Lightning

- validation_step, val_dataloader are now optional.
- enabled multiple dataloaders for validation.
- support for latest test-tube logger optimized for PT 1.2.0.
- lr_scheduler now activated after epoch

0.4.0 is the first public release after a short period testing with public users. Thanks for all the help ironing out bugs to get Lightning to run on everything from notebooks to local to server machines.

This release includes:
- Extensively tested code.
- Cleaner API to accommodate the various research use cases

New features
- No need for experiment object in trainer.
- Training continuation (not just weights, but also epoch, global step, etc...)
- if the folder the checkpoint callback uses has weights, it loads the last weights automatically.
- training step and validation step don't reduce outputs automatically anymore. This fixes issues with reducing generated outputs for example (images, text).
- 16-bit can now be used with a single GPU (no DP or DDP in this case). bypasses issue with NVIDIA apex and PT compatibility for DP+16-bit training.

Simplified data loader.

Added a decorator to do lazy loading internally:

Old:
python
property
def tng_dataloader(self):
if self._tng_dataloader is None:
self._tng_dataloader = DataLoader(...)
return self.tng_dataloder


Now:

python
ptl.data_loader
def tng_dataloader(self):
return DataLoader(...)
`

Fully tested!

Includes:
- Code coverage (99%)
- Full tests that run multiple models in different configs
- Full tests that test specific functionality in trainer.

* fix: don't start zmq when fast_queue=false by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/417
* fix release ci by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/416

New Contributors
* rittik9 made their first contribution in https://github.com/Lightning-AI/LitServe/pull/375
* ali-alshaar7 made their first contribution in https://github.com/Lightning-AI/LitServe/pull/380

**Full Changelog**: https://github.com/Lightning-AI/LitServe/compare/v0.2.5...v0.2.6

What's Changed
* Set LitServer.stream using LitSpec.stream by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/398
* fix openai usage info for non-streaming response by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/399
* [pre-commit.ci] pre-commit suggestions by pre-commit-ci in https://github.com/Lightning-AI/LitServe/pull/400
* Async continuous batching loop by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/401
* add validation for `stream=False` with `yield` usage by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/402
* fix callback runner to execute after predict by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/406
* integrate zmq by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/403
* warn users when predict/unbatch output length is not same as requests by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/408
* move built in loops inside classes by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/409
* add justus and thomas as codeowners by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/410
* enable multiple workers for ZMQ by aniketmaurya in https://github.com/Lightning-AI/LitServe/pull/411


**Full Changelog**: https://github.com/Lightning-AI/LitServe/compare/v0.2.6.dev2...v0.2.6.dev3