weight = 'weights/volt-small-scannet.pth'
resume = False
evaluate = True
test_only = False
seed = 5793371
save_path = 'exp/scannet/2026-07-23_152153'
num_worker = 16
batch_size = 8
gradient_accumulation_steps = 1
batch_size_val = None
batch_size_test = None
epoch = 75
eval_epoch = 75
clip_grad = 10.0
use_ema = False
ema_decay = 0.999
dataset_ratios = None
sync_bn = False
enable_amp = True
amp_dtype = 'float16'
empty_cache = False
empty_cache_per_epoch = False
find_unused_parameters = True
enable_wandb = True
wandb_project = 'Volt'
wandb_key = None
mix_prob = 0.8
param_dicts = None
hooks = [
    dict(
        type='CheckpointLoader',
        keywords='module.backbone',
        replacement='module.backbone'),
    dict(type='IterationTimer', warmup_iter=2),
    dict(type='InformationWriter'),
    dict(
        type='InsSegEvaluator',
        segment_ignore_index=(-1, 0, 1),
        instance_ignore_index=-1),
    dict(type='CheckpointSaver', save_freq=None),
    dict(type='PreciseEvaluator', test_last=False)
]
train = dict(type='DefaultTrainer')
test = dict(
    type='InsSegTester',
    verbose=True,
    segment_ignore_index=(-1, 0, 1),
    instance_ignore_index=-1)
CLASS_LABELS_20 = ('wall', 'floor', 'cabinet', 'bed', 'chair', 'sofa', 'table',
                   'door', 'window', 'bookshelf', 'picture', 'counter', 'desk',
                   'curtain', 'refrigerator', 'shower curtain', 'toilet',
                   'sink', 'bathtub', 'otherfurniture')
VALID_CLASS_IDS_20 = (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 24, 28,
                      33, 34, 36, 39)
class_names = ('wall', 'floor', 'cabinet', 'bed', 'chair', 'sofa', 'table',
               'door', 'window', 'bookshelf', 'picture', 'counter', 'desk',
               'curtain', 'refrigerator', 'shower curtain', 'toilet', 'sink',
               'bathtub', 'otherfurniture')
class_ids = (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 24, 28, 33, 34, 36,
             39)
num_classes = 20
segment_ignore_index = (-1, 0, 1)
semantic_num_classes = 18
num_channels = 128
model = dict(
    type='SPFormer-v1m1',
    backbone=dict(
        type='Volt',
        in_channels=6,
        embed_dim=384,
        depth=12,
        num_heads=6,
        mlp_ratio=4,
        init_values=None,
        qk_norm=True,
        drop_path=0.3,
        stride=5,
        kernel_size=5,
        increase_drop_path=True,
        up_mlp_dim=128),
    decoder=dict(
        type='SPFormerDecoder',
        num_class=18,
        in_channel=128,
        num_layer=6,
        num_query=400,
        d_model=384,
        nhead=8,
        hidden_dim=1024,
        dropout=0.0,
        activation_fn='gelu',
        iter_pred=True,
        attn_mask=True,
        use_query_pos=False,
        use_score=False,
        use_param_query=True),
    criterion=dict(
        type='SPFormerCriterion',
        matcher=dict(
            type='SPFormerHungarianMatcher',
            costs=[
                dict(type='SPFormerQueryClassificationCost', weight=0.5),
                dict(type='SPFormerMaskBCECost', weight=1.0),
                dict(type='SPFormerMaskDiceCost', weight=1.0)
            ]),
        loss_weight=[0.2, 1.0, 1.0, 0.5],
        num_classes=18,
        non_object_weight=0.1,
        fix_dice_loss_weight=False,
        iter_matcher=True,
        fix_mean_loss=True),
    semantic_num_classes=18,
    semantic_ignore_index=-1,
    segment_ignore_index=(-1, 0, 1),
    instance_ignore_index=-1,
    topk_insts=100,
    score_thr=0.0,
    npoint_thr=100,
    nms=True)
optimizer = dict(type='AdamW', lr=0.0003, weight_decay=0.1)
scheduler = dict(type='PolyLR')
dataset_type = 'ScanNetDataset'
data_root = 'data/scannet'
data = dict(
    num_classes=20,
    ignore_index=-1,
    names=('wall', 'floor', 'cabinet', 'bed', 'chair', 'sofa', 'table', 'door',
           'window', 'bookshelf', 'picture', 'counter', 'desk', 'curtain',
           'refrigerator', 'shower curtain', 'toilet', 'sink', 'bathtub',
           'otherfurniture'),
    ids=(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 24, 28, 33, 34, 36,
         39),
    train=dict(
        type='ScanNetDataset',
        split='train',
        data_root='data/scannet',
        transform=[
            dict(type='CenterShift', apply_z=True),
            dict(
                type='RandomDropout',
                dropout_ratio=0.2,
                dropout_application_ratio=0.5),
            dict(type='RandomFlip', p=0.5),
            dict(
                type='RandomRotate',
                angle=[-1, 1],
                axis='z',
                center=[0, 0, 0],
                p=0.95),
            dict(
                type='RandomRotate',
                angle=[-0.041666666666666664, 0.041666666666666664],
                axis='x',
                p=0.95),
            dict(
                type='RandomRotate',
                angle=[-0.041666666666666664, 0.041666666666666664],
                axis='y',
                p=0.95),
            dict(type='RandomScale', scale=[0.8, 1.2]),
            dict(
                type='RandomShift',
                shift=[(-0.5, 0.5), (-0.5, 0.5), (-0.5, 0.5)]),
            dict(
                type='ElasticDistortion',
                distortion_params=[[0.2, 0.4], [0.8, 1.6]]),
            dict(type='ChromaticAutoContrast', p=0.5, blend_factor=None),
            dict(type='ChromaticTranslation', p=0.95, ratio=0.1),
            dict(type='ChromaticJitter', p=0.95, std=0.05),
            dict(type='SphereCrop', sample_rate=0.6, mode='random'),
            dict(type='SphereCrop', point_max=102400, mode='random'),
            dict(
                type='Copy',
                keys_dict=dict(
                    coord='origin_coord',
                    instance='origin_instance',
                    segment='origin_segment')),
            dict(
                type='Update',
                keys_dict=dict(index_valid_keys=[
                    'coord', 'color', 'normal', 'strength', 'segment',
                    'instance', 'image_coord', 'image_mask'
                ])),
            dict(
                type='GridSample',
                grid_size=0.02,
                hash_type='fnv',
                mode='train',
                return_grid_coord=True,
                return_inverse=True),
            dict(type='NormalizeColor'),
            dict(
                type='InstanceParser',
                segment_ignore_index=(-1, 0, 1),
                instance_ignore_index=-1),
            dict(type='ToTensor'),
            dict(
                type='Collect',
                keys=('coord', 'origin_coord', 'grid_coord', 'segment',
                      'origin_segment', 'instance', 'origin_instance',
                      'superpoint', 'inverse'),
                feat_keys=('color', 'normal'),
                offset_keys_dict=dict(
                    offset='coord', origin_offset='origin_coord'))
        ],
        test_mode=False,
        loop=1),
    val=dict(
        type='ScanNetDataset',
        split='val',
        data_root='data/scannet',
        transform=[
            dict(type='MeanShift'),
            dict(
                type='RandomRotate',
                angle=[0.35, 0.35],
                axis='z',
                center=[0, 0, 0],
                always_apply=True),
            dict(
                type='Copy',
                keys_dict=dict(
                    coord='origin_coord',
                    instance='origin_instance',
                    segment='origin_segment')),
            dict(
                type='Update',
                keys_dict=dict(index_valid_keys=[
                    'coord', 'color', 'normal', 'strength', 'segment',
                    'instance', 'image_coord', 'image_mask'
                ])),
            dict(
                type='GridSample',
                grid_size=0.02,
                hash_type='fnv',
                mode='train',
                deterministic=True,
                return_grid_coord=True,
                return_inverse=True),
            dict(type='NormalizeColor'),
            dict(
                type='InstanceParser',
                segment_ignore_index=(-1, 0, 1),
                instance_ignore_index=-1),
            dict(type='ToTensor'),
            dict(
                type='Collect',
                keys=('coord', 'origin_coord', 'grid_coord', 'segment',
                      'origin_segment', 'instance', 'origin_instance',
                      'superpoint', 'inverse', 'name'),
                feat_keys=('color', 'normal'),
                offset_keys_dict=dict(
                    offset='coord', origin_offset='origin_coord'))
        ],
        test_mode=False),
    test=dict(
        type='ScanNetDataset',
        split='val',
        data_root='data/scannet',
        transform=[
            dict(type='MeanShift'),
            dict(
                type='RandomRotate',
                angle=[0.35, 0.35],
                axis='z',
                center=[0, 0, 0],
                always_apply=True),
            dict(
                type='Copy',
                keys_dict=dict(
                    coord='origin_coord',
                    instance='origin_instance',
                    segment='origin_segment')),
            dict(
                type='Update',
                keys_dict=dict(index_valid_keys=[
                    'coord', 'color', 'normal', 'strength', 'segment',
                    'instance', 'image_coord', 'image_mask'
                ])),
            dict(
                type='GridSample',
                grid_size=0.02,
                hash_type='fnv',
                mode='train',
                deterministic=True,
                return_grid_coord=True,
                return_inverse=True),
            dict(type='NormalizeColor'),
            dict(
                type='InstanceParser',
                segment_ignore_index=(-1, 0, 1),
                instance_ignore_index=-1),
            dict(type='ToTensor'),
            dict(
                type='Collect',
                keys=('coord', 'origin_coord', 'grid_coord', 'segment',
                      'origin_segment', 'instance', 'origin_instance',
                      'superpoint', 'inverse', 'name'),
                feat_keys=('color', 'normal'),
                offset_keys_dict=dict(
                    offset='coord', origin_offset='origin_coord'))
        ],
        test_mode=False))
num_worker_per_gpu = 8
batch_size_per_gpu = 4
batch_size_val_per_gpu = 1
batch_size_test_per_gpu = 1