Skip to content

lcbench

class LCBenchConfig
dataclass #

Bases: YAHPOConfig

A LCBench Config.

Note:#

For momentum, the paper seems to suggest it's (0.1, 0.9) but the configspace says (0.1, 0.99), going with the code version.

def validate() #

Validate this is a correct config.

Source code in src/mfpbench/yahpo/benchmarks/lcbench.py 
28
29
30
31
32
33
34
35
36
def validate(self) -> None:
    """Validate this is a correct config."""
    assert 16 <= self.batch_size <= 512
    assert 1e-04 <= self.learning_rate <= 0.1
    assert 0.1 <= self.momentum <= 0.99
    assert 1e-05 <= self.weight_decay <= 0.1
    assert 1 <= self.num_layers <= 5
    assert 64 <= self.max_units <= 1024
    assert 0.0 <= self.max_dropout <= 1.0

class LCBenchResult
dataclass #

Bases: YAHPOResult[LCBenchConfig, int]

score: float
prop #

The score of interest.

error: float
prop #

The error of interest.

test_score: float
prop #

The score on the test set.

test_error: float
prop #

The score on the test set.

val_score: float
prop #

The score on the validation set.

val_error: float
prop #

The score on the validation set.

cost: float
prop #

Time taken in seconds to train the config (assumed to be seconds).

class LCBenchBenchmark #

Bases: YAHPOBenchmark

yahpo_instances
classvar attr #

('3945', '7593', '34539', '126025', '126026', '126029', '146212', '167104', '167149', '167152', '167161', '167168', '167181', '167184', '167185', '167190', '167200', '167201', '168329', '168330', '168331', '168335', '168868', '168908', '168910', '189354', '189862', '189865', '189866', '189873', '189905', '189906', '189908', '189909')