# Copyright 2022 Cerebras Systems.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import numpy as np
import torch
import torchvision
from cerebras.modelzoo.data.vision.classification.dataset_factory import (
Processor,
VisionSubset,
)
[docs]class Flowers102Processor(Processor):
[docs] def __init__(self, params):
super().__init__(params)
self.allowable_split = ["train", "val", "test"]
self.num_classes = 102
def create_dataset(self, use_training_transforms=True, split="train"):
self.check_split_valid(split)
transform, target_transform = self.process_transform(
use_training_transforms
)
dataset = torchvision.datasets.Flowers102(
root=self.data_dir,
split=split,
transform=transform,
target_transform=target_transform,
download=False,
)
return dataset
def create_vtab_dataset(
self, use_1k_sample=True, train_split_percent=None, seed=42
):
train_transform, train_target_transform = self.process_transform(
use_training_transforms=True
)
eval_transform, eval_target_transform = self.process_transform(
use_training_transforms=False
)
train_set = torchvision.datasets.Flowers102(
root=self.data_dir,
split="train",
transform=train_transform,
target_transform=train_target_transform,
download=False,
)
val_set = torchvision.datasets.Flowers102(
root=self.data_dir,
split="val",
transform=eval_transform,
target_transform=eval_target_transform,
download=False,
)
test_set = torchvision.datasets.Flowers102(
root=self.data_dir,
split="test",
transform=eval_transform,
target_transform=eval_target_transform,
download=False,
)
rng = np.random.default_rng(seed)
train_sample_idx = self.create_shuffled_idx(len(train_set), rng)
val_sample_idx = self.create_shuffled_idx(len(val_set), rng)
if use_1k_sample:
train_set = VisionSubset(train_set, train_sample_idx[:800])
val_set = VisionSubset(val_set, val_sample_idx[:200])
return train_set, val_set, test_set
else:
if train_split_percent:
# if train_split_percent is specified, the training set and
# validation set are combined and then split according to the
# specified percentage.
split_percent = [train_split_percent, 100 - train_split_percent]
train_set_splits = self.split_dataset(
train_set, split_percent, seed
)
val_set_splits = self.split_dataset(
val_set, split_percent, seed
)
# update transform so that the split dataset at index 0 uses
# training transform while those at index 1 uses eval transform
train_set_splits[0].set_transforms(train_transform)
val_set_splits[0].set_transforms(train_transform)
train_set_splits[1].set_transforms(eval_transform)
val_set_splits[1].set_transforms(eval_transform)
new_train_set = torch.utils.data.ConcatDataset(
[train_set_splits[0], val_set_splits[0]]
)
new_val_set = torch.utils.data.ConcatDataset(
[train_set_splits[1], val_set_splits[1]]
)
return new_train_set, new_val_set, test_set
else:
return train_set, val_set, test_set