH_VERSION >= '1.6.0':
|
||||
It can also be a dict containing arguments of GradScalar.
|
||||
Defaults to 512. For Pytorch >= 1.6, mmcv uses official
|
||||
implementation of GradScaler. If you use a dict version of
|
||||
loss_scale to create GradScaler, plese refer to:
|
||||
loss_scale to create GradScaler, please refer to:
|
||||
https://pytorch.org/docs/stable/amp.html#torch.cuda.amp.GradScaler
|
||||
for the parameters.
|
||||
|
||||
|
||||
@ -10,7 +10,7 @@ from .hook import HOOKS, Hook
|
||||
|
||||
@HOOKS.register_module()
|
||||
class ProfilerHook(Hook):
|
||||
"""Profiler to analyze perfromance during training.
|
||||
"""Profiler to analyze performance during training.
|
||||
|
||||
PyTorch Profiler is a tool that allows the collection of the performance
|
||||
metrics during the training. More details on Profiler can be found at
|
||||
@ -67,7 +67,7 @@ class ProfilerHook(Hook):
|
||||
from torch import profiler # torch version >= 1.8.1
|
||||
except ImportError:
|
||||
raise ImportError('profiler is the new feature of torch1.8.1, '
|
||||
f'but your verison is {torch.__version__}')
|
||||
f'but your version is {torch.__version__}')
|
||||
|
||||
assert isinstance(by_epoch, bool), '``by_epoch`` should be a boolean.'
|
||||
self.by_epoch = by_epoch
|
||||
@ -120,10 +120,10 @@ class ProfilerHook(Hook):
|
||||
trace_type = trace_cfg.pop('type') # log_trace handler
|
||||
if trace_type == 'log_trace':
|
||||
|
||||
def _log_hanlder(prof):
|
||||
def _log_handler(prof):
|
||||
print(prof.key_averages().table(**trace_cfg))
|
||||
|
||||
_on_trace_ready = _log_hanlder
|
||||
_on_trace_ready = _log_handler
|
||||
elif trace_type == 'tb_trace': # tensorboard_trace handler
|
||||
try:
|
||||
import torch_tb_profiler # noqa: F401
|
||||
|
||||
@ -51,7 +51,7 @@ class DefaultOptimizerConstructor:
|
||||
``dcn_offset_lr_mult``. If you wish to apply both of them to the
|
||||
offset layer in deformable convs, set ``dcn_offset_lr_mult``
|
||||
to the original ``dcn_offset_lr_mult`` * ``bias_lr_mult``.
|
||||
2. If the option ``dcn_offset_lr_mult`` is used, the construtor will
|
||||
2. If the option ``dcn_offset_lr_mult`` is used, the constructor will
|
||||
apply it to all the DCN layers in the model. So be carefull when
|
||||
the model contains multiple DCN layers in places other than
|
||||
backbone.
|
||||
|
||||
@ -1,12 +1,12 @@
|
||||
# flake8: noqa
|
||||
from .init_plugins import is_tensorrt_plugin_loaded, load_tensorrt_plugin
|
||||
from .tensorrt_utils import (TRTWraper, load_trt_engine, onnx2trt,
|
||||
from .tensorrt_utils import (TRTWraper, TRTWrapper, load_trt_engine, onnx2trt,
|
||||
save_trt_engine)
|
||||
|
||||
# load tensorrt plugin lib
|
||||
load_tensorrt_plugin()
|
||||
|
||||
__all__ = [
|
||||
'onnx2trt', 'save_trt_engine', 'load_trt_engine', 'TRTWraper',
|
||||
'is_tensorrt_plugin_loaded'
|
||||
'onnx2trt', 'save_trt_engine', 'load_trt_engine', 'TRTWrapper',
|
||||
'TRTWraper', 'is_tensorrt_plugin_loaded'
|
||||
]
|
||||
|
||||
@ -1,3 +1,5 @@
|
||||
import warnings
|
||||
|
||||
import numpy as np
|
||||
import onnx
|
||||
import tensorrt as trt
|
||||
@ -40,7 +42,7 @@ def preprocess_onnx(onnx_model):
|
||||
elif name in init_dict:
|
||||
raw_data = init_dict[name].raw_data
|
||||
else:
|
||||
raise ValueError(f'{name} not found in node or initilizer.')
|
||||
raise ValueError(f'{name} not found in node or initializer.')
|
||||
return np.frombuffer(raw_data, typ).item()
|
||||
|
||||
nrof_node = len(nodes)
|
||||
@ -225,8 +227,8 @@ def torch_device_from_trt(device):
|
||||
return TypeError('%s is not supported by torch' % device)
|
||||
|
||||
|
||||
class TRTWraper(torch.nn.Module):
|
||||
"""TensorRT engine Wraper.
|
||||
class TRTWrapper(torch.nn.Module):
|
||||
"""TensorRT engine Wrapper.
|
||||
|
||||
Arguments:
|
||||
engine (tensorrt.ICudaEngine): TensorRT engine to wrap
|
||||
@ -239,7 +241,7 @@ class TRTWraper(torch.nn.Module):
|
||||
"""
|
||||
|
||||
def __init__(self, engine, input_names=None, output_names=None):
|
||||
super(TRTWraper, self).__init__()
|
||||
super(TRTWrapper, self).__init__()
|
||||
self.engine = engine
|
||||
if isinstance(self.engine, str):
|
||||
self.engine = load_trt_engine(engine)
|
||||
@ -247,7 +249,7 @@ class TRTWraper(torch.nn.Module):
|
||||
if not isinstance(self.engine, trt.ICudaEngine):
|
||||
raise TypeError('engine should be str or trt.ICudaEngine')
|
||||
|
||||
self._register_state_dict_hook(TRTWraper._on_state_dict)
|
||||
self._register_state_dict_hook(TRTWrapper._on_state_dict)
|
||||
self.context = self.engine.create_execution_context()
|
||||
|
||||
# get input and output names from engine
|
||||
@ -310,3 +312,11 @@ class TRTWraper(torch.nn.Module):
|
||||
torch.cuda.current_stream().cuda_stream)
|
||||
|
||||
return outputs
|
||||
|
||||
|
||||
class TRTWraper(TRTWrapper):
|
||||
|
||||
def __init__(self, *args, **kwargs):
|
||||
super().__init__(*args, **kwargs)
|
||||
warnings.warn('TRTWraper will be deprecated in'
|
||||
' future. Please use TRTWrapper instead')
|
||||
|
||||
@ -266,7 +266,7 @@ def requires_executable(prerequisites):
|
||||
|
||||
|
||||
def deprecated_api_warning(name_dict, cls_name=None):
|
||||
"""A decorator to check if some argments are deprecate and try to replace
|
||||
"""A decorator to check if some arguments are deprecate and try to replace
|
||||
deprecate src_arg_name to dst_arg_name.
|
||||
|
||||
Args:
|
||||
|
||||
@ -15,7 +15,7 @@ def imshow(img, win_name='', wait_time=0):
|
||||
wait_time (int): Value of waitKey param.
|
||||
"""
|
||||
cv2.imshow(win_name, imread(img))
|
||||
if wait_time == 0: # prevent from hangning if windows was closed
|
||||
if wait_time == 0: # prevent from hanging if windows was closed
|
||||
while True:
|
||||
ret = cv2.waitKey(1)
|
||||
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user