multiml.task.pytorch.PytorchBaseTask

class multiml.task.pytorch.PytorchBaseTask(device='cpu', gpu_ids=None, torchinfo=False, amp=False, torch_compile=False, dataset_args=None, dataloader_args=None, batch_sampler=False, metric_sample=1, **kwargs)

Base task for PyTorch model.

Examples

>>> # your pytorch model
>>> class MyPytorchModel(nn.Module):
>>>     def __init__(self, inputs=2, outputs=2):
>>>         super(MyPytorchModel, self).__init__()
>>>
>>>         self.fc1 = nn.Linear(inputs, outputs)
>>>         self.relu = nn.ReLU()
>>>
>>>     def forward(self, x):
>>>         return self.relu(self.fc1(x))
>>>
>>> # create task instance
>>> task = PytorchBaseTask(storegate=storegate,
>>>                        model=MyPytorchModel,
>>>                        input_var_names=('x0', 'x1'),
>>>                        output_var_names='outputs-pytorch',
>>>                        true_var_names='labels',
>>>                        optimizer='SGD',
>>>                        optimizer_args=dict(lr=0.1),
>>>                        loss='CrossEntropyLoss')
>>> task.set_hps({'num_epochs': 5})
>>> task.execute()
>>> task.finalize()

__init__(device='cpu', gpu_ids=None, torchinfo=False, amp=False, torch_compile=False, dataset_args=None, dataloader_args=None, batch_sampler=False, metric_sample=1, **kwargs)

Initialize the pytorch base task.

Parameters:

device (str or obj) – pytorch device, e.g. ‘cpu’, ‘cuda’.
gpu_ids (list) – GPU identifiers, e.g. [0, 1, 2]. data_parallel mode is enabled if gpu_ids is given.
torchinfo (bool) – show torchinfo summary after model compile.
amp (bool) – (expert option) enable amp mode.
torch_compile (bool) – (expert option) enable torch.compile.
dataset_args (dict) – args passed to default DataSet creation.
dataloader_args (dict) – args passed to default DataLoader creation.
batch_sampler (bool) – user batch_sampler or not.
metric_sample (float or int) – sampling ratio for running metrics.

Methods

`__init__`([device, gpu_ids, torchinfo, amp, ...])	Initialize the pytorch base task.
`add_device`(data, device)	Add data to device.
`build_model`()	Build model.
`compile`()	Compile pytorch ml objects.
`compile_device`()	Compile device.
`compile_loss`()	Compile pytorch loss.
`compile_model`()	Compile pytorch model.
`compile_optimizer`()	Compile pytorch optimizer and scheduler.
`compile_var_names`()	Compile var_names.
`do_test`()	Perform test phase or not.
`do_train`()	Perform train phase or not.
`do_valid`()	Perform valid phase or not.
`dump_model`([extra_args])	Dump current pytorch model.
`execute`()	Execute a task.
`finalize`()	Finalize base task.
`fit`([train_data, valid_data, dataloaders, ...])	Train model over epoch.
`fit_predict`([fit_args, predict_args])	Fit and predict model.
`fix_submodule`(target)	Fix given parameters of model.
`get_batch_sampler`(phase, dataset)	Returns batch sampler.
`get_dataset`([data, phase, preload, callbacks])	Returns dataset from given ndarray data.
`get_input_true_data`(phase)	Get input and true data.
`get_input_var_shapes`([phase])	Get shape of input_var_names.
`get_metadata`(metadata_key)	Returns metadata.
`get_pred_index`()	Returns prediction index passed to loss calculation.
`get_storegate_dataset`(phase[, preload, ...])	Returns storegate dataset.
`get_tensor_dataset`(data[, callbacks])	Returns tensor dataset from given ndarray data.
`get_unique_id`()	Returns unique identifier of task.
`load_metadata`()	Load metadata.
`load_model`()	Load pre-trained pytorch model weights.
`predict`([data, dataloader, phase, label])	Predict model.
`predict_update`([data, phase])	Predict and update data in StoreGate.
`prepare_dataloader`([data, phase, ...])	Prepare dataloader.
`prepare_dataloaders`([phases, dataset_args, ...])	Prepare dataloaders for all phases.
`set_hps`(params)	Set hyperparameters to this task.
`show_info`()	Print information.
`step_batch`(data, phase[, label])	Process batch data and update weights.
`step_epoch`(epoch, phase, dataloader[, label])	Process model for given epoch and phase.
`step_loss`(outputs, labels)	Process loss function.
`step_model`(inputs)	Process model.
`step_optimizer`(loss)	Process optimizer.
`update`(data[, phase])	Update data in storegate.

Attributes

`input_saver_key`	Return input_saver_key.
`input_var_names`	Returns input_var_names.
`job_id`	Return job_id of task.
`ml`	Returns ML data class.
`name`	Return name of task.
`output_saver_key`	Return output_saver_key.
`output_var_names`	Returns output_var_names.
`phases`	Returns ML phases.
`pool_id`	Return pool_id of task.
`pred_var_names`	Returns pred_var_names.
`save_var_names`	Returns save_var_names.
`saver`	Return saver of task.
`storegate`	Return storegate of task.
`subtask_id`	Return subtask_id of task.
`task_id`	Return task_id of task.
`trial_id`	Return trial_id of task.
`true_var_names`	Returns true_var_names.

__init__(device='cpu', gpu_ids=None, torchinfo=False, amp=False, torch_compile=False, dataset_args=None, dataloader_args=None, batch_sampler=False, metric_sample=1, **kwargs)

Initialize the pytorch base task.

Parameters:

device (str or obj) – pytorch device, e.g. ‘cpu’, ‘cuda’.
gpu_ids (list) – GPU identifiers, e.g. [0, 1, 2]. data_parallel mode is enabled if gpu_ids is given.
torchinfo (bool) – show torchinfo summary after model compile.
amp (bool) – (expert option) enable amp mode.
torch_compile (bool) – (expert option) enable torch.compile.
dataset_args (dict) – args passed to default DataSet creation.
dataloader_args (dict) – args passed to default DataLoader creation.
batch_sampler (bool) – user batch_sampler or not.
metric_sample (float or int) – sampling ratio for running metrics.