* WIP

* Fixed web server

* Got the basics woprking

* Fixed some logic, trapped some error conditions

Also changed "SERVO_ESC" drivetrain type, which doesn't really mean anything, to "I2C_SERVO", which actually describes the board used

* removed RC class stub

* Basic port done

* cleanup

* Fixed mode switch

* Fixed joystick import issue

* Changed option name to "PIGPIO_RC" which is more clear

* changed rc controller option config to "pigpio_rc"

* typo

* Fixed comment to align with nomenclature

* Fixed auto-record on throttle

Did a little other cleanup

* Cosmetic cleanup per review

* Fixed auto-record so it is only on when desired

Did some other cleanup of unneeded constants and added a debug option

* Changes per review

* WIP

Fixed web server

Got the basics woprking

Fixed some logic, trapped some error conditions

Also changed "SERVO_ESC" drivetrain type, which doesn't really mean anything, to "I2C_SERVO", which actually describes the board used

removed RC class stub

Basic port done

cleanup

Fixed mode switch

Fixed joystick import issue

Changed option name to "PIGPIO_RC" which is more clear

changed rc controller option config to "pigpio_rc"

typo

Fixed comment to align with nomenclature

Fixed auto-record on throttle

Did a little other cleanup

Cosmetic cleanup per review

Fixed auto-record so it is only on when desired

Did some other cleanup of unneeded constants and added a debug option

Changes per review

* Addressed code review

Co-authored-by: Chris Anderson <gca2@earthlink.net>

Revert to passing full path to model in training call which got accidentally broken in 4.2 master. (#873)

* Fix delete_last_n_records in tub.
* Pep-8: remove whitespace
* Add test.

* Update to CircuitPython OLED library

* Per code review

* Added the OLED library to Nano installation, too

Co-authored-by: Chris Anderson <gca2@earthlink.net>

Co-authored-by: Chris Anderson <gca2@earthlink.net>

As suggested by @DocGarbanzo. It works!

* - Update tensorflow version to 1.15.0
- Add doc for Jetpack 4.4
- Remove opencv installation

* Add dataset and pre-trained model page

* fix issue #634 (#646)

- remove jquery dependency
- add clearTimeout

* Allow overriding WEB_CONTROL_PORT from env variable

* pin h5py version

* fix h5py version for mac

* * Update doc with donkey train command. (#709)

* Update doc with developer section for building own models in donkey 4.1
* Integrate changes from PR feedback

* Donkey 4.x is now official (#723)

* Update nano install docs - python3-opencv vs libopencv-python

* change slack to discord on readme

* Donkeycar 4.x release.  (#644)

## Major Improvements

- New Datastore.
- More ways to pre-process image data for training. 
- Use the 2.x version of Tensorflow. 
- Lots of other minor improvements.

* Improvements to the car app and handling of KerasPilot parts: (#648)

* Improvements to the car app and handling of KerasPilot parts:
*...

It was redundant to the DRIVE_TRAIN_TYPE and  CONTROLLER_TYPE options

* added virtual lidar support
* removed useless line

Small updates for RC Reveiver and Lidar

Combined lidar tests into single file and marked as skip, because it requires the necessary hardwar
Updated docs for RC receiver

Donkey UI: Add disclaimer and documentation for car connector screen

Kivy UI fixes

Change tub select label against toggle button to allow not creating a new directory, i.e. call rsync pi@donkeypi:~/mycar/data/ instead of rsync pi@donkeypi:~/mycar/data. The latter command will create a new data/ folder within the target data/ folder which we don't want.

Fix ui issues: check for existing config entries in ssh poll and correct path to start screen image.

Added support for Odometry

Support running webcontroller and physical joystick at the same time

Also: extensions to training supporting comments, auto-naming, transfer models. Small improvement in deleting/restoring tub records in bulk.

(cherry picked from commit 6159015ac57713c7be24028275acb0938cd436ca)

Update workflow after rename of ubuntu conda file.

(cherry picked from commit 9768aca4d04e8296f528a0f435397f7eb29a8ef8)

* Take out Tensorflow from automated windows installation because it's causing …  …an error with latest conda
* Changed pc.yml to windows.yml
* Rename pc.yml to windows.yml
* Create ubuntu.yml

(cherry picked from commit 8d5740d5d3622f21dbd633faaf05438a091bdee0)

Avoids ambiguity between computer type and OS type (PCs can run both Windows and Ubuntu)

(cherry picked from commit 8ac25295895ca34e3351d284d9fc61c376d3bacd)

Conda says not to put "--user" in yml scripts so we'll have the users do this manually until we get to the bottom of the TF 2.2 installation errors on Windows

(cherry picked from commit 81732ab8cde9567b2d6c59e4bd83fad79f2d4c9d)

* Predictions are now scaled between [0, 1] for training and validation

Previously the predictions were between [-1, 1] for velocity and
steering, but PyTorch Lightning v1.2.4 introduced changes that
prevented this from working.

Now scaling to be between [0, 1] for training and inference and
then unscaling the predictions for running.

* Replaced accuracy and precision metrics with mean squared error

It didn't make sense to use accuracy and precision when predicted
values could take on a continious range of values. MSE is a better
metric.

PyTorch Lightning also would throw an error when trying to use
accuracy or precision with non-integer tensors.

I'll add it into the documentation to manually install with the --user flag after the conda evn is built