Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly
Sign up
Appearance settings

Commit 56a2ac1

Browse files
ravinkohlidengdifan
andauthored
[RELEASE] v0.2.1 (#475)
* [FIX] Documentation and docker workflow file (#449) * fixes to documentation and docker * fix to docker * Apply suggestions from code review * add change log for release (#450) * [FIX] release docs (#452) * Release 0.2 * Release 0.2.0 * fix docs new line * [FIX] ADD forecasting init design to pip data files (#459) * add forecasting_init.json to data files under setup * avoid undefined reference in scale_value * checks for time series dataset split (#464) * checks for time series dataset split * maint * Update autoPyTorch/datasets/time_series_dataset.py Co-authored-by: Ravin Kohli <13005107+ravinkohli@users.noreply.github.com> Co-authored-by: Ravin Kohli <13005107+ravinkohli@users.noreply.github.com> * [FIX] Numerical stability scaling for timeseries forecasting tasks (#467) * resolve rebase conflict * add checks for scaling factors * flake8 fix * resolve conflict * [FIX] pipeline options in `fit_pipeline` (#466) * fix update of pipeline config options in fit pipeline * fix flake and test * suggestions from review * [FIX] results management and visualisation with missing test data (#465) * add flexibility to avoid checking for test scores * fix flake and test * fix bug in tests * suggestions from review * [ADD] Robustly refit models in final ensemble in parallel (#471) * add parallel model runner and update running traditional classifiers * update pipeline config to pipeline options * working refit function * fix mypy and flake * suggestions from review * fix mypy and flake * suggestions from review * finish documentation * fix tests * add test for parallel model runner * fix flake * fix tests * fix traditional prediction for refit * suggestions from review * add warning for failed processing of results * remove unnecessary change * update autopytorch version number * update autopytorch version number and the example file * [DOCS] Release notes v0.2.1 (#476) * Release 0.2.1 * add release docs * Update docs/releases.rst Co-authored-by: Difan Deng <33290713+dengdifan@users.noreply.github.com>
1 parent c138dff commit 56a2ac1

37 files changed

+2814
-306
lines changed

‎autoPyTorch/__version__.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
11
"""Version information."""
22

33
# The following line *must* be the last in the module, exactly as formatted:
4-
__version__ = "0.2"
4+
__version__ = "0.2.1"

‎autoPyTorch/api/base_task.py

Lines changed: 263 additions & 158 deletions
Large diffs are not rendered by default.

‎autoPyTorch/constants.py

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -58,6 +58,10 @@
5858
"forecasting tasks! Please run \n pip install autoPyTorch[forecasting] \n to "\
5959
"install the corresponding dependencies!"
6060

61+
# This value is applied to ensure numerical stability: Sometimes we want to rescale some values: value / scale.
62+
# We make the scale value to be 1 if it is smaller than this value to ensure that the scaled value will not resutl in
63+
# overflow
64+
VERY_SMALL_VALUE = 1e-12
6165

6266
# The constant values for time series forecasting comes from
6367
# https://github.com/rakshitha123/TSForecasting/blob/master/experiments/deep_learning_experiments.py
@@ -78,3 +82,6 @@
7882

7983
# To avoid that we get a sequence that is too long to be fed to a network
8084
MAX_WINDOW_SIZE_BASE = 500
85+
86+
# AutoPyTorch optionally allows network inference or metrics calculation for the following datasets
87+
OPTIONAL_INFERENCE_CHOICES = ('test',)

‎autoPyTorch/datasets/time_series_dataset.py

Lines changed: 11 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -693,6 +693,17 @@ def __init__(self,
693693

694694
self.splits = self.get_splits_from_resampling_strategy() # type: ignore[assignment]
695695

696+
valid_splits = []
697+
for i, split in enumerate(self.splits):
698+
if len(split[0]) > 0:
699+
valid_splits.append(split)
700+
701+
if len(valid_splits) == 0:
702+
raise ValueError(f'The passed value for {n_prediction_steps} is unsuited for the current dataset, please '
703+
'consider reducing n_prediction_steps')
704+
705+
self.splits = valid_splits
706+
696707
# TODO doing experiments to give the most proper way of defining these two values
697708
if lagged_value is None:
698709
try:

‎autoPyTorch/ensemble/abstract_ensemble.py

Lines changed: 12 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -9,6 +9,9 @@
99
class AbstractEnsemble(object):
1010
__metaclass__ = ABCMeta
1111

12+
def __init__(self):
13+
self.identifiers_: List[Tuple[int, int, float]] = []
14+
1215
@abstractmethod
1316
def fit(
1417
self,
@@ -76,3 +79,12 @@ def get_validation_performance(self) -> float:
7679
Returns:
7780
Score
7881
"""
82+
83+
def update_identifiers(
84+
self,
85+
replace_identifiers_mapping: Dict[Tuple[int, int, float], Tuple[int, int, float]]
86+
) -> None:
87+
identifiers = self.identifiers_.copy()
88+
for i, identifier in enumerate(self.identifiers_):
89+
identifiers[i] = replace_identifiers_mapping.get(identifier, identifier)
90+
self.identifiers_ = identifiers

‎autoPyTorch/evaluation/abstract_evaluator.py

Lines changed: 17 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -195,7 +195,8 @@ def get_additional_run_info(self) -> Dict[str, Any]:
195195
Can be found in autoPyTorch/pipeline/components/setup/traditional_ml/estimator_configs
196196
"""
197197
return {'pipeline_configuration': self.configuration,
198-
'trainer_configuration': self.pipeline.named_steps['model_trainer'].choice.model.get_config()}
198+
'trainer_configuration': self.pipeline.named_steps['model_trainer'].choice.model.get_config(),
199+
'configuration_origin': 'traditional'}
199200

200201
def get_pipeline_representation(self) -> Dict[str, str]:
201202
return self.pipeline.get_pipeline_representation()
@@ -347,7 +348,7 @@ class AbstractEvaluator(object):
347348
348349
An evaluator is an object that:
349350
+ constructs a pipeline (i.e. a classification or regression estimator) for a given
350-
pipeline_config and run settings (budget, seed)
351+
pipeline_options and run settings (budget, seed)
351352
+ Fits and trains this pipeline (TrainEvaluator) or tests a given
352353
configuration (TestEvaluator)
353354
@@ -369,7 +370,7 @@ class AbstractEvaluator(object):
369370
The amount of epochs/time a configuration is allowed to run.
370371
budget_type (str):
371372
The budget type. Currently, only epoch and time are allowed.
372-
pipeline_config (Optional[Dict[str, Any]]):
373+
pipeline_options (Optional[Dict[str, Any]]):
373374
Defines the content of the pipeline being evaluated. For example, it
374375
contains pipeline specific settings like logging name, or whether or not
375376
to use tensorboard.
@@ -430,7 +431,7 @@ def __init__(self, backend: Backend,
430431
budget: float,
431432
configuration: Union[int, str, Configuration],
432433
budget_type: str = None,
433-
pipeline_config: Optional[Dict[str, Any]] = None,
434+
pipeline_options: Optional[Dict[str, Any]] = None,
434435
seed: int = 1,
435436
output_y_hat_optimization: bool = True,
436437
num_run: Optional[int] = None,
@@ -523,10 +524,10 @@ def __init__(self, backend: Backend,
523524
self._init_params = init_params
524525

525526
assert self.pipeline_class is not None, "Could not infer pipeline class"
526-
pipeline_config = pipeline_config if pipeline_config is not None \
527+
pipeline_options = pipeline_options if pipeline_options is not None \
527528
else self.pipeline_class.get_default_pipeline_options()
528-
self.budget_type = pipeline_config['budget_type'] if budget_type is None else budget_type
529-
self.budget = pipeline_config[self.budget_type] if budget == 0 else budget
529+
self.budget_type = pipeline_options['budget_type'] if budget_type is None else budget_type
530+
self.budget = pipeline_options[self.budget_type] if budget == 0 else budget
530531

531532
self.num_run = 0 if num_run is None else num_run
532533

@@ -539,7 +540,7 @@ def __init__(self, backend: Backend,
539540
port=logger_port,
540541
)
541542

542-
self._init_fit_dictionary(logger_port=logger_port, pipeline_config=pipeline_config, metrics_dict=metrics_dict)
543+
self._init_fit_dictionary(logger_port=logger_port, pipeline_options=pipeline_options, metrics_dict=metrics_dict)
543544
self.Y_optimization: Optional[np.ndarray] = None
544545
self.Y_actual_train: Optional[np.ndarray] = None
545546
self.pipelines: Optional[List[BaseEstimator]] = None
@@ -597,7 +598,7 @@ def _init_datamanager_info(
597598
def _init_fit_dictionary(
598599
self,
599600
logger_port: int,
600-
pipeline_config: Dict[str, Any],
601+
pipeline_options: Dict[str, Any],
601602
metrics_dict: Optional[Dict[str, List[str]]] = None,
602603
) -> None:
603604
"""
@@ -608,7 +609,7 @@ def _init_fit_dictionary(
608609
Logging is performed using a socket-server scheme to be robust against many
609610
parallel entities that want to write to the same file. This integer states the
610611
socket port for the communication channel.
611-
pipeline_config (Dict[str, Any]):
612+
pipeline_options (Dict[str, Any]):
612613
Defines the content of the pipeline being evaluated. For example, it
613614
contains pipeline specific settings like logging name, or whether or not
614615
to use tensorboard.
@@ -634,7 +635,7 @@ def _init_fit_dictionary(
634635
'optimize_metric': self.metric.name
635636
})
636637

637-
self.fit_dictionary.update(pipeline_config)
638+
self.fit_dictionary.update(pipeline_options)
638639
# If the budget is epochs, we want to limit that in the fit dictionary
639640
if self.budget_type == 'epochs':
640641
self.fit_dictionary['epochs'] = self.budget
@@ -805,6 +806,11 @@ def finish_up(self, loss: Dict[str, float], train_loss: Dict[str, float],
805806
if test_loss is not None:
806807
additional_run_info['test_loss'] = test_loss
807808

809+
# Add information to additional info that can be useful for other functionalities
810+
additional_run_info['configuration'] = self.configuration \
811+
if not isinstance(self.configuration, Configuration) else self.configuration.get_dictionary()
812+
additional_run_info['budget'] = self.budget
813+
808814
rval_dict = {'loss': cost,
809815
'additional_run_info': additional_run_info,
810816
'status': status}

‎autoPyTorch/evaluation/tae.py

Lines changed: 9 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -123,7 +123,7 @@ def __init__(
123123
abort_on_first_run_crash: bool,
124124
pynisher_context: str,
125125
multi_objectives: List[str],
126-
pipeline_config: Optional[Dict[str, Any]] = None,
126+
pipeline_options: Optional[Dict[str, Any]] = None,
127127
initial_num_run: int = 1,
128128
stats: Optional[Stats] = None,
129129
run_obj: str = 'quality',
@@ -198,13 +198,13 @@ def __init__(
198198
self.disable_file_output = disable_file_output
199199
self.init_params = init_params
200200

201-
self.budget_type = pipeline_config['budget_type'] if pipeline_config is not None else budget_type
201+
self.budget_type = pipeline_options['budget_type'] if pipeline_options is not None else budget_type
202202

203-
self.pipeline_config: Dict[str, Union[int, str, float]] = dict()
204-
if pipeline_config is None:
205-
pipeline_config = replace_string_bool_to_bool(json.load(open(
203+
self.pipeline_options: Dict[str, Union[int, str, float]] = dict()
204+
if pipeline_options is None:
205+
pipeline_options = replace_string_bool_to_bool(json.load(open(
206206
os.path.join(os.path.dirname(__file__), '../configs/default_pipeline_options.json'))))
207-
self.pipeline_config.update(pipeline_config)
207+
self.pipeline_options.update(pipeline_options)
208208

209209
self.logger_port = logger_port
210210
if self.logger_port is None:
@@ -225,7 +225,7 @@ def __init__(
225225
def _check_and_get_default_budget(self) -> float:
226226
budget_type_choices_tabular = ('epochs', 'runtime')
227227
budget_choices = {
228-
budget_type: float(self.pipeline_config.get(budget_type, np.inf))
228+
budget_type: float(self.pipeline_options.get(budget_type, np.inf))
229229
for budget_type in budget_type_choices_tabular
230230
}
231231

@@ -234,7 +234,7 @@ def _check_and_get_default_budget(self) -> float:
234234
budget_type_choices = budget_type_choices_tabular + FORECASTING_BUDGET_TYPE
235235

236236
# budget is defined by epochs by default
237-
budget_type = str(self.pipeline_config.get('budget_type', 'epochs'))
237+
budget_type = str(self.pipeline_options.get('budget_type', 'epochs'))
238238
if self.budget_type is not None:
239239
budget_type = self.budget_type
240240

@@ -361,7 +361,7 @@ def run(
361361
init_params=init_params,
362362
budget=budget,
363363
budget_type=self.budget_type,
364-
pipeline_config=self.pipeline_config,
364+
pipeline_options=self.pipeline_options,
365365
logger_port=self.logger_port,
366366
all_supported_metrics=self.all_supported_metrics,
367367
search_space_updates=self.search_space_updates

‎autoPyTorch/evaluation/test_evaluator.py

Lines changed: 5 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -51,7 +51,7 @@ class TestEvaluator(AbstractEvaluator):
5151
The amount of epochs/time a configuration is allowed to run.
5252
budget_type (str):
5353
The budget type, which can be epochs or time
54-
pipeline_config (Optional[Dict[str, Any]]):
54+
pipeline_options (Optional[Dict[str, Any]]):
5555
Defines the content of the pipeline being evaluated. For example, it
5656
contains pipeline specific settings like logging name, or whether or not
5757
to use tensorboard.
@@ -113,7 +113,7 @@ def __init__(
113113
budget: float,
114114
configuration: Union[int, str, Configuration],
115115
budget_type: str = None,
116-
pipeline_config: Optional[Dict[str, Any]] = None,
116+
pipeline_options: Optional[Dict[str, Any]] = None,
117117
seed: int = 1,
118118
output_y_hat_optimization: bool = False,
119119
num_run: Optional[int] = None,
@@ -141,7 +141,7 @@ def __init__(
141141
budget_type=budget_type,
142142
logger_port=logger_port,
143143
all_supported_metrics=all_supported_metrics,
144-
pipeline_config=pipeline_config,
144+
pipeline_options=pipeline_options,
145145
search_space_updates=search_space_updates
146146
)
147147

@@ -206,7 +206,7 @@ def eval_test_function(
206206
include: Optional[Dict[str, Any]],
207207
exclude: Optional[Dict[str, Any]],
208208
disable_file_output: Optional[List[Union[str, DisableFileOutputParameters]]] = None,
209-
pipeline_config: Optional[Dict[str, Any]] = None,
209+
pipeline_options: Optional[Dict[str, Any]] = None,
210210
budget_type: str = None,
211211
init_params: Optional[Dict[str, Any]] = None,
212212
logger_port: Optional[int] = None,
@@ -230,7 +230,7 @@ def eval_test_function(
230230
budget_type=budget_type,
231231
logger_port=logger_port,
232232
all_supported_metrics=all_supported_metrics,
233-
pipeline_config=pipeline_config,
233+
pipeline_options=pipeline_options,
234234
search_space_updates=search_space_updates)
235235

236236
evaluator.fit_predict_and_loss()

‎autoPyTorch/evaluation/time_series_forecasting_train_evaluator.py

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -40,7 +40,7 @@ class TimeSeriesForecastingTrainEvaluator(TrainEvaluator):
4040
The amount of epochs/time a configuration is allowed to run.
4141
budget_type (str):
4242
The budget type, which can be epochs or time
43-
pipeline_config (Optional[Dict[str, Any]]):
43+
pipeline_options (Optional[Dict[str, Any]]):
4444
Defines the content of the pipeline being evaluated. For example, it
4545
contains pipeline specific settings like logging name, or whether or not
4646
to use tensorboard.
@@ -106,7 +106,7 @@ def __init__(self, backend: Backend, queue: Queue,
106106
metric: autoPyTorchMetric,
107107
budget: float,
108108
budget_type: str = None,
109-
pipeline_config: Optional[Dict[str, Any]] = None,
109+
pipeline_options: Optional[Dict[str, Any]] = None,
110110
configuration: Optional[Configuration] = None,
111111
seed: int = 1,
112112
output_y_hat_optimization: bool = True,
@@ -138,7 +138,7 @@ def __init__(self, backend: Backend, queue: Queue,
138138
logger_port=logger_port,
139139
keep_models=keep_models,
140140
all_supported_metrics=all_supported_metrics,
141-
pipeline_config=pipeline_config,
141+
pipeline_options=pipeline_options,
142142
search_space_updates=search_space_updates
143143
)
144144
self.datamanager = backend.load_datamanager()
@@ -456,7 +456,7 @@ def forecasting_eval_train_function(
456456
include: Optional[Dict[str, Any]],
457457
exclude: Optional[Dict[str, Any]],
458458
disable_file_output: Optional[List[Union[str, DisableFileOutputParameters]]] = None,
459-
pipeline_config: Optional[Dict[str, Any]] = None,
459+
pipeline_options: Optional[Dict[str, Any]] = None,
460460
budget_type: str = None,
461461
init_params: Optional[Dict[str, Any]] = None,
462462
logger_port: Optional[int] = None,
@@ -490,7 +490,7 @@ def forecasting_eval_train_function(
490490
The amount of epochs/time a configuration is allowed to run.
491491
budget_type (str):
492492
The budget type, which can be epochs or time
493-
pipeline_config (Optional[Dict[str, Any]]):
493+
pipeline_options (Optional[Dict[str, Any]]):
494494
Defines the content of the pipeline being evaluated. For example, it
495495
contains pipeline specific settings like logging name, or whether or not
496496
to use tensorboard.
@@ -550,7 +550,7 @@ def forecasting_eval_train_function(
550550
budget_type=budget_type,
551551
logger_port=logger_port,
552552
all_supported_metrics=all_supported_metrics,
553-
pipeline_config=pipeline_config,
553+
pipeline_options=pipeline_options,
554554
search_space_updates=search_space_updates,
555555
max_budget=max_budget,
556556
min_num_test_instances=min_num_test_instances,

‎autoPyTorch/evaluation/train_evaluator.py

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -60,7 +60,7 @@ class TrainEvaluator(AbstractEvaluator):
6060
The amount of epochs/time a configuration is allowed to run.
6161
budget_type (str):
6262
The budget type, which can be epochs or time
63-
pipeline_config (Optional[Dict[str, Any]]):
63+
pipeline_options (Optional[Dict[str, Any]]):
6464
Defines the content of the pipeline being evaluated. For example, it
6565
contains pipeline specific settings like logging name, or whether or not
6666
to use tensorboard.
@@ -121,7 +121,7 @@ def __init__(self, backend: Backend, queue: Queue,
121121
budget: float,
122122
configuration: Union[int, str, Configuration],
123123
budget_type: str = None,
124-
pipeline_config: Optional[Dict[str, Any]] = None,
124+
pipeline_options: Optional[Dict[str, Any]] = None,
125125
seed: int = 1,
126126
output_y_hat_optimization: bool = True,
127127
num_run: Optional[int] = None,
@@ -149,7 +149,7 @@ def __init__(self, backend: Backend, queue: Queue,
149149
budget_type=budget_type,
150150
logger_port=logger_port,
151151
all_supported_metrics=all_supported_metrics,
152-
pipeline_config=pipeline_config,
152+
pipeline_options=pipeline_options,
153153
search_space_updates=search_space_updates
154154
)
155155

@@ -420,7 +420,7 @@ def eval_train_function(
420420
include: Optional[Dict[str, Any]],
421421
exclude: Optional[Dict[str, Any]],
422422
disable_file_output: Optional[List[Union[str, DisableFileOutputParameters]]] = None,
423-
pipeline_config: Optional[Dict[str, Any]] = None,
423+
pipeline_options: Optional[Dict[str, Any]] = None,
424424
budget_type: str = None,
425425
init_params: Optional[Dict[str, Any]] = None,
426426
logger_port: Optional[int] = None,
@@ -452,7 +452,7 @@ def eval_train_function(
452452
The amount of epochs/time a configuration is allowed to run.
453453
budget_type (str):
454454
The budget type, which can be epochs or time
455-
pipeline_config (Optional[Dict[str, Any]]):
455+
pipeline_options (Optional[Dict[str, Any]]):
456456
Defines the content of the pipeline being evaluated. For example, it
457457
contains pipeline specific settings like logging name, or whether or not
458458
to use tensorboard.
@@ -506,7 +506,7 @@ def eval_train_function(
506506
budget_type=budget_type,
507507
logger_port=logger_port,
508508
all_supported_metrics=all_supported_metrics,
509-
pipeline_config=pipeline_config,
509+
pipeline_options=pipeline_options,
510510
search_space_updates=search_space_updates,
511511
)
512512
evaluator.fit_predict_and_loss()

0 commit comments

Comments
(0)

AltStyle によって変換されたページ (->オリジナル) /