def _train_desc(self, model_desc:ModelDesc, conf_train:Config)->MetricsStats:
"""Train given description"""
# region conf vars
conf_trainer = conf_train['trainer']
conf_loader = conf_train['loader']
trainer_title = conf_trainer['title']
epochs = conf_trainer['epochs']
drop_path_prob = conf_trainer['drop_path_prob']
# endregion
logger.pushd(trainer_title)
if epochs == 0:
# nothing to pretrain, save time
metrics_stats = MetricsStats(model_desc, None, None)
else:
model = nas_utils.model_from_desc(model_desc,
droppath=drop_path_prob>0.0,
affine=True)
# get data
train_dl, val_dl = self.get_data(conf_loader)
assert train_dl is not None
trainer = Trainer(conf_trainer, model, checkpoint=None)
train_metrics = trainer.fit(train_dl, val_dl)
metrics_stats = Search._create_metrics_stats(model, train_metrics, self.finalizers)
logger.popd()
return metrics_stats
def eval_arch(conf_eval: Config, cell_builder: Optional[CellBuilder]):
logger.pushd('eval_arch')
# region conf vars
conf_loader = conf_eval['loader']
model_filename = conf_eval['model_filename']
metric_filename = conf_eval['metric_filename']
conf_checkpoint = conf_eval['checkpoint']
resume = conf_eval['resume']
conf_train = conf_eval['trainer']
# endregion
if cell_builder:
cell_builder.register_ops()
model = create_model(conf_eval)
# get data
train_dl, _, test_dl = data.get_data(conf_loader)
assert train_dl is not None and test_dl is not None
checkpoint = nas_utils.create_checkpoint(conf_checkpoint, resume)
trainer = Trainer(conf_train, model, checkpoint)
train_metrics = trainer.fit(train_dl, test_dl)
train_metrics.save(metric_filename)
# save model
if model_filename:
model_filename = utils.full_path(model_filename)
ml_utils.save_model(model, model_filename)
logger.info({'model_save_path': model_filename})
logger.popd()
def train_test()->Metrics:
conf = common.get_conf()
conf_eval = conf['nas']['eval']
# region conf vars
conf_loader = conf_eval['loader']
conf_trainer = conf_eval['trainer']
# endregion
conf_trainer['validation']['freq']=1
conf_trainer['epochs'] = 1
conf_loader['train_batch'] = 128
conf_loader['test_batch'] = 4096
conf_loader['cutout'] = 0
conf_trainer['drop_path_prob'] = 0.0
conf_trainer['grad_clip'] = 0.0
conf_trainer['aux_weight'] = 0.0
Net = cifar10_models.resnet34
model = Net().to(torch.device('cuda'))
# get data
data_loaders = data.get_data(conf_loader)
assert data_loaders.train_dl is not None and data_loaders.test_dl is not None
trainer = Trainer(conf_trainer, model, None)
trainer.fit(data_loaders)
met = trainer.get_metrics()
return met
def train_test(conf_eval: Config):
logger = get_logger()
# region conf vars
conf_loader = conf_eval['loader']
conf_trainer = conf_eval['trainer']
# endregion
conf_trainer['validation']['freq'] = 1
conf_trainer['epochs'] = 10
conf_loader['train_batch'] = 128
conf_loader['test_batch'] = 4096
conf_loader['cutout'] = 0
conf_trainer['drop_path_prob'] = 0.0
conf_trainer['grad_clip'] = 0.0
conf_trainer['aux_weight'] = 0.0
device = torch.device(conf_eval['device'])
Net = cifar10_models.resnet34
model = Net().to(device)
# get data
train_dl, _, test_dl = data.get_data(conf_loader)
assert train_dl is not None and test_dl is not None
trainer = Trainer(conf_trainer, model, device, None, False)
trainer.fit(train_dl, test_dl)
def train_model_desc(self, model_desc:ModelDesc, conf_train:Config)\
->Optional[ModelMetrics]:
"""Train given description"""
# region conf vars
conf_trainer = conf_train['trainer']
conf_loader = conf_train['loader']
trainer_title = conf_trainer['title']
epochs = conf_trainer['epochs']
drop_path_prob = conf_trainer['drop_path_prob']
# endregion
# if epochs ==0 then nothing to train, so save time
if epochs <= 0:
return None
logger.pushd(trainer_title)
model = Model(model_desc, droppath=drop_path_prob > 0.0, affine=True)
# get data
data_loaders = self.get_data(conf_loader)
trainer = Trainer(conf_trainer, model, checkpoint=None)
train_metrics = trainer.fit(data_loaders)
logger.popd()
return ModelMetrics(model, train_metrics)
def train_model(self, conf_train: Config, model: nn.Module,
checkpoint: Optional[CheckPoint]) -> Metrics:
conf_loader = conf_train['loader']
conf_train = conf_train['trainer']
# get data
train_dl, test_dl = self.get_data(conf_loader)
trainer = Trainer(conf_train, model, checkpoint)
train_metrics = trainer.fit(train_dl, test_dl)
return train_metrics
def train_test(conf_eval: Config):
conf_loader = conf_eval['loader']
conf_trainer = conf_eval['trainer']
# create model
Net = cifar10_models.resnet34
model = Net().to(torch.device('cuda', 0))
# get data
train_dl, _, test_dl = data.get_data(conf_loader)
# train!
trainer = Trainer(conf_trainer, model)
trainer.fit(train_dl, test_dl)
def main():
#6, 7, 9, 10, 16
#model = model_builder.build(model_builder.EXAMPLE_DESC_MATRIX, model_builder.EXAMPLE_VERTEX_OPS)
nsds = Nasbench101Dataset('~/dataroot/nasbench_ds/nasbench_full.pkl')
conf = common_init(config_filepath='confs/algos/nasbench101.yaml')
conf_eval = conf['nas']['eval']
conf_loader = conf_eval['loader']
conf_trainer = conf_eval['trainer']
model = nsds.create_model(5) # 401277 is same model as example
data_loaders = data.get_data(conf_loader)
trainer = Trainer(conf_trainer, model)
trainer.fit(data_loaders)
