Maximum resources #3142
Maximum resources #3142
Conversation
|
Job PR-3142-7a800f8 is done. |
tabular/src/autogluon/tabular/models/tabular_nn/torch/tabular_nn_torch.py
Outdated
Show resolved
Hide resolved
tabular/src/autogluon/tabular/models/fastainn/tabular_nn_fastai.py
Outdated
Show resolved
Hide resolved
tabular/src/autogluon/tabular/models/tabular_nn/torch/tabular_nn_torch.py
Outdated
Show resolved
Hide resolved
| total_resources: Optional[Dict[str, Union[int, float]]] = None, | ||
| parallel_hpo: bool = False, | ||
| **kwargs | ||
| ): |
There was a problem hiding this comment.
add docstring explaining this, add return type
| return kwargs | ||
|
|
||
| def _preprocess_fit_resources(self, silent=False, total_resources=None, parallel_hpo=False, **kwargs): |
| return kwargs | ||
|
|
||
| def _preprocess_fit_resources(self, silent=False, total_resources=None, parallel_hpo=False, **kwargs): |
|
Job PR-3142-bd6372d is done. |
|
Merging as unit tests for previous commits have passed, the most recent commit only added comments |
|
Job PR-3142-f76a039 is done. |
|
Job PR-3142-178cf94 is done. |
|
Job PR-3142-d22642f is done. |
Issue #, if available:
torch models training will be slowed by the usage of virtual cores.
Description of changes:
Example run output with a newly launched cluster of 8 m5.24xlarge machine:
The training time matches a local run and appear to be normal now
Fitting 1 L1 models ... Fitting model: NeuralNetFastAI_BAG_L1 ... Fitting 8 child models (S1F1 - S1F8) | Fitting with ParallelDistributedFoldFittingStrategy 0.9224 = Validation score (accuracy) 286.06s = Training runtime 3.44s = Validation runtime Fitting model: WeightedEnsemble_L2 ... 0.9224 = Validation score (accuracy) 0.08s = Training runtime 0.04s = Validation runtime AutoGluon training complete, total runtime = 298.45s ... Best model: "WeightedEnsemble_L2"By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.