tests for maml and opti #38
Conversation
There was a problem hiding this comment.
Hey @Riddhi2587, I am sorry for poking in the draft pr.
I have added some comments; please go through them. I will do another review after these comments are addressed.
If you have made these changes in the local repository and haven't updated the remote branch, ignore these changes.
| # self.assertIsInstance(params,) | ||
| self.assertIsInstance(apply_fn, Callable) | ||
| self.assertIsInstance(adapt_fn, Callable) | ||
| self.assertIsInstance(loss_fn, Callable) | ||
| state = Opti.create(params, apply_fn, adapt_fn, loss_fn, tx) | ||
| self.assertIsInstance(state, MetaTrainState) |
There was a problem hiding this comment.
What you are doing here is very thoughtful and is in the right direction. But the problem with this approach is as the repo gets bigger, writing these assert statements will be very painful. So how about automating all this?
There exists a tool for doing this called mypy. Mypy is a static type checker tool , and as we are including type hints, it makes sense to add mypy tool in repo.
There was a problem hiding this comment.
I can set up mypy for the repository which will do type-checking tasks on the library and on the test modules.
looping in @veds12
| self.assertIsInstance(state, MetaTrainState) | ||
| self.assertIsInstance(metrics, List[Callable[[ndarray, ndarray], ndarray]]) | ||
| meta_train = Opti.meta_train_step(state, tasks, metrics) | ||
| self.assertIsInstance(meta_train, Tuple[MetaTrainState, ndarray, List[ndarray]]) |
|
|
||
| sys.path.append('C:users/riddh/git-repos/jeta/jeta') | ||
| # from opti_trainer import * | ||
| import opti_trainer as opt |
There was a problem hiding this comment.
clever hack! but this will only work on your local system.
init.py is empty, this is why you are not able to import opti_trainer and other functions/ class here.
You can only import opti_trainer if the above init file has an import statement for opti_trainer.
So by adding following line in init.py,
from .opti_trainer import OptiTrainerthere will be no need to write this: sys.path.append('C:users/riddh/git-repos/jeta/jeta') .
then to import the OptiTrainer class in the tests module, use
from jeta import OptiTrainer|
|
||
| sys.path.append('C:users/riddh/git-repos/jeta/jeta') | ||
| import maml |
|
@Riddhi2587 Any updates? |
No description provided.