[Logistic Regression] Support fit on two classes #343
Conversation
|
build |
| ) -> Callable[[FitInputType, Dict[str, Any]], Dict[str, Any],]: | ||
| num_workers = self.num_workers | ||
| array_order = self._fit_array_order() | ||
| num_classes = dataset.select(alias.label).distinct().count() |
There was a problem hiding this comment.
seems num_classes's not been used.
|
|
||
| # filter only supported params | ||
| init_parameters = { | ||
| k: v for k, v in init_parameters.items() if k in supported_params |
There was a problem hiding this comment.
Seems supported_params is always an empty list?
There was a problem hiding this comment.
Yeah, working on C++/Cython support: rapidsai/cuml#5516
Will address this in the next PR.
There was a problem hiding this comment.
Will you be integrating the init params in this PR, now that the cuml init params PR is merged? It would help flesh out the param mapping and the compat tests.
There was a problem hiding this comment.
Yeah. not sure actually.
The init PR will update a certain amount of the codes in multiple places of this PR. May make this PR too long and introduce extra reviewing overhead.
Thinking of getting this one merged. Then I will create a new PR for transform and init and update all the tests in test_logistic_regression.py.
| @@ -0,0 +1,111 @@ | |||
| from typing import Any, Dict, Tuple | |||
There was a problem hiding this comment.
Would be good to have a 'compat' test similar to other classes with branches for non-supported apis.
There was a problem hiding this comment.
'compat' test requires vector input type. Currently spark rapids ml converts vector input type into float64 that has no support in cuml C++/Cython 23.08 yet.
Do we want to work out a workaround?
|
build |
| feature_type, | ||
| ) = super()._pre_process_data(dataset) | ||
|
|
||
| # if input format is vectorUDT, convert data type from float32 |
Signed-off-by: Jinfeng <[email protected]>
Signed-off-by: Jinfeng <[email protected]>
Signed-off-by: Jinfeng <[email protected]>
Signed-off-by: Jinfeng <[email protected]>
Signed-off-by: Jinfeng <[email protected]>
Signed-off-by: Jinfeng <[email protected]>
|
build |
Require installing cuml 23.08 nightly to run the tests.