GitLab

Co-authored-by: Lijiao <Lijiaoa@outlook.com>

* Fix PyTorchImageClassificationTrainer's training

It seems that the current process only calculates the loss and gradient, and does not use the optimizer for optimization. Therefore, the model is not actually trained, and its accuracy on the Web UI remains unchanged.

* Add intermediate reports for every epoch

For now, intermediate reports and final reports are consistent, and they are displayed once after all epochs have finished. This may not meet our expectations. We hope that intermediate reports can provide us with the validation results after each epoch.

* Fix import path

* Fix conflicted version

* fix outputs and python process not killed

* change cleanup stop environment logic

* fix bug

* use same default metric component

* review self

* rename boolean type variable

* del whichGraph field

* rename my_theme to nni_theme
Co-authored-by: Lijiao <Lijiaoa@outlook.com>

* Refactor fast-test

* Fix yaml syntax error

* Fix syntax issues

* Update requirements

* Update requirements

* Fix pylint

* Fix pylint

* Fix pylint

* Cache pip packages for lint

* Add wheel as dependency

* Isolate setup dependencies

* Isolate docs pipeline

* Fix swig issue

* refactor util config

* merge NNICTL_HOME_DIR and NNI_HOME_DIR

* add default logDir and fix Config logdir

* fix save/load

* modify updater and tensorboard

* rename

* rename

* fix local clusterMetaData

* fix local config convert

* fix ut

* fix ut

* del unused code

* fix cursor

Support show customized tuner/assessor name and get optimize_mode from tuner/advisor/assessor (#3317)