make init bias better, 
I think `detach` is more safer than `data`

Eliminate 1 op during training and inference.

* CLI fire prep updates

* revert unintentional TF export change

* Sidestep os.path.relpath() Windows bug

os.path.relpath() seems to have a major bug on Windows due to Windows horrible path handling. This fix attempts to sidestep the issue.

```
File "C:\Users\mkokg/.cache\torch\hub\ultralytics_yolov5_master\export.py", line 64, in
ROOT = Path(os.path.relpath(ROOT, Path.cwd())) # relative
File "C:\Users\mkokg\AppData\Local\Programs\Python\Python310\lib\ntpath.py", line 718, in relpath
raise ValueError("path is on mount %r, start on mount %r" % (
ValueError: path is on mount 'C:', start on mount 'D:'
```

* Update yolo.py

* Update yolo.py

* Update yolo.py

* Update export.py

* Remove named arguments where possible

Speed improvements.

* Update yolo.py

* Update yolo.py

* Update yolo.py

* Update `check_anchor_order()`

Use mean area per output layer for added stability.

* Check in pixel-space not grid-space fix

Slightly more efficient than .to(device)

Fix for Colab hub error:


```python
import yaml

with open('yolov5s.yaml', errors='ignore') as f:
     d = yaml.safe_load(f)  # model dict

print(d)

---------------------------------------------------------------------------
ReaderError                               Traceback (most recent call last)
<ipython-input-8-1150b5143073> in <module>()
      2 
      3 with open('yolov5s.yaml', errors='ignore') as f:
----> 4      d = yaml.safe_load(f)  # model dict
      5 
      6 print(d)

6 frames
/usr/local/lib/python3.7/dist-packages/yaml/reader.py in check_printable(self, data)
    142             position = self.index+(len(self.buffer)-self.pointer)+match.start()
    143             raise ReaderError(self.name, position, ord(character),
--> 144                     'unicode', "special characters are not allowed")
    145 
    146     def update(self, length):

ReaderError: unacceptable character #x1f680: special characters are not allowed
  in "yolov5s.yaml", position 9
```

* Update model yamls for v6.0

* Add python models/yolo.py --test

* Ghost fix

Checking for `onnx_dynamic` first should suppress the warning:

```log
TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if self.grid[i].shape[2:4] != x[i].shape[2:4] or self.onnx_dynamic
```

* Fix float zeros format

* 255 to integer

* Update `set_logging()`

* Update export.py

* pre-commit fixes

* Update LoadImages

* Update LoadStreams

* Update print_args

* Single LOGGER definition

* yolo.py fix

Co-authored-by: pre-commit <pre-commit@example.com>

* Meshgrid `indexing='ij'` for PyTorch 1.10

Will not merge currently as breaks backwards compatibility.

* Meshgrid `indexing='ij'` for PyTorch 1.10

Will not merge currently as breaks backwards compatibility.

* Add check_version hard argument

* Update comment

* define pre-commit

* add CI code

* configure

* apply pre-commit

* fstring

* apply MD

* pre-commit

* Update torch_utils.py

* Update print strings

* notes

* Cleanup code-format.yml

* Update setup.cfg

* Update .pre-commit-config.yaml

Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* refactor anchors and anchor_grid in Detect Layer

* fix CI failures by adding compatibility

* fix tf failure

* fix different devices errors

* Cleanup

* fix anchors overwriting issue

* better refactoring

* Remove self.anchor_grid shape check (redundant with self.grid check)

Also PEP8 / 120 line width

* Convert _make_grid() from static to dynamic method

* Remove anchor_grid.to(device)

clone() should already clone to same device as self.anchors

* fix different devices error

Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* fix different devices bug

* extend _apply() instead of to() for a general fix

* Only apply if Detect() is last layer

Co-authored-by: Jebastin Nadar <njebastin10@gmail.com>

* Indent fix

* Add comment to yolo.py

* Add comment to common.py

Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Clip TTA Augmented Tails

Experimental TTA update.

* Update yolo.py

* Update yolo.py

* Update yolo.py

* Update yolo.py

* Fix E722, do not use bare 'except'

* Remove used codes

* Add FileNotFoundError in LoadImagesAndLabels

* Remove AssertionError

* Ignore LoadImagesAndLabels

* Ignore downloads.py

* Ignore torch_utils.py

* Ignore train.py

* Ignore datasets.py

* Enable utils/download.py

* Fixing exception in thop

* Remove unused code

* Fixing exception in LoadImagesAndLabels

* Fixing exception in exif_size

* Fixing exception in parse_model

* Ignore exceptions in requests

* Revert the exception as suggested

* Revert the exception as suggested

* Fix relative `ROOT` Pytorch Hub custom model bug

* Update yolo.py

* Update relative `ROOT` logic

* python 3.9 Path().is_relative_to() removal

* Update `sys.path.append(str(ROOT))`

* Cleanup

* Add suffix checks

* Cleanup

* Cleanup2

* Cleanup3

Profiling fix copies input to Detect layer to circumvent inplace changes to the feature maps.

* Add `SPPF()` layer

* Cleanup

* Add credit

* Implement new headers

* Reformat 1

* Reformat 2

* Reformat 3 - math

* Reformat 4 - yaml

* Add yolov5s-ghost.yaml

* Finish C3Ghost

* Add C3Ghost to list

* Add C3Ghost to number of repeats if statement

* Fixes

* Cleanup

Co-authored-by: yangyuantao <yangyuantao@transai.cn>

* more simpler code for DWConvClass

more simpler code for DWConvClass

* remove DWConv function

* Replace DWConvClass with DWConv

* Add `DWConvClass()`

* Cleanup

* Cleanup2