_base_ = '../../../_base_/default_runtime.py'

model = dict(
    type='Recognizer3D',
    backbone=dict(
        type='ResNet3dSlowOnly',
        depth=50,
        conv1_kernel=(1, 7, 7),
        inflate=(0, 0, 1, 1)),
    cls_head=dict(
        type='I3DHead',
        in_channels=2048,
        num_classes=60,
        dropout_ratio=0.5,
        average_clips='prob'),
    data_preprocessor=dict(
        type='ActionDataPreprocessor',
        mean=[123.675, 116.28, 103.53],
        std=[58.395, 57.12, 57.375],
        format_shape='NCTHW'))

dataset_type = 'PoseDataset'
data_root = 'data/nturgbd_videos/'
ann_file = 'data/skeleton/ntu60_2d.pkl'

train_pipeline = [
    dict(type='MMUniformSampleFrames', clip_len=dict(RGB=8), num_clips=1),
    dict(type='MMDecode'),
    dict(type='MMCompact', hw_ratio=1., allow_imgpad=True),
    dict(type='Resize', scale=(256, 256), keep_ratio=False),
    dict(type='RandomResizedCrop', area_range=(0.56, 1.0)),
    dict(type='Resize', scale=(224, 224), keep_ratio=False),
    dict(type='Flip', flip_ratio=0.5),
    dict(type='FormatShape', input_format='NCTHW'),
    dict(type='PackActionInputs')
]
val_pipeline = [
    dict(
        type='MMUniformSampleFrames',
        clip_len=dict(RGB=8),
        num_clips=1,
        test_mode=True),
    dict(type='MMDecode'),
    dict(type='MMCompact', hw_ratio=1., allow_imgpad=True),
    dict(type='Resize', scale=(224, 224), keep_ratio=False),
    dict(type='FormatShape', input_format='NCTHW'),
    dict(type='PackActionInputs')
]
test_pipeline = [
    dict(
        type='MMUniformSampleFrames',
        clip_len=dict(RGB=8),
        num_clips=10,
        test_mode=True),
    dict(type='MMDecode'),
    dict(type='MMCompact', hw_ratio=1., allow_imgpad=True),
    dict(type='Resize', scale=(224, 224), keep_ratio=False),
    dict(type='FormatShape', input_format='NCTHW'),
    dict(type='PackActionInputs')
]

train_dataloader = dict(
    batch_size=12,
    num_workers=8,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=True),
    dataset=dict(
        type='RepeatDataset',
        times=10,
        dataset=dict(
            type=dataset_type,
            ann_file=ann_file,
            data_prefix=dict(video=data_root),
            split='xsub_train',
            pipeline=train_pipeline)))
val_dataloader = dict(
    batch_size=12,
    num_workers=8,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type=dataset_type,
        ann_file=ann_file,
        data_prefix=dict(video=data_root),
        split='xsub_val',
        pipeline=val_pipeline,
        test_mode=True))
test_dataloader = dict(
    batch_size=1,
    num_workers=8,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type=dataset_type,
        ann_file=ann_file,
        data_prefix=dict(video=data_root),
        split='xsub_val',
        pipeline=test_pipeline,
        test_mode=True))

val_evaluator = [dict(type='AccMetric')]
test_evaluator = val_evaluator

train_cfg = dict(
    type='EpochBasedTrainLoop', max_epochs=18, val_begin=1, val_interval=1)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')

param_scheduler = [
    dict(
        type='CosineAnnealingLR',
        eta_min=0,
        T_max=18,
        by_epoch=True,
        convert_to_iter_based=True)
]

optim_wrapper = dict(
    optimizer=dict(type='SGD', lr=0.15, momentum=0.9, weight_decay=0.0001),
    clip_grad=dict(max_norm=40, norm_type=2))

# Default setting for scaling LR automatically
#   - `enable` means enable scaling LR automatically
#       or not by default.
#   - `base_batch_size` = (8 GPUs) x (12 samples per GPU).
auto_scale_lr = dict(enable=False, base_batch_size=96)