Doc is needed on how to better customize the sub Loop

[tshu-w]

I tried something like

            epoch_loop = self.trainer.fit_loop.epoch_loop
            epoch_loop.connect(val_loop=MyEvaluationLoop(verbose=False))
            self.trainer.fit_loop.connect(epoch_loop=epoch_loop)

but failed with:

  File "/home1/wangtianshu/universal-blocker/run", line 35, in <module>
    main()
  File "/home1/wangtianshu/universal-blocker/run", line 22, in main
    cli = LitCLI(
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/utilities/cli.py", line 564, in __init__
    self._run_subcommand(self.subcommand)
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/utilities/cli.py", line 835, in _run_subcommand
    fn(**fn_kwargs)
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 771, in fit
    self._call_and_handle_interrupt(
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 724, in _call_and_handle_interrupt
    return trainer_fn(*args, **kwargs)
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 812, in _fit_impl
    results = self._run(model, ckpt_path=self.ckpt_path)
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 1237, in _run
    results = self._run_stage()
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 1324, in _run_stage
    return self._run_train()
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 1346, in _run_train
    self._run_sanity_check()
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 1407, in _run_sanity_check
    val_loop._reload_evaluation_dataloaders()
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/loops/dataloader/evaluation_loop.py", line 236, in _reload_evaluation_dataloaders
    if self.trainer.testing:
  File "/home1/wangtianshu/.local/share/conda/envs/blocking/lib/python3.9/site-packages/pytorch_lightning/loops/base.py", line 58, in trainer
    raise RuntimeError("The loop is not attached to a Trainer.")
RuntimeError: The loop is not attached to a Trainer.

It works with following setup, but I think detailed doc will help other user:

            fit_loop = self.trainer.fit_loop
            epoch_loop = fit_loop.epoch_loop
            epoch_loop.connect(val_loop=MyEvaluationLoop(verbose=False))
            fit_loop.connect(epoch_loop=epoch_loop)
            self.trainer.fit_loop = fit_loop