Update DeepSpeed version requirement in Dockerfile

[awaelchli]

What does this PR do?

Fixes PR merge blocking on master.

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is free to review the PR once the tests have passed.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

Make sure you had fun coding 🙃

[pep8speaks]

Hello @awaelchli! Thanks for opening this PR.

• In the file pl_examples/repro_vae.py:

Line 206:36: W292 no newline at end of file

Do see the Hitchhiker's guide to code style

[codecov]

Codecov Report

Merging #7326 (e857eeb) into master (e0c64f0) will increase coverage by 3%.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master   #7326    +/-   ##
=======================================
+ Coverage      87%     90%    +3%     
=======================================
  Files         200     200            
  Lines       12865   13051   +186     
=======================================
+ Hits        11210   11739   +529     
+ Misses       1655    1312   -343     

[leezu]

@awaelchli is there a tracking issue or other resource detailing the problems with deepspeed>0.3.14?

[awaelchli]

Yes, a failing test after Deepspeed updated to a new version
https://dev.azure.com/PytorchLightning/pytorch-lightning/_build/results?buildId=7176&view=logs&j=3afc50db-e620-5b81-6016-870a6976ad29&t=d9f671c5-a304-5675-5394-961fd7f98b9b

_______________________ test_deepspeed_multigpu_stage_3 ________________________

tmpdir = local('/tmp/pytest-of-AzDevOps_azpcontainer/pytest-7/test_deepspeed_multigpu_stage_0')
deepspeed_config = {'optimizer': {'params': {'lr': 3e-05}, 'type': 'SGD'}, 'scheduler': {'params': {'last_batch_iteration': -1, 'warmup_max_lr': 3e-05, 'warmup_min_lr': 0, 'warmup_num_steps': 100}, 'type': 'WarmupLR'}}

    @RunIf(min_gpus=2, deepspeed=True, special=True)
    def test_deepspeed_multigpu_stage_3(tmpdir, deepspeed_config):
        """
        Test to ensure ZeRO Stage 3 works with a parallel model.
        """
        model = ModelParallelBoringModel()
        trainer = Trainer(
            plugins=[DeepSpeedPlugin(stage=3)],
            default_root_dir=tmpdir,
            gpus=2,
            fast_dev_run=True,
            precision=16,
        )
>       trainer.fit(model)

tests/plugins/test_deepspeed_plugin.py:459: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
pytorch_lightning/trainer/trainer.py:869: in fit
    self._run(model)
pytorch_lightning/trainer/trainer.py:473: in _run
    self.pre_dispatch()
pytorch_lightning/trainer/trainer.py:498: in pre_dispatch
    self.accelerator.pre_dispatch(self)
pytorch_lightning/accelerators/accelerator.py:108: in pre_dispatch
    self.training_type_plugin.pre_dispatch()
pytorch_lightning/plugins/training_type/deepspeed.py:241: in pre_dispatch
    self.init_deepspeed()
pytorch_lightning/plugins/training_type/deepspeed.py:258: in init_deepspeed
    self._initialize_deepspeed_train(model)
pytorch_lightning/plugins/training_type/deepspeed.py:287: in _initialize_deepspeed_train
    model, optimizer, _, lr_scheduler = deepspeed.initialize(
/usr/local/lib/python3.8/dist-packages/deepspeed/__init__.py:120: in initialize
    engine = DeepSpeedEngine(args=args,
/usr/local/lib/python3.8/dist-packages/deepspeed/runtime/engine.py:149: in __init__
    self._configure_distributed_model(model)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = DeepSpeedEngine(
  (module): LightningDeepSpeedModule(
    (module): ModelParallelBoringModel(
      (layer): Linear(in_features=32, out_features=2, bias=True)
      (linear): Linear(in_features=32, out_features=2, bias=True)
    )
  )
)
model = LightningDeepSpeedModule(
  (module): ModelParallelBoringModel(
    (layer): Linear(in_features=32, out_features=2, bias=True)
    (linear): Linear(in_features=32, out_features=2, bias=True)
  )
)

    def _configure_distributed_model(self, model):
        self.module = model
        if self.fp16_enabled():
            if self.zero_optimization_partition_weights() and any(
                [hasattr(param,
                         'ds_id') for param in self.module.parameters()]):
>               assert all([param.dtype == torch.half for param in self.module.parameters()]), f"Model must initialized in fp16 mode for ZeRO Stage 3."
E               AssertionError: Model must initialized in fp16 mode for ZeRO Stage 3.

/usr/local/lib/python3.8/dist-packages/deepspeed/runtime/engine.py:569: AssertionError