Table of Contents
Shortcuts

BatchSizeScheduler#

class ignite.handlers.param_scheduler.BatchSizeScheduler(dataloader, scheduler_fn, save_history=False)[source]#

Scheduler to update the batch size of a DataLoader’s BatchSampler during training.

Parameters:

  • dataloader (DataLoader) – torch DataLoader with a BatchSampler.

  • scheduler_fn (Callable[[int], int]) – a callable that takes an event index and returns the new batch size (int).

  • save_history (bool) – whether to log the batch size values to engine.state.param_history, (default=False).

Examples

from collections import OrderedDict

import torch
from torch import nn, optim

from ignite.engine import *
from ignite.handlers import *
from ignite.metrics import *
from ignite.metrics.clustering import *
from ignite.metrics.fairness import *
from ignite.metrics.rec_sys import *
from ignite.metrics.regression import *
from ignite.utils import *

# create default evaluator for doctests

def eval_step(engine, batch):
    return batch

default_evaluator = Engine(eval_step)

# create default optimizer for doctests

param_tensor = torch.zeros([1], requires_grad=True)
default_optimizer = torch.optim.SGD([param_tensor], lr=0.1)

# create default trainer for doctests
# as handlers could be attached to the trainer,
# each test must define his own trainer using `.. testsetup:`

def get_default_trainer():

    def train_step(engine, batch):
        return batch

    return Engine(train_step)

# create default model for doctests

default_model = nn.Sequential(OrderedDict([
    ('base', nn.Linear(4, 2)),
    ('fc', nn.Linear(2, 1))
]))

manual_seed(666)

from torch.utils.data import DataLoader, TensorDataset
import torch

default_trainer = get_default_trainer()

dataset = TensorDataset(torch.arange(100))
dataloader = DataLoader(dataset, batch_size=10)

# Linearly increase batch size from 10 to 50 over 5 epochs
scheduler = BatchSizeScheduler(dataloader, scheduler_fn=lambda e: 10 + e * 10)

default_trainer.add_event_handler(Events.EPOCH_STARTED, scheduler)

New in version 0.5.4.

Methods

attach

Attach the handler to the engine.

get_param

Method to get current batch size value.

simulate_values

Method to simulate scheduled batch size values during num_events events.
attach(engine, event=Events.EPOCH_STARTED)[source]#

Attach the handler to the engine.

Parameters:
Return type:

None

get_param()[source]#

Method to get current batch size value.

Return type:

int

classmethod simulate_values(num_events, **scheduler_kwargs)[source]#

Method to simulate scheduled batch size values during num_events events.

Parameters:

  • num_events (int) – number of events during the simulation.

  • scheduler_kwargs (Any) – scheduler configuration kwargs (requires scheduler_fn).

Returns:

list of [event_index, batch_size] pairs.

Return type:

list[list[int]]

×

Search Docs