def from_config(config):
if not config:
return None
config_dict = config.to_dict()
config_dict = remove_empty_keys(config_dict)
return SubGraph.objects.create(definition=json.dumps(config_dict))
def from_config(config):
if not config:
return None
config_dict = config.to_dict()
config_dict = remove_empty_keys(config_dict)
return AgentMemory.objects.create(**config_dict)
def from_config(config):
if not config:
return None
config_dict = config.to_dict()
config_dict.pop('memory_config', None)
config_dict['memory'] = AgentMemory.from_config(config.memory_config)
config_dict = remove_empty_keys(config_dict)
return Agent.objects.create(**config_dict)
def from_config(config):
if not config:
return None
config_dict = config.to_dict()
config_dict['definition'] = json.dumps(
config_dict.get('definition', {}))
config_dict = remove_empty_keys(config_dict)
pipeline = Pipeline.objects.create(**config_dict)
return pipeline
def from_config(config):
config_dict = BaseExperiment.from_config(config)
config_dict['agent'] = Agent.from_config(config.agent_config)
config_dict.pop('agent_config', None)
config_dict['environment'] = Environment.from_config(
config.environment_config)
config_dict.pop('environment_config', None)
eval_metrics = config_dict.pop('eval_metrics')
config_dict = remove_empty_keys(config_dict)
exp = RLExperiment.objects.create(**config_dict)
exp.eval_metrics = eval_metrics
return exp
def from_config(config):
config_dict = BaseExperiment.from_config(config)
config_dict['estimator'] = Estimator.from_config(
config.estimator_config)
config_dict.pop('estimator_config', None)
config_dict['train_input_data'] = InputData.from_config(
config.train_input_data_config)
config_dict.pop('train_input_data_config', None)
config_dict['eval_input_data'] = InputData.from_config(
config.eval_input_data_config)
config_dict.pop('eval_input_data_config', None)
eval_metrics = config_dict.pop('eval_metrics')
config_dict = remove_empty_keys(config_dict)
exp = Experiment.objects.create(**config_dict)
exp.eval_metrics = eval_metrics
return exp
def from_config(config):
if not config:
return None
params = {}
fields = [f.name for f in PolyaxonModel._meta.get_fields()]
config_dict = config.to_dict()
if not isinstance(config_dict.get('summaries', []), list):
config_dict['summaries'] = [config_dict['summaries']]
config_dict['params'] = params
config_dict['loss'] = Loss.from_config(config.loss_config)
config_dict.pop('loss_config', None)
config_dict['optimizer'] = Optimizer.from_config(
config.optimizer_config)
config_dict.pop('optimizer_config', None)
config_dict['graph'] = SubGraph.from_config(config.graph_config)
config_dict.pop('graph_config', None)
config_dict['encoder'] = Encoder.from_config(config.encoder_config)
config_dict.pop('encoder_config', None)
config_dict['decoder'] = Decoder.from_config(config.decoder_config)
config_dict.pop('decoder_config', None)
config_dict['bridge'] = Bridge.from_config(config.bridge_config)
config_dict.pop('bridge_config', None)
# also remove eval_metrics_config
config_dict.pop('eval_metrics_config', None)
# Rest of the keys should go to params
keys = list(config_dict.keys())
for key in keys:
if key not in fields:
params[key] = config_dict.pop(key)
config_dict['params'] = params
config_dict = remove_empty_keys(config_dict)
model = PolyaxonModel.objects.create(**config_dict)
model.eval_metrics = [
Metric.from_config(m) for m in config.eval_metrics_config
]
return model