[Refactor] Deprecate imgs_per_gpu and use samples_per_gpu (#204)

* [Refactor] change imgs_per_gpu to samples_per_gpu in config files

* [Docs] change imgs_per_gpu to samples_per_gpu in docs

* [Refactor] change imgs_per_gpu to samples_per_gpu in codes and add warnings

* [Fix] fix isort

* [Docs] fix docs format

* [Refactor] add related UT codes

* [Fix] fix isort

configs/benchmarks/classification/_base_/datasets/cifar10.py

@@ -20,7 +20,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=128,
+    samples_per_gpu=128,
     workers_per_gpu=2,
     train=dict(
         type=dataset_type,

configs/benchmarks/classification/_base_/datasets/imagenet.py

@@ -23,7 +23,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # total 32x8=256, 8GPU linear cls
+    samples_per_gpu=32,  # total 32x8=256, 8GPU linear cls
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,

configs/benchmarks/classification/_base_/datasets/inaturalist2018.py

@@ -23,7 +23,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # total 32x8=256, 8GPU linear cls
+    samples_per_gpu=32,  # total 32x8=256, 8GPU linear cls
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,

configs/benchmarks/classification/_base_/datasets/places205.py

@@ -25,7 +25,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # total 32x8=256, 8GPU linear cls
+    samples_per_gpu=32,  # total 32x8=256, 8GPU linear cls
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,

configs/benchmarks/classification/imagenet/imagenet_10percent/resnet50_head1_4xb64-steplr1e-1-20e_in1k-10pct.py

@@ -10,7 +10,7 @@ model = dict(backbone=dict(norm_cfg=dict(type='SyncBN')))
 
 # dataset settings
 data = dict(
-    imgs_per_gpu=64,  # total 64x4=256
+    samples_per_gpu=64,  # total 64x4=256
     train=dict(
         data_source=dict(ann_file='data/imagenet/meta/train_10pct.txt')))
 

configs/benchmarks/classification/imagenet/imagenet_1percent/resnet50_head1_4xb64-steplr1e-1-20e_in1k-1pct.py

@@ -10,7 +10,7 @@ model = dict(backbone=dict(norm_cfg=dict(type='SyncBN')))
 
 # dataset settings
 data = dict(
-    imgs_per_gpu=64,  # total 64x4=256
+    samples_per_gpu=64,  # total 64x4=256
     train=dict(
         data_source=dict(ann_file='data/imagenet/meta/train_1percent.txt')))
 

configs/benchmarks/classification/imagenet/resnet50_8xb512-coslr-90e_in1k.py

@@ -8,7 +8,7 @@ _base_ = [
 model = dict(backbone=dict(frozen_stages=4))
 
 # dataset summary
-data = dict(imgs_per_gpu=512)  # total 512*8=4096, 8GPU linear cls
+data = dict(samples_per_gpu=512)  # total 512*8=4096, 8GPU linear cls
 
 # simsiam setting
 # runtime settings

configs/benchmarks/classification/imagenet/vit-small-p16_8xb128-coslr-90e_in1k.py

@@ -9,7 +9,7 @@ _base_ = [
 model = dict(backbone=dict(frozen_stages=12, norm_eval=True))
 
 # dataset summary
-data = dict(imgs_per_gpu=128)  # total 128*8=1024, 8 GPU linear cls
+data = dict(samples_per_gpu=128)  # total 128*8=1024, 8 GPU linear cls
 
 # optimizer
 optimizer = dict(type='SGD', lr=12, momentum=0.9, weight_decay=0.)

configs/benchmarks/classification/svm_voc07.py

@@ -5,7 +5,7 @@ split_name = ['voc07_trainval', 'voc07_test']
 img_norm_cfg = dict(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
 
 data = dict(
-    imgs_per_gpu=32,
+    samples_per_gpu=32,
     workers_per_gpu=4,
     extract=dict(
         type=dataset_type,

configs/benchmarks/classification/tsne_imagenet.py

@@ -4,7 +4,7 @@ name = 'imagenet_val'
 img_norm_cfg = dict(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
 
 data = dict(
-    imgs_per_gpu=8,
+    samples_per_gpu=8,
     workers_per_gpu=4,
     extract=dict(
         type='SingleViewDataset',

configs/selfsup/_base_/datasets/imagenet_byol.py

@@ -51,7 +51,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # total 32*8(gpu)=256
+    samples_per_gpu=32,  # total 32*8(gpu)=256
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,

configs/selfsup/_base_/datasets/imagenet_deepcluster.py

@@ -31,7 +31,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=64,  # 64*8
+    samples_per_gpu=64,  # 64*8
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,
@@ -49,7 +49,7 @@ custom_hooks = [
     dict(
         type='DeepClusterHook',
         extractor=dict(
-            imgs_per_gpu=128,
+            samples_per_gpu=128,
             workers_per_gpu=8,
             dataset=dict(
                 type=dataset_type,

configs/selfsup/_base_/datasets/imagenet_mocov1.py

@@ -23,7 +23,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # total 32*8=256
+    samples_per_gpu=32,  # total 32*8=256
     workers_per_gpu=4,
     drop_last=True,
     train=dict(

configs/selfsup/_base_/datasets/imagenet_mocov2.py

@@ -30,7 +30,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # total 32*8=256
+    samples_per_gpu=32,  # total 32*8=256
     workers_per_gpu=4,
     drop_last=True,
     train=dict(

configs/selfsup/_base_/datasets/imagenet_mocov3.py

@@ -51,7 +51,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=256,  # 256*16(gpu)=4096
+    samples_per_gpu=256,  # 256*16(gpu)=4096
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,

configs/selfsup/_base_/datasets/imagenet_npid.py

@@ -23,7 +23,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # total 32*8
+    samples_per_gpu=32,  # total 32*8
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,

configs/selfsup/_base_/datasets/imagenet_odc.py

@@ -31,7 +31,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=64,  # 64*8
+    samples_per_gpu=64,  # 64*8
     sampling_replace=True,
     workers_per_gpu=4,
     train=dict(
@@ -50,7 +50,7 @@ custom_hooks = [
     dict(
         type='DeepClusterHook',
         extractor=dict(
-            imgs_per_gpu=128,
+            samples_per_gpu=128,
             workers_per_gpu=8,
             dataset=dict(
                 type=dataset_type,

configs/selfsup/_base_/datasets/imagenet_relative-loc.py

@@ -21,7 +21,7 @@ prefetch = False
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=64,  # 64 x 8 = 512
+    samples_per_gpu=64,  # 64 x 8 = 512
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,

configs/selfsup/_base_/datasets/imagenet_rotation-pred.py

@@ -23,7 +23,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=16,  # (16*4) x 8 = 512
+    samples_per_gpu=16,  # (16*4) x 8 = 512
     workers_per_gpu=2,
     train=dict(
         type=dataset_type,

configs/selfsup/_base_/datasets/imagenet_simclr.py

@@ -29,7 +29,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # total 32*8
+    samples_per_gpu=32,  # total 32*8
     workers_per_gpu=4,
     train=dict(
         type=dataset_type,

configs/selfsup/_base_/datasets/imagenet_swav_mcrop-2-6.py

@@ -51,7 +51,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # total 32*8=256
+    samples_per_gpu=32,  # total 32*8=256
     workers_per_gpu=4,
     drop_last=True,
     train=dict(

configs/selfsup/byol/byol_resnet50_8xb256-fp16-accum2-coslr-200e_in1k.py

@@ -6,7 +6,7 @@ _base_ = [
 ]
 
 # dataset summary
-data = dict(imgs_per_gpu=256)
+data = dict(samples_per_gpu=256)
 
 # additional hooks
 # interval for accumulate gradient, total 8*256*2(interval)=4096

configs/selfsup/mocov3/mocov3_vit-small-p16_32xb128-fp16-coslr-300e_in1k-224.py

@@ -58,7 +58,8 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=128, train=dict(pipelines=[train_pipeline1, train_pipeline2]))
+    samples_per_gpu=128,
+    train=dict(pipelines=[train_pipeline1, train_pipeline2]))
 
 # MoCo v3 use the same momentum update method as BYOL
 custom_hooks = [dict(type='MomentumUpdateHook')]

configs/selfsup/simclr/simclr_resnet50_8xb64-coslr-200e_in1k.py

@@ -1,4 +1,4 @@
 _base_ = 'simclr_resnet50_8xb32-coslr-200e_in1k.py'
 
 # dataset summary
-data = dict(imgs_per_gpu=64)  # total 64*8
+data = dict(samples_per_gpu=64)  # total 64*8

configs/selfsup/swav/swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96.py

@@ -13,7 +13,7 @@ custom_hooks = [
     dict(
         type='SwAVHook',
         priority='VERY_HIGH',
-        batch_size={{_base_.data.imgs_per_gpu}},
+        batch_size={{_base_.data.samples_per_gpu}},
         epoch_queue_starts=15,
         crops_for_assign=[0, 1],
         feat_dim=128,

docs/en/tutorials/0_config.md

@@ -207,7 +207,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # Batch size of a single GPU, total 32*8=256
+    samples_per_gpu=32,  # Batch size of a single GPU, total 32*8=256
     workers_per_gpu=4,  # Worker to pre-fetch data for each single GPU
     drop_last=True,  # Whether to drop the last batch of data
     train=dict(
@@ -304,7 +304,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,
+    samples_per_gpu=32,
     workers_per_gpu=4,
     drop_last=True,
     train=dict(type=dataset_type, type=data_source, data_prefix=...),

docs/en/tutorials/4_schedule.md

@@ -154,14 +154,14 @@ When there is not enough computation resource, the batch size can only be set to
 Here is an example:
 
 ```python
-data = dict(imgs_per_gpu=64)
+data = dict(samples_per_gpu=64)
 optimizer_config = dict(type="DistOptimizerHook", update_interval=4)
 ```
 
 Indicates that during training, back-propagation is performed every 4 iters. And the above is equivalent to:
 
 ```python
-data = dict(imgs_per_gpu=256)
+data = dict(samples_per_gpu=256)
 optimizer_config = dict(type="OptimizerHook")
 ```
 

docs/en/tutorials/6_benchmarks.md

@@ -70,7 +70,7 @@ bash tools/benchmarks/classification/slurm_train_linear.sh ${PARTITION} ${JOB_NA
 ```
 
 Remarks:
-- The default GPU number is 8. When changing GPUS, please also change imgs_per_gpu in the config file accordingly to ensure the total batch size is 256.
+- The default GPU number is 8. When changing GPUS, please also change `samples_per_gpu` in the config file accordingly to ensure the total batch size is 256.
 - `CONFIG`: Use config files under `configs/benchmarks/classification/`, excluding svm_voc07.py and tsne_imagenet.py and imagenet_*percent folders.
 - `PRETRAIN`: the pretrained model file.
 

docs/zh_cn/tutorials/0_config.md

@@ -207,7 +207,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,  # Batch size of a single GPU, total 32*8=256
+    samples_per_gpu=32,  # Batch size of a single GPU, total 32*8=256
     workers_per_gpu=4,  # Worker to pre-fetch data for each single GPU
     drop_last=True,  # Whether to drop the last batch of data
     train=dict(
@@ -304,7 +304,7 @@ if not prefetch:
 
 # dataset summary
 data = dict(
-    imgs_per_gpu=32,
+    samples_per_gpu=32,
     workers_per_gpu=4,
     drop_last=True,
     train=dict(type=dataset_type, type=data_source, data_prefix=...),

docs/zh_cn/tutorials/4_schedule.md

@@ -154,14 +154,14 @@ optimizer_config = dict(grad_clip=dict(max_norm=35, norm_type=2))
 用例如下：
 
 ```python
-data = dict(imgs_per_gpu=64)
+data = dict(samples_per_gpu=64)
 optimizer_config = dict(type="DistOptimizerHook", update_interval=4)
 ```
 
 表示训练时，每 4 个 iter 执行一次反向传播。由于此时单张 GPU 上的批次大小为 64，也就等价于单张 GPU 上一次迭代的批次大小为 256，也即：
 
 ```python
-data = dict(imgs_per_gpu=256)
+data = dict(samples_per_gpu=256)
 optimizer_config = dict(type="OptimizerHook")
 ```
 

docs/zh_cn/tutorials/6_benchmarks.md

@@ -70,7 +70,7 @@ bash tools/benchmarks/classification/slurm_train_linear.sh ${PARTITION} ${JOB_NA
 ```
 
 Remarks:
-- The default GPU number is 8. When changing GPUS, please also change imgs_per_gpu in the config file accordingly to ensure the total batch size is 256.
+- The default GPU number is 8. When changing GPUS, please also change `samples_per_gpu` in the config file accordingly to ensure the total batch size is 256.
 - `CONFIG`: Use config files under `configs/benchmarks/classification/`, excluding svm_voc07.py and tsne_imagenet.py and imagenet_*percent folders.
 - `PRETRAIN`: the pretrained model file.
 

mmselfsup/apis/train.py

@@ -75,13 +75,26 @@ def train_model(model,
 
     # prepare data loaders
     dataset = dataset if isinstance(dataset, (list, tuple)) else [dataset]
+    if 'imgs_per_gpu' in cfg.data:
+        logger.warning('"imgs_per_gpu" is deprecated. '
+                       'Please use "samples_per_gpu" instead')
+        if 'samples_per_gpu' in cfg.data:
+            logger.warning(
+                f'Got "imgs_per_gpu"={cfg.data.imgs_per_gpu} and '
+                f'"samples_per_gpu"={cfg.data.samples_per_gpu}, "imgs_per_gpu"'
+                f'={cfg.data.imgs_per_gpu} is used in this experiments')
+        else:
+            logger.warning(
+                'Automatically set "samples_per_gpu"="imgs_per_gpu"='
+                f'{cfg.data.imgs_per_gpu} in this experiments')
+        cfg.data.samples_per_gpu = cfg.data.imgs_per_gpu
 
     data_loaders = [
         build_dataloader(
             ds,
-            cfg.data.imgs_per_gpu,
-            cfg.data.workers_per_gpu,
-            # cfg.gpus will be ignored if distributed
+            samples_per_gpu=cfg.data.samples_per_gpu,
+            workers_per_gpu=cfg.data.workers_per_gpu,
+            # `num_gpus` will be ignored if distributed
             num_gpus=len(cfg.gpu_ids),
             dist=distributed,
             replace=getattr(cfg.data, 'sampling_replace', False),
@@ -161,7 +174,7 @@ def train_model(model,
         val_dataset = build_dataset(cfg.data.val)
         val_dataloader = build_dataloader(
             val_dataset,
-            imgs_per_gpu=cfg.data.imgs_per_gpu,
+            samples_per_gpu=cfg.data.samples_per_gpu,
             workers_per_gpu=cfg.data.workers_per_gpu,
             dist=distributed,
             shuffle=False,

mmselfsup/core/hooks/deepcluster_hook.py

@@ -7,6 +7,7 @@ from mmcv.utils import print_log
 
 from mmselfsup.utils import Extractor
 from mmselfsup.utils import clustering as _clustering
+from mmselfsup.utils import get_root_logger
 
 
 @HOOKS.register_module()
@@ -41,6 +42,23 @@ class DeepClusterHook(Hook):
             interval=1,
             dist_mode=True,
             data_loaders=None):
+
+        logger = get_root_logger()
+        if 'imgs_per_gpu' in extractor:
+            logger.warning('"imgs_per_gpu" is deprecated. '
+                           'Please use "samples_per_gpu" instead')
+            if 'samples_per_gpu' in extractor:
+                logger.warning(
+                    f'Got "imgs_per_gpu"={extractor["imgs_per_gpu"]} and '
+                    f'"samples_per_gpu"={extractor["samples_per_gpu"]}, '
+                    f'"imgs_per_gpu"={extractor["imgs_per_gpu"]} is used in '
+                    f'this experiments')
+            else:
+                logger.warning(
+                    'Automatically set "samples_per_gpu"="imgs_per_gpu"='
+                    f'{extractor["imgs_per_gpu"]} in this experiments')
+            extractor['samples_per_gpu'] = extractor['imgs_per_gpu']
+
         self.extractor = Extractor(dist_mode=dist_mode, **extractor)
         self.clustering_type = clustering.pop('type')
         self.clustering_cfg = clustering

mmselfsup/datasets/builder.py

@@ -1,6 +1,7 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 import platform
 import random
+import warnings
 from functools import partial
 
 import numpy as np
@@ -45,8 +46,9 @@ def build_dataset(cfg, default_args=None):
 
 
 def build_dataloader(dataset,
-                     imgs_per_gpu,
-                     workers_per_gpu,
+                     imgs_per_gpu=None,
+                     samples_per_gpu=None,
+                     workers_per_gpu=1,
                      num_gpus=1,
                      dist=True,
                      shuffle=True,
@@ -62,10 +64,13 @@ def build_dataloader(dataset,
 
     Args:
         dataset (Dataset): A PyTorch dataset.
-        imgs_per_gpu (int): Number of images on each GPU, i.e., batch size of
-            each GPU.
+        imgs_per_gpu (int): (Deprecated, please use samples_per_gpu) Number of
+            images on each GPU, i.e., batch size of each GPU. Defaults to None.
+        samples_per_gpu (int): Number of images on each GPU, i.e., batch size
+            of each GPU. Defaults to None.
         workers_per_gpu (int): How many subprocesses to use for data loading
-            for each GPU.
+            for each GPU. `persistent_workers` option needs num_workers > 0.
+            Defaults to 1.
         num_gpus (int): Number of GPUs. Only used in non-distributed training.
         dist (bool): Distributed training/test or not. Defaults to True.
         shuffle (bool): Whether to shuffle the data at every epoch.
@@ -85,18 +90,33 @@ def build_dataloader(dataset,
     Returns:
         DataLoader: A PyTorch dataloader.
     """
+    if imgs_per_gpu is None and samples_per_gpu is None:
+        raise ValueError(
+            'Please inidcate number of images on each GPU, ',
+            '"imgs_per_gpu" and "samples_per_gpu" can not be "None" at the ',
+            'same time. "imgs_per_gpu" is deprecated, please use ',
+            '"samples_per_gpu".')
+
+    if imgs_per_gpu is not None:
+        warnings.warn(f'Got "imgs_per_gpu"={imgs_per_gpu} and '
+                      f'"samples_per_gpu"={samples_per_gpu}, "imgs_per_gpu"'
+                      f'={imgs_per_gpu} is used in this experiments. '
+                      'Automatically set "samples_per_gpu"="imgs_per_gpu"='
+                      f'{imgs_per_gpu} in this experiments')
+        samples_per_gpu = imgs_per_gpu
+
     rank, world_size = get_dist_info()
     if dist:
         sampler = DistributedSampler(
             dataset, world_size, rank, shuffle=shuffle, replace=replace)
         shuffle = False
-        batch_size = imgs_per_gpu
+        batch_size = samples_per_gpu
         num_workers = workers_per_gpu
     else:
         if replace:
             return NotImplemented
         sampler = None  # TODO: set replace
-        batch_size = num_gpus * imgs_per_gpu
+        batch_size = num_gpus * samples_per_gpu
         num_workers = num_gpus * workers_per_gpu
 
     init_fn = partial(
@@ -117,7 +137,7 @@ def build_dataloader(dataset,
         batch_size=batch_size,
         sampler=sampler,
         num_workers=num_workers,
-        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
+        collate_fn=partial(collate, samples_per_gpu=samples_per_gpu),
         pin_memory=pin_memory,
         shuffle=shuffle,
         worker_init_fn=init_fn,

mmselfsup/utils/extractor.py

@@ -11,8 +11,8 @@ class Extractor(object):
     Args:
         dataset (Dataset | dict): A PyTorch dataset or dict that indicates
             the dataset.
-        imgs_per_gpu (int): Number of images on each GPU, i.e., batch size of
-            each GPU.
+        samples_per_gpu (int): Number of images on each GPU, i.e., batch size
+            of each GPU.
         workers_per_gpu (int): How many subprocesses to use for data loading
             for each GPU.
         dist_mode (bool): Use distributed extraction or not. Defaults to False.
@@ -25,7 +25,7 @@ class Extractor(object):
 
     def __init__(self,
                  dataset,
-                 imgs_per_gpu,
+                 samples_per_gpu,
                  workers_per_gpu,
                  dist_mode=False,
                  persistent_workers=True,
@@ -40,8 +40,8 @@ class Extractor(object):
                             f'not {type(dataset)}')
         self.data_loader = datasets.build_dataloader(
             self.dataset,
-            imgs_per_gpu,
-            workers_per_gpu,
+            samples_per_gpu=samples_per_gpu,
+            workers_per_gpu=workers_per_gpu,
             dist=dist_mode,
             shuffle=False,
             persistent_workers=persistent_workers,

tests/test_data/test_datasets/test_builder.py

@@ -3,7 +3,7 @@ from unittest.mock import ANY
 import pytest
 
 from mmselfsup.datasets import (ConcatDataset, DeepClusterDataset,
-                                RepeatDataset, build_dataset)
+                                RepeatDataset, build_dataloader, build_dataset)
 
 DATASET_CONFIG = dict(
     type='DeepClusterDataset',
@@ -50,3 +50,19 @@ DATASET_CONFIG = dict(
 ])
 def test_build_dataset(cfg, expected_type):
     assert isinstance(build_dataset(cfg), expected_type)
+
+
+def test_build_dataloader():
+    dataset = build_dataset(DATASET_CONFIG)
+
+    with pytest.raises(ValueError):
+        data_loader = build_dataloader(dataset)
+
+    data_loader = build_dataloader(
+        dataset,
+        imgs_per_gpu=1,
+        samples_per_gpu=None,
+        dist=False,
+    )
+    assert len(data_loader) == 2
+    assert data_loader.batch_size == 1

tools/analysis_tools/visualize_tsne.py

@@ -156,9 +156,22 @@ def main():
     dataset.data_source.data_infos = tmp_infos
     logger.info(f'Apply t-SNE to visualize {len(dataset)} samples.')
 
+    if 'imgs_per_gpu' in cfg.data:
+        logger.warning('"imgs_per_gpu" is deprecated. '
+                       'Please use "samples_per_gpu" instead')
+        if 'samples_per_gpu' in cfg.data:
+            logger.warning(
+                f'Got "imgs_per_gpu"={cfg.data.imgs_per_gpu} and '
+                f'"samples_per_gpu"={cfg.data.samples_per_gpu}, "imgs_per_gpu"'
+                f'={cfg.data.imgs_per_gpu} is used in this experiments')
+        else:
+            logger.warning(
+                'Automatically set "samples_per_gpu"="imgs_per_gpu"='
+                f'{cfg.data.imgs_per_gpu} in this experiments')
+        cfg.data.samples_per_gpu = cfg.data.imgs_per_gpu
     data_loader = build_dataloader(
         dataset,
-        imgs_per_gpu=dataset_cfg.data.imgs_per_gpu,
+        samples_per_gpu=dataset_cfg.data.samples_per_gpu,
         workers_per_gpu=dataset_cfg.data.workers_per_gpu,
         dist=distributed,
         shuffle=False)

tools/benchmarks/classification/dist_train_linear.sh

@@ -6,7 +6,7 @@ set -x
 CFG=$1  # use cfgs under "configs/benchmarks/classification/imagenet/*.py"
 PRETRAIN=$2  # pretrained model
 PY_ARGS=${@:3}
-GPUS=${GPUS:-8}  # When changing GPUS, please also change imgs_per_gpu in the config file accordingly to ensure the total batch size is 256.
+GPUS=${GPUS:-8}  # When changing GPUS, please also change samples_per_gpu in the config file accordingly to ensure the total batch size is 256.
 PORT=${PORT:-29500}
 
 # set work_dir according to config path and pretrained model to distinguish different models

tools/benchmarks/classification/slurm_train_linear.sh

@@ -8,7 +8,7 @@ JOB_NAME=$2
 CFG=$3  # use cfgs under "configs/benchmarks/classification/imagenet/*.py"
 PRETRAIN=$4  # pretrained model
 PY_ARGS=${@:5}
-GPUS=${GPUS:-8}  # When changing GPUS, please also change imgs_per_gpu in the config file accordingly to ensure the total batch size is 256.
+GPUS=${GPUS:-8}  # When changing GPUS, please also change samples_per_gpu in the config file accordingly to ensure the total batch size is 256.
 GPUS_PER_NODE=${GPUS_PER_NODE:-8}
 CPUS_PER_TASK=${CPUS_PER_TASK:-5}
 PORT=${PORT:-29500}

tools/benchmarks/classification/svm_voc07/extract.py

@@ -96,9 +96,23 @@ def main():
     # build the dataloader
     dataset_cfg = mmcv.Config.fromfile(args.dataset_config)
     dataset = build_dataset(dataset_cfg.data.extract)
+    if 'imgs_per_gpu' in cfg.data:
+        logger.warning('"imgs_per_gpu" is deprecated. '
+                       'Please use "samples_per_gpu" instead')
+        if 'samples_per_gpu' in cfg.data:
+            logger.warning(
+                f'Got "imgs_per_gpu"={cfg.data.imgs_per_gpu} and '
+                f'"samples_per_gpu"={cfg.data.samples_per_gpu}, "imgs_per_gpu"'
+                f'={cfg.data.imgs_per_gpu} is used in this experiments')
+        else:
+            logger.warning(
+                'Automatically set "samples_per_gpu"="imgs_per_gpu"='
+                f'{cfg.data.imgs_per_gpu} in this experiments')
+        cfg.data.samples_per_gpu = cfg.data.imgs_per_gpu
+
     data_loader = build_dataloader(
         dataset,
-        imgs_per_gpu=dataset_cfg.data.imgs_per_gpu,
+        samples_per_gpu=dataset_cfg.data.samples_per_gpu,
         workers_per_gpu=dataset_cfg.data.workers_per_gpu,
         dist=distributed,
         shuffle=False)

tools/test.py

@@ -96,9 +96,22 @@ def main():
 
     # build the dataloader
     dataset = build_dataset(cfg.data.val)
+    if 'imgs_per_gpu' in cfg.data:
+        logger.warning('"imgs_per_gpu" is deprecated. '
+                       'Please use "samples_per_gpu" instead')
+        if 'samples_per_gpu' in cfg.data:
+            logger.warning(
+                f'Got "imgs_per_gpu"={cfg.data.imgs_per_gpu} and '
+                f'"samples_per_gpu"={cfg.data.samples_per_gpu}, "imgs_per_gpu"'
+                f'={cfg.data.imgs_per_gpu} is used in this experiments')
+        else:
+            logger.warning(
+                'Automatically set "samples_per_gpu"="imgs_per_gpu"='
+                f'{cfg.data.imgs_per_gpu} in this experiments')
+        cfg.data.samples_per_gpu = cfg.data.imgs_per_gpu
     data_loader = build_dataloader(
         dataset,
-        imgs_per_gpu=cfg.data.imgs_per_gpu,
+        samples_per_gpu=cfg.data.samples_per_gpu,
         workers_per_gpu=cfg.data.workers_per_gpu,
         dist=distributed,
         shuffle=False)