Raise error if max duration is in epochs and dataloader is infinite

[dakinggg]

What does this PR do?

Raises an error if the train dataloader is infinite and max duration is specified in epochs.

UX:

In [11]: trainer.fit()
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[11], line 1
----> 1 trainer.fit()

File ~/github/composer/composer/trainer/trainer.py:1693, in Trainer.fit(self, train_dataloader, train_dataloader_label, train_subset_num_batches, duration, reset_time, schedulers, scale_schedule_ratio, step_schedulers_every_batch, eval_dataloader, eval_subset_num_batches, eval_interval, grad_accum, device_train_microbatch_size, precision)
   1690     _raise_missing_argument_exception('max_duration')
   1692 if self.state.dataloader_len is None and self.state.max_duration.unit == TimeUnit.EPOCH:
-> 1693     raise ValueError(
   1694         ('max_duration cannot be specified in epochs when using an infinite dataloader. Please either '
   1695          'provide a dataloader with a length, specify max_duration in batches, samples, or tokens, or provide '
   1696          'train_subset_num_batches.'))
   1698 if self.state.max_duration <= self.state.timestamp.get(self.state.max_duration.unit) and not reset_time:
   1699     raise ValueError(
   1700         (f'The max_duration ({self.state.max_duration}) is less than or equal to the elapsed training duration '
   1701          f'({self.state.timestamp.get(self.state.max_duration.unit)}). No training would occur. '
   1702          'Please provide the `duration` or specify `reset_time=True` in Trainer.fit().'))

ValueError: max_duration cannot be specified in epochs when using an infinite dataloader. Please either provide a dataloader with a length, specify max_duration in batches, samples, or tokens, or provide train_subset_num_batches.

What issue(s) does this change relate to?

Closes CO-1738

Before submitting

• Have you read the contributor guidelines?
• Was this change discussed/approved in a GitHub issue first? It is much more likely to be merged if so.
• Did you update any related docs and document your change?
• Did you update any related tests and add any new tests related to your change? (see testing)
• Did you run the tests locally to make sure they pass?
• Did you run pre-commit on your change? (see the pre-commit section of prerequisites)

[mvpatel2000]

Lgtm. Maybe extend test to ensure infinite dataloader works in other cases described in error msg by parameterizing test

[eracah]

LGTM

[dakinggg]

Good call, added some more tests.