* Update ci-testing.yml

* Update ci-testing.yml

* Update ci-testing.yml

* Update docker.yml

Fixed FP32 TRT  model export
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

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

Simplify and update `--save-period` epoch 0

* Pin `protobuf==3.20.1`

* Update requirements.txt

* Update ci-testing.yml

Pin protobuf==3.20.1

Resolve protobuf 4 issues.

Avoids openvino-dev conflicts

* Add warning emoji on `--conf > 0.001`

* Update val.py

* Update and rename ci-testing.yml to ci.yml

* Update ci.yml

* Update ci.yml

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update and rename ci.yml to ci-tests.yml

* Update ci-tests.yml

* Update ci-tests.yml

* Rename ci-tests.yml to ci-testing.yml
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Attempt to resolve Windows issues

* Add `psutil` and `ipython` to requirements.txt

Lightweight packages used by YOLOv5 for system utilization (psutil) and interactive notebooks (IPython)

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* sort alphabetically
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Remove `formats` variable to avoid `pd` conflict

* Update export.py

Update requirements.txt

* Reduce AMP check to detections verification

More robust and faster

* Update general.py

* Update general.py

* Increase NMS fixed time limit to 200ms

* Update general.py

* Apple Metal Performance Shader (MPS) device support

Following https://pytorch.org/blog/introducing-accelerated-pytorch-training-on-mac/

Should work with Apple M1 devices with PyTorch nightly installed with command `--device mps`. Usage examples:
```bash
python train.py --device mps
python detect.py --device mps
python val.py --device mps
```

* Update device strategy to fix MPS issue

* Bug Fixed: OpenVINO metadata

* Bug Fixed: OpenVINO metadata

* Update export.py

* Update export.py
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

* Rename OpenVINO meta.yaml to model name

* Rename OpenVINO meta.yaml to model name

* Rename OpenVINO meta.yaml to model name

* fix

* Write .yaml file when exporting model to openvino

Write .yaml file automatically when exporting model to openvino to be used during inference

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update export.py

* Update export.py

* Load metadata on inference

* Update common.py
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

Resolves Ultralytics HUB CI errors.

Adjust to 5%, fixes failing Colab AMP check with V100 (1.5% different) with 200% safety margin.

Resolves https://github.com/ultralytics/yolov5/discussions/7931

* Code refactor for general.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update restapi.py
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Add PyTorch AMP check

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Cleanup

* Cleanup

* Cleanup

* Robust for DDP

* Fixes

* Add amp enabled boolean to check_train_batch_size

* Simplify

* space to prefix
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Resolves https://github.com/ultralytics/yolov5/issues/7911#issuecomment-1133671255

Locate file in subdirectories.

* Add DWConvTranspose2d() module

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Add DWConvTranspose2d() module

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix

* Fix
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Fix for https://github.com/ultralytics/yolov5/discussions/7901

* Improve mAP0.5-0.95

Two changes provided
1. Added limit on the maximum number of detections for each image likewise pycocotools
2. Rework process_batch function

Changes #2 solved issue #4251
I also independently encountered the problem described in issue #4251 that the values for the same thresholds do not match when changing the limits in the torch.linspace function.
These changes solve this problem.

Currently during validation yolov5x.pt model the following results were obtained:
from yolov5 validation
               Class     Images     Labels          P          R     mAP@.5 mAP@.5:.95: 100%|██████████| 157/157 [01:07<00:00,  2.33it/s]
                 all       5000      36335      0.743      0.626      0.682      0.506
from pycocotools
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.505
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.685

These results are very close, although not completely pass the competition issue #2258.
I think it's problem with false positive bboxes matched ignored criteria, but this is not actual for custom datasets and does not require an additional solution.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Remove line to retain pycocotools results

* Update val.py

* Update val.py

* Remove to device op

* Higher precision int conversion

* Update val.py
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Added Windows cmd to count GPU devices

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