2023-03-03 13:59:33,312 - mmseg - INFO - Multi-processing start method is `None`
2023-03-03 13:59:33,327 - mmseg - INFO - OpenCV num_threads is `128
2023-03-03 13:59:33,327 - mmseg - INFO - OMP num threads is 1
2023-03-03 13:59:33,410 - mmseg - INFO - Environment info:
------------------------------------------------------------
sys.platform: linux
Python: 3.7.16 (default, Jan 17 2023, 22:20:44) [GCC 11.2.0]
CUDA available: True
GPU 0,1,2,3,4,5,6,7: NVIDIA A100-SXM4-80GB
CUDA_HOME: /mnt/petrelfs/laizeqiang/miniconda3/envs/torch
NVCC: Cuda compilation tools, release 11.6, V11.6.124
GCC: gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44)
PyTorch: 1.13.1
PyTorch compiling details: PyTorch built with:
  - GCC 9.3
  - C++ Version: 201402
  - Intel(R) oneAPI Math Kernel Library Version 2021.4-Product Build 20210904 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v2.6.0 (Git Hash 52b5f107dd9cf10910aaa19cb47f3abf9b349815)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - LAPACK is enabled (usually provided by MKL)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 11.6
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_61,code=sm_61;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86;-gencode;arch=compute_37,code=compute_37
  - CuDNN 8.3.2  (built against CUDA 11.5)
  - Magma 2.6.1
  - Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.6, CUDNN_VERSION=8.3.2, CXX_COMPILER=/opt/rh/devtoolset-9/root/usr/bin/c++, CXX_FLAGS= -fabi-version=11 -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_KINETO -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -DEDGE_PROFILER_USE_KINETO -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wunused-local-typedefs -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=1.13.1, USE_CUDA=ON, USE_CUDNN=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, USE_ROCM=OFF, 

TorchVision: 0.14.1
OpenCV: 4.7.0
MMCV: 1.7.1
MMCV Compiler: GCC 9.3
MMCV CUDA Compiler: 11.6
MMSegmentation: 0.30.0+ad87029
------------------------------------------------------------

2023-03-03 13:59:33,411 - mmseg - INFO - Distributed training: True
2023-03-03 13:59:34,043 - mmseg - INFO - Config:
norm_cfg = dict(type='SyncBN', requires_grad=True)
checkpoint = 'pretrained/segformer_mit-b2_512x512_160k_ade20k_20220620_114047-64e4feca.pth'
model = dict(
    type='EncoderDecoderFreeze',
    freeze_parameters=['backbone', 'decode_head'],
    pretrained=
    'pretrained/segformer_mit-b2_512x512_160k_ade20k_20220620_114047-64e4feca.pth',
    backbone=dict(
        type='MixVisionTransformerCustomInitWeights',
        in_channels=3,
        embed_dims=64,
        num_stages=4,
        num_layers=[3, 4, 6, 3],
        num_heads=[1, 2, 5, 8],
        patch_sizes=[7, 3, 3, 3],
        sr_ratios=[8, 4, 2, 1],
        out_indices=(0, 1, 2, 3),
        mlp_ratio=4,
        qkv_bias=True,
        drop_rate=0.0,
        attn_drop_rate=0.0,
        drop_path_rate=0.1),
    decode_head=dict(
        type='SegformerHeadUnetFCHeadSingleStep',
        pretrained=
        'pretrained/segformer_mit-b2_512x512_160k_ade20k_20220620_114047-64e4feca.pth',
        dim=128,
        out_dim=256,
        unet_channels=272,
        dim_mults=[1, 1, 1],
        cat_embedding_dim=16,
        in_channels=[64, 128, 320, 512],
        in_index=[0, 1, 2, 3],
        channels=256,
        dropout_ratio=0.1,
        num_classes=151,
        norm_cfg=dict(type='SyncBN', requires_grad=True),
        align_corners=False,
        ignore_index=0,
        loss_decode=dict(
            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
    train_cfg=dict(),
    test_cfg=dict(mode='whole'))
dataset_type = 'ADE20K151Dataset'
data_root = 'data/ade/ADEChallengeData2016'
img_norm_cfg = dict(
    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
crop_size = (512, 512)
train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations', reduce_zero_label=False),
    dict(type='Resize', img_scale=(2048, 512), ratio_range=(0.5, 2.0)),
    dict(type='RandomCrop', crop_size=(512, 512), cat_max_ratio=0.75),
    dict(type='RandomFlip', prob=0.5),
    dict(type='PhotoMetricDistortion'),
    dict(
        type='Normalize',
        mean=[123.675, 116.28, 103.53],
        std=[58.395, 57.12, 57.375],
        to_rgb=True),
    dict(type='Pad', size=(512, 512), pad_val=0, seg_pad_val=0),
    dict(type='DefaultFormatBundle'),
    dict(type='Collect', keys=['img', 'gt_semantic_seg'])
]
test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(
        type='MultiScaleFlipAug',
        img_scale=(2048, 512),
        flip=False,
        transforms=[
            dict(type='Resize', keep_ratio=True),
            dict(type='RandomFlip'),
            dict(
                type='Normalize',
                mean=[123.675, 116.28, 103.53],
                std=[58.395, 57.12, 57.375],
                to_rgb=True),
            dict(type='Pad', size_divisor=16, pad_val=0, seg_pad_val=0),
            dict(type='ImageToTensor', keys=['img']),
            dict(type='Collect', keys=['img'])
        ])
]
data = dict(
    samples_per_gpu=4,
    workers_per_gpu=4,
    train=dict(
        type='ADE20K151Dataset',
        data_root='data/ade/ADEChallengeData2016',
        img_dir='images/training',
        ann_dir='annotations/training',
        pipeline=[
            dict(type='LoadImageFromFile'),
            dict(type='LoadAnnotations', reduce_zero_label=False),
            dict(type='Resize', img_scale=(2048, 512), ratio_range=(0.5, 2.0)),
            dict(type='RandomCrop', crop_size=(512, 512), cat_max_ratio=0.75),
            dict(type='RandomFlip', prob=0.5),
            dict(type='PhotoMetricDistortion'),
            dict(
                type='Normalize',
                mean=[123.675, 116.28, 103.53],
                std=[58.395, 57.12, 57.375],
                to_rgb=True),
            dict(type='Pad', size=(512, 512), pad_val=0, seg_pad_val=0),
            dict(type='DefaultFormatBundle'),
            dict(type='Collect', keys=['img', 'gt_semantic_seg'])
        ]),
    val=dict(
        type='ADE20K151Dataset',
        data_root='data/ade/ADEChallengeData2016',
        img_dir='images/validation',
        ann_dir='annotations/validation',
        pipeline=[
            dict(type='LoadImageFromFile'),
            dict(
                type='MultiScaleFlipAug',
                img_scale=(2048, 512),
                flip=False,
                transforms=[
                    dict(type='Resize', keep_ratio=True),
                    dict(type='RandomFlip'),
                    dict(
                        type='Normalize',
                        mean=[123.675, 116.28, 103.53],
                        std=[58.395, 57.12, 57.375],
                        to_rgb=True),
                    dict(
                        type='Pad', size_divisor=16, pad_val=0, seg_pad_val=0),
                    dict(type='ImageToTensor', keys=['img']),
                    dict(type='Collect', keys=['img'])
                ])
        ]),
    test=dict(
        type='ADE20K151Dataset',
        data_root='data/ade/ADEChallengeData2016',
        img_dir='images/validation',
        ann_dir='annotations/validation',
        pipeline=[
            dict(type='LoadImageFromFile'),
            dict(
                type='MultiScaleFlipAug',
                img_scale=(2048, 512),
                flip=False,
                transforms=[
                    dict(type='Resize', keep_ratio=True),
                    dict(type='RandomFlip'),
                    dict(
                        type='Normalize',
                        mean=[123.675, 116.28, 103.53],
                        std=[58.395, 57.12, 57.375],
                        to_rgb=True),
                    dict(
                        type='Pad', size_divisor=16, pad_val=0, seg_pad_val=0),
                    dict(type='ImageToTensor', keys=['img']),
                    dict(type='Collect', keys=['img'])
                ])
        ]))
log_config = dict(
    interval=50, hooks=[dict(type='TextLoggerHook', by_epoch=False)])
dist_params = dict(backend='nccl')
log_level = 'INFO'
load_from = None
resume_from = None
workflow = [('train', 1)]
cudnn_benchmark = True
optimizer = dict(
    type='AdamW', lr=0.00015, betas=[0.9, 0.96], weight_decay=0.045)
optimizer_config = dict()
lr_config = dict(
    policy='step',
    warmup='linear',
    warmup_iters=1000,
    warmup_ratio=1e-06,
    step=10000,
    gamma=0.5,
    min_lr=1e-06,
    by_epoch=False)
runner = dict(type='IterBasedRunner', max_iters=80000)
checkpoint_config = dict(by_epoch=False, interval=8000)
evaluation = dict(
    interval=8000, metric='mIoU', pre_eval=True, save_best='mIoU')
work_dir = './work_dirs/segformer_mit_b2_segformer_head_unet_fc_single_step_ade_pretrained_freeze_embed_80k_ade20k151'
gpu_ids = range(0, 8)
auto_resume = True

2023-03-03 13:59:38,432 - mmseg - INFO - Set random seed to 97773280, deterministic: False
2023-03-03 13:59:38,757 - mmseg - INFO - Parameters in backbone freezed!
2023-03-03 13:59:38,758 - mmseg - INFO - Trainable parameters in SegformerHeadUnetFCHeadSingleStep: ['unet.init_conv.weight', 'unet.init_conv.bias', 'unet.time_mlp.1.weight', 'unet.time_mlp.1.bias', 'unet.time_mlp.3.weight', 'unet.time_mlp.3.bias', 'unet.downs.0.0.mlp.1.weight', 'unet.downs.0.0.mlp.1.bias', 'unet.downs.0.0.block1.proj.weight', 'unet.downs.0.0.block1.proj.bias', 'unet.downs.0.0.block1.norm.weight', 'unet.downs.0.0.block1.norm.bias', 'unet.downs.0.0.block2.proj.weight', 'unet.downs.0.0.block2.proj.bias', 'unet.downs.0.0.block2.norm.weight', 'unet.downs.0.0.block2.norm.bias', 'unet.downs.0.1.mlp.1.weight', 'unet.downs.0.1.mlp.1.bias', 'unet.downs.0.1.block1.proj.weight', 'unet.downs.0.1.block1.proj.bias', 'unet.downs.0.1.block1.norm.weight', 'unet.downs.0.1.block1.norm.bias', 'unet.downs.0.1.block2.proj.weight', 'unet.downs.0.1.block2.proj.bias', 'unet.downs.0.1.block2.norm.weight', 'unet.downs.0.1.block2.norm.bias', 'unet.downs.0.2.fn.fn.to_qkv.weight', 'unet.downs.0.2.fn.fn.to_out.0.weight', 'unet.downs.0.2.fn.fn.to_out.0.bias', 'unet.downs.0.2.fn.fn.to_out.1.g', 'unet.downs.0.2.fn.norm.g', 'unet.downs.0.3.weight', 'unet.downs.0.3.bias', 'unet.downs.1.0.mlp.1.weight', 'unet.downs.1.0.mlp.1.bias', 'unet.downs.1.0.block1.proj.weight', 'unet.downs.1.0.block1.proj.bias', 'unet.downs.1.0.block1.norm.weight', 'unet.downs.1.0.block1.norm.bias', 'unet.downs.1.0.block2.proj.weight', 'unet.downs.1.0.block2.proj.bias', 'unet.downs.1.0.block2.norm.weight', 'unet.downs.1.0.block2.norm.bias', 'unet.downs.1.1.mlp.1.weight', 'unet.downs.1.1.mlp.1.bias', 'unet.downs.1.1.block1.proj.weight', 'unet.downs.1.1.block1.proj.bias', 'unet.downs.1.1.block1.norm.weight', 'unet.downs.1.1.block1.norm.bias', 'unet.downs.1.1.block2.proj.weight', 'unet.downs.1.1.block2.proj.bias', 'unet.downs.1.1.block2.norm.weight', 'unet.downs.1.1.block2.norm.bias', 'unet.downs.1.2.fn.fn.to_qkv.weight', 'unet.downs.1.2.fn.fn.to_out.0.weight', 'unet.downs.1.2.fn.fn.to_out.0.bias', 'unet.downs.1.2.fn.fn.to_out.1.g', 'unet.downs.1.2.fn.norm.g', 'unet.downs.1.3.weight', 'unet.downs.1.3.bias', 'unet.downs.2.0.mlp.1.weight', 'unet.downs.2.0.mlp.1.bias', 'unet.downs.2.0.block1.proj.weight', 'unet.downs.2.0.block1.proj.bias', 'unet.downs.2.0.block1.norm.weight', 'unet.downs.2.0.block1.norm.bias', 'unet.downs.2.0.block2.proj.weight', 'unet.downs.2.0.block2.proj.bias', 'unet.downs.2.0.block2.norm.weight', 'unet.downs.2.0.block2.norm.bias', 'unet.downs.2.1.mlp.1.weight', 'unet.downs.2.1.mlp.1.bias', 'unet.downs.2.1.block1.proj.weight', 'unet.downs.2.1.block1.proj.bias', 'unet.downs.2.1.block1.norm.weight', 'unet.downs.2.1.block1.norm.bias', 'unet.downs.2.1.block2.proj.weight', 'unet.downs.2.1.block2.proj.bias', 'unet.downs.2.1.block2.norm.weight', 'unet.downs.2.1.block2.norm.bias', 'unet.downs.2.2.fn.fn.to_qkv.weight', 'unet.downs.2.2.fn.fn.to_out.0.weight', 'unet.downs.2.2.fn.fn.to_out.0.bias', 'unet.downs.2.2.fn.fn.to_out.1.g', 'unet.downs.2.2.fn.norm.g', 'unet.downs.2.3.weight', 'unet.downs.2.3.bias', 'unet.ups.0.0.mlp.1.weight', 'unet.ups.0.0.mlp.1.bias', 'unet.ups.0.0.block1.proj.weight', 'unet.ups.0.0.block1.proj.bias', 'unet.ups.0.0.block1.norm.weight', 'unet.ups.0.0.block1.norm.bias', 'unet.ups.0.0.block2.proj.weight', 'unet.ups.0.0.block2.proj.bias', 'unet.ups.0.0.block2.norm.weight', 'unet.ups.0.0.block2.norm.bias', 'unet.ups.0.0.res_conv.weight', 'unet.ups.0.0.res_conv.bias', 'unet.ups.0.1.mlp.1.weight', 'unet.ups.0.1.mlp.1.bias', 'unet.ups.0.1.block1.proj.weight', 'unet.ups.0.1.block1.proj.bias', 'unet.ups.0.1.block1.norm.weight', 'unet.ups.0.1.block1.norm.bias', 'unet.ups.0.1.block2.proj.weight', 'unet.ups.0.1.block2.proj.bias', 'unet.ups.0.1.block2.norm.weight', 'unet.ups.0.1.block2.norm.bias', 'unet.ups.0.1.res_conv.weight', 'unet.ups.0.1.res_conv.bias', 'unet.ups.0.2.fn.fn.to_qkv.weight', 'unet.ups.0.2.fn.fn.to_out.0.weight', 'unet.ups.0.2.fn.fn.to_out.0.bias', 'unet.ups.0.2.fn.fn.to_out.1.g', 'unet.ups.0.2.fn.norm.g', 'unet.ups.0.3.1.weight', 'unet.ups.0.3.1.bias', 'unet.ups.1.0.mlp.1.weight', 'unet.ups.1.0.mlp.1.bias', 'unet.ups.1.0.block1.proj.weight', 'unet.ups.1.0.block1.proj.bias', 'unet.ups.1.0.block1.norm.weight', 'unet.ups.1.0.block1.norm.bias', 'unet.ups.1.0.block2.proj.weight', 'unet.ups.1.0.block2.proj.bias', 'unet.ups.1.0.block2.norm.weight', 'unet.ups.1.0.block2.norm.bias', 'unet.ups.1.0.res_conv.weight', 'unet.ups.1.0.res_conv.bias', 'unet.ups.1.1.mlp.1.weight', 'unet.ups.1.1.mlp.1.bias', 'unet.ups.1.1.block1.proj.weight', 'unet.ups.1.1.block1.proj.bias', 'unet.ups.1.1.block1.norm.weight', 'unet.ups.1.1.block1.norm.bias', 'unet.ups.1.1.block2.proj.weight', 'unet.ups.1.1.block2.proj.bias', 'unet.ups.1.1.block2.norm.weight', 'unet.ups.1.1.block2.norm.bias', 'unet.ups.1.1.res_conv.weight', 'unet.ups.1.1.res_conv.bias', 'unet.ups.1.2.fn.fn.to_qkv.weight', 'unet.ups.1.2.fn.fn.to_out.0.weight', 'unet.ups.1.2.fn.fn.to_out.0.bias', 'unet.ups.1.2.fn.fn.to_out.1.g', 'unet.ups.1.2.fn.norm.g', 'unet.ups.1.3.1.weight', 'unet.ups.1.3.1.bias', 'unet.ups.2.0.mlp.1.weight', 'unet.ups.2.0.mlp.1.bias', 'unet.ups.2.0.block1.proj.weight', 'unet.ups.2.0.block1.proj.bias', 'unet.ups.2.0.block1.norm.weight', 'unet.ups.2.0.block1.norm.bias', 'unet.ups.2.0.block2.proj.weight', 'unet.ups.2.0.block2.proj.bias', 'unet.ups.2.0.block2.norm.weight', 'unet.ups.2.0.block2.norm.bias', 'unet.ups.2.0.res_conv.weight', 'unet.ups.2.0.res_conv.bias', 'unet.ups.2.1.mlp.1.weight', 'unet.ups.2.1.mlp.1.bias', 'unet.ups.2.1.block1.proj.weight', 'unet.ups.2.1.block1.proj.bias', 'unet.ups.2.1.block1.norm.weight', 'unet.ups.2.1.block1.norm.bias', 'unet.ups.2.1.block2.proj.weight', 'unet.ups.2.1.block2.proj.bias', 'unet.ups.2.1.block2.norm.weight', 'unet.ups.2.1.block2.norm.bias', 'unet.ups.2.1.res_conv.weight', 'unet.ups.2.1.res_conv.bias', 'unet.ups.2.2.fn.fn.to_qkv.weight', 'unet.ups.2.2.fn.fn.to_out.0.weight', 'unet.ups.2.2.fn.fn.to_out.0.bias', 'unet.ups.2.2.fn.fn.to_out.1.g', 'unet.ups.2.2.fn.norm.g', 'unet.ups.2.3.weight', 'unet.ups.2.3.bias', 'unet.mid_block1.mlp.1.weight', 'unet.mid_block1.mlp.1.bias', 'unet.mid_block1.block1.proj.weight', 'unet.mid_block1.block1.proj.bias', 'unet.mid_block1.block1.norm.weight', 'unet.mid_block1.block1.norm.bias', 'unet.mid_block1.block2.proj.weight', 'unet.mid_block1.block2.proj.bias', 'unet.mid_block1.block2.norm.weight', 'unet.mid_block1.block2.norm.bias', 'unet.mid_attn.fn.fn.to_qkv.weight', 'unet.mid_attn.fn.fn.to_out.weight', 'unet.mid_attn.fn.fn.to_out.bias', 'unet.mid_attn.fn.norm.g', 'unet.mid_block2.mlp.1.weight', 'unet.mid_block2.mlp.1.bias', 'unet.mid_block2.block1.proj.weight', 'unet.mid_block2.block1.proj.bias', 'unet.mid_block2.block1.norm.weight', 'unet.mid_block2.block1.norm.bias', 'unet.mid_block2.block2.proj.weight', 'unet.mid_block2.block2.proj.bias', 'unet.mid_block2.block2.norm.weight', 'unet.mid_block2.block2.norm.bias', 'unet.final_res_block.mlp.1.weight', 'unet.final_res_block.mlp.1.bias', 'unet.final_res_block.block1.proj.weight', 'unet.final_res_block.block1.proj.bias', 'unet.final_res_block.block1.norm.weight', 'unet.final_res_block.block1.norm.bias', 'unet.final_res_block.block2.proj.weight', 'unet.final_res_block.block2.proj.bias', 'unet.final_res_block.block2.norm.weight', 'unet.final_res_block.block2.norm.bias', 'unet.final_res_block.res_conv.weight', 'unet.final_res_block.res_conv.bias', 'unet.final_conv.weight', 'unet.final_conv.bias', 'conv_seg_new.weight', 'conv_seg_new.bias']
2023-03-03 13:59:38,758 - mmseg - INFO - Parameters in decode_head freezed!
2023-03-03 13:59:38,778 - mmseg - INFO - load checkpoint from local path: pretrained/segformer_mit-b2_512x512_160k_ade20k_20220620_114047-64e4feca.pth
2023-03-03 13:59:39,026 - mmseg - WARNING - The model and loaded state dict do not match exactly

unexpected key in source state_dict: decode_head.conv_seg.weight, decode_head.conv_seg.bias, decode_head.convs.0.conv.weight, decode_head.convs.0.bn.weight, decode_head.convs.0.bn.bias, decode_head.convs.0.bn.running_mean, decode_head.convs.0.bn.running_var, decode_head.convs.0.bn.num_batches_tracked, decode_head.convs.1.conv.weight, decode_head.convs.1.bn.weight, decode_head.convs.1.bn.bias, decode_head.convs.1.bn.running_mean, decode_head.convs.1.bn.running_var, decode_head.convs.1.bn.num_batches_tracked, decode_head.convs.2.conv.weight, decode_head.convs.2.bn.weight, decode_head.convs.2.bn.bias, decode_head.convs.2.bn.running_mean, decode_head.convs.2.bn.running_var, decode_head.convs.2.bn.num_batches_tracked, decode_head.convs.3.conv.weight, decode_head.convs.3.bn.weight, decode_head.convs.3.bn.bias, decode_head.convs.3.bn.running_mean, decode_head.convs.3.bn.running_var, decode_head.convs.3.bn.num_batches_tracked, decode_head.fusion_conv.conv.weight, decode_head.fusion_conv.bn.weight, decode_head.fusion_conv.bn.bias, decode_head.fusion_conv.bn.running_mean, decode_head.fusion_conv.bn.running_var, decode_head.fusion_conv.bn.num_batches_tracked

2023-03-03 13:59:39,040 - mmseg - INFO - load checkpoint from local path: pretrained/segformer_mit-b2_512x512_160k_ade20k_20220620_114047-64e4feca.pth
2023-03-03 13:59:39,262 - mmseg - WARNING - The model and loaded state dict do not match exactly

unexpected key in source state_dict: backbone.layers.0.0.projection.weight, backbone.layers.0.0.projection.bias, backbone.layers.0.0.norm.weight, backbone.layers.0.0.norm.bias, backbone.layers.0.1.0.norm1.weight, backbone.layers.0.1.0.norm1.bias, backbone.layers.0.1.0.attn.attn.in_proj_weight, backbone.layers.0.1.0.attn.attn.in_proj_bias, backbone.layers.0.1.0.attn.attn.out_proj.weight, backbone.layers.0.1.0.attn.attn.out_proj.bias, backbone.layers.0.1.0.attn.sr.weight, backbone.layers.0.1.0.attn.sr.bias, backbone.layers.0.1.0.attn.norm.weight, backbone.layers.0.1.0.attn.norm.bias, backbone.layers.0.1.0.norm2.weight, backbone.layers.0.1.0.norm2.bias, backbone.layers.0.1.0.ffn.layers.0.weight, backbone.layers.0.1.0.ffn.layers.0.bias, backbone.layers.0.1.0.ffn.layers.1.weight, backbone.layers.0.1.0.ffn.layers.1.bias, backbone.layers.0.1.0.ffn.layers.4.weight, backbone.layers.0.1.0.ffn.layers.4.bias, backbone.layers.0.1.1.norm1.weight, backbone.layers.0.1.1.norm1.bias, backbone.layers.0.1.1.attn.attn.in_proj_weight, backbone.layers.0.1.1.attn.attn.in_proj_bias, backbone.layers.0.1.1.attn.attn.out_proj.weight, backbone.layers.0.1.1.attn.attn.out_proj.bias, backbone.layers.0.1.1.attn.sr.weight, backbone.layers.0.1.1.attn.sr.bias, backbone.layers.0.1.1.attn.norm.weight, backbone.layers.0.1.1.attn.norm.bias, backbone.layers.0.1.1.norm2.weight, backbone.layers.0.1.1.norm2.bias, backbone.layers.0.1.1.ffn.layers.0.weight, backbone.layers.0.1.1.ffn.layers.0.bias, backbone.layers.0.1.1.ffn.layers.1.weight, backbone.layers.0.1.1.ffn.layers.1.bias, backbone.layers.0.1.1.ffn.layers.4.weight, backbone.layers.0.1.1.ffn.layers.4.bias, backbone.layers.0.1.2.norm1.weight, backbone.layers.0.1.2.norm1.bias, backbone.layers.0.1.2.attn.attn.in_proj_weight, backbone.layers.0.1.2.attn.attn.in_proj_bias, backbone.layers.0.1.2.attn.attn.out_proj.weight, backbone.layers.0.1.2.attn.attn.out_proj.bias, backbone.layers.0.1.2.attn.sr.weight, backbone.layers.0.1.2.attn.sr.bias, backbone.layers.0.1.2.attn.norm.weight, backbone.layers.0.1.2.attn.norm.bias, backbone.layers.0.1.2.norm2.weight, backbone.layers.0.1.2.norm2.bias, backbone.layers.0.1.2.ffn.layers.0.weight, backbone.layers.0.1.2.ffn.layers.0.bias, backbone.layers.0.1.2.ffn.layers.1.weight, backbone.layers.0.1.2.ffn.layers.1.bias, backbone.layers.0.1.2.ffn.layers.4.weight, backbone.layers.0.1.2.ffn.layers.4.bias, backbone.layers.0.2.weight, backbone.layers.0.2.bias, backbone.layers.1.0.projection.weight, backbone.layers.1.0.projection.bias, backbone.layers.1.0.norm.weight, backbone.layers.1.0.norm.bias, backbone.layers.1.1.0.norm1.weight, backbone.layers.1.1.0.norm1.bias, backbone.layers.1.1.0.attn.attn.in_proj_weight, backbone.layers.1.1.0.attn.attn.in_proj_bias, backbone.layers.1.1.0.attn.attn.out_proj.weight, backbone.layers.1.1.0.attn.attn.out_proj.bias, backbone.layers.1.1.0.attn.sr.weight, backbone.layers.1.1.0.attn.sr.bias, backbone.layers.1.1.0.attn.norm.weight, backbone.layers.1.1.0.attn.norm.bias, backbone.layers.1.1.0.norm2.weight, backbone.layers.1.1.0.norm2.bias, backbone.layers.1.1.0.ffn.layers.0.weight, backbone.layers.1.1.0.ffn.layers.0.bias, backbone.layers.1.1.0.ffn.layers.1.weight, backbone.layers.1.1.0.ffn.layers.1.bias, backbone.layers.1.1.0.ffn.layers.4.weight, backbone.layers.1.1.0.ffn.layers.4.bias, backbone.layers.1.1.1.norm1.weight, backbone.layers.1.1.1.norm1.bias, backbone.layers.1.1.1.attn.attn.in_proj_weight, backbone.layers.1.1.1.attn.attn.in_proj_bias, backbone.layers.1.1.1.attn.attn.out_proj.weight, backbone.layers.1.1.1.attn.attn.out_proj.bias, backbone.layers.1.1.1.attn.sr.weight, backbone.layers.1.1.1.attn.sr.bias, backbone.layers.1.1.1.attn.norm.weight, backbone.layers.1.1.1.attn.norm.bias, backbone.layers.1.1.1.norm2.weight, backbone.layers.1.1.1.norm2.bias, backbone.layers.1.1.1.ffn.layers.0.weight, backbone.layers.1.1.1.ffn.layers.0.bias, backbone.layers.1.1.1.ffn.layers.1.weight, backbone.layers.1.1.1.ffn.layers.1.bias, backbone.layers.1.1.1.ffn.layers.4.weight, backbone.layers.1.1.1.ffn.layers.4.bias, backbone.layers.1.1.2.norm1.weight, backbone.layers.1.1.2.norm1.bias, backbone.layers.1.1.2.attn.attn.in_proj_weight, backbone.layers.1.1.2.attn.attn.in_proj_bias, backbone.layers.1.1.2.attn.attn.out_proj.weight, backbone.layers.1.1.2.attn.attn.out_proj.bias, backbone.layers.1.1.2.attn.sr.weight, backbone.layers.1.1.2.attn.sr.bias, backbone.layers.1.1.2.attn.norm.weight, backbone.layers.1.1.2.attn.norm.bias, backbone.layers.1.1.2.norm2.weight, backbone.layers.1.1.2.norm2.bias, backbone.layers.1.1.2.ffn.layers.0.weight, backbone.layers.1.1.2.ffn.layers.0.bias, backbone.layers.1.1.2.ffn.layers.1.weight, backbone.layers.1.1.2.ffn.layers.1.bias, backbone.layers.1.1.2.ffn.layers.4.weight, backbone.layers.1.1.2.ffn.layers.4.bias, backbone.layers.1.1.3.norm1.weight, backbone.layers.1.1.3.norm1.bias, backbone.layers.1.1.3.attn.attn.in_proj_weight, backbone.layers.1.1.3.attn.attn.in_proj_bias, backbone.layers.1.1.3.attn.attn.out_proj.weight, backbone.layers.1.1.3.attn.attn.out_proj.bias, backbone.layers.1.1.3.attn.sr.weight, backbone.layers.1.1.3.attn.sr.bias, backbone.layers.1.1.3.attn.norm.weight, backbone.layers.1.1.3.attn.norm.bias, backbone.layers.1.1.3.norm2.weight, backbone.layers.1.1.3.norm2.bias, backbone.layers.1.1.3.ffn.layers.0.weight, backbone.layers.1.1.3.ffn.layers.0.bias, backbone.layers.1.1.3.ffn.layers.1.weight, backbone.layers.1.1.3.ffn.layers.1.bias, backbone.layers.1.1.3.ffn.layers.4.weight, backbone.layers.1.1.3.ffn.layers.4.bias, backbone.layers.1.2.weight, backbone.layers.1.2.bias, backbone.layers.2.0.projection.weight, backbone.layers.2.0.projection.bias, backbone.layers.2.0.norm.weight, backbone.layers.2.0.norm.bias, backbone.layers.2.1.0.norm1.weight, backbone.layers.2.1.0.norm1.bias, backbone.layers.2.1.0.attn.attn.in_proj_weight, backbone.layers.2.1.0.attn.attn.in_proj_bias, backbone.layers.2.1.0.attn.attn.out_proj.weight, backbone.layers.2.1.0.attn.attn.out_proj.bias, backbone.layers.2.1.0.attn.sr.weight, backbone.layers.2.1.0.attn.sr.bias, backbone.layers.2.1.0.attn.norm.weight, backbone.layers.2.1.0.attn.norm.bias, backbone.layers.2.1.0.norm2.weight, backbone.layers.2.1.0.norm2.bias, backbone.layers.2.1.0.ffn.layers.0.weight, backbone.layers.2.1.0.ffn.layers.0.bias, backbone.layers.2.1.0.ffn.layers.1.weight, backbone.layers.2.1.0.ffn.layers.1.bias, backbone.layers.2.1.0.ffn.layers.4.weight, backbone.layers.2.1.0.ffn.layers.4.bias, backbone.layers.2.1.1.norm1.weight, backbone.layers.2.1.1.norm1.bias, backbone.layers.2.1.1.attn.attn.in_proj_weight, backbone.layers.2.1.1.attn.attn.in_proj_bias, backbone.layers.2.1.1.attn.attn.out_proj.weight, backbone.layers.2.1.1.attn.attn.out_proj.bias, backbone.layers.2.1.1.attn.sr.weight, backbone.layers.2.1.1.attn.sr.bias, backbone.layers.2.1.1.attn.norm.weight, backbone.layers.2.1.1.attn.norm.bias, backbone.layers.2.1.1.norm2.weight, backbone.layers.2.1.1.norm2.bias, backbone.layers.2.1.1.ffn.layers.0.weight, backbone.layers.2.1.1.ffn.layers.0.bias, backbone.layers.2.1.1.ffn.layers.1.weight, backbone.layers.2.1.1.ffn.layers.1.bias, backbone.layers.2.1.1.ffn.layers.4.weight, backbone.layers.2.1.1.ffn.layers.4.bias, backbone.layers.2.1.2.norm1.weight, backbone.layers.2.1.2.norm1.bias, backbone.layers.2.1.2.attn.attn.in_proj_weight, backbone.layers.2.1.2.attn.attn.in_proj_bias, backbone.layers.2.1.2.attn.attn.out_proj.weight, backbone.layers.2.1.2.attn.attn.out_proj.bias, backbone.layers.2.1.2.attn.sr.weight, backbone.layers.2.1.2.attn.sr.bias, backbone.layers.2.1.2.attn.norm.weight, backbone.layers.2.1.2.attn.norm.bias, backbone.layers.2.1.2.norm2.weight, backbone.layers.2.1.2.norm2.bias, backbone.layers.2.1.2.ffn.layers.0.weight, backbone.layers.2.1.2.ffn.layers.0.bias, backbone.layers.2.1.2.ffn.layers.1.weight, backbone.layers.2.1.2.ffn.layers.1.bias, backbone.layers.2.1.2.ffn.layers.4.weight, backbone.layers.2.1.2.ffn.layers.4.bias, backbone.layers.2.1.3.norm1.weight, backbone.layers.2.1.3.norm1.bias, backbone.layers.2.1.3.attn.attn.in_proj_weight, backbone.layers.2.1.3.attn.attn.in_proj_bias, backbone.layers.2.1.3.attn.attn.out_proj.weight, backbone.layers.2.1.3.attn.attn.out_proj.bias, backbone.layers.2.1.3.attn.sr.weight, backbone.layers.2.1.3.attn.sr.bias, backbone.layers.2.1.3.attn.norm.weight, backbone.layers.2.1.3.attn.norm.bias, backbone.layers.2.1.3.norm2.weight, backbone.layers.2.1.3.norm2.bias, backbone.layers.2.1.3.ffn.layers.0.weight, backbone.layers.2.1.3.ffn.layers.0.bias, backbone.layers.2.1.3.ffn.layers.1.weight, backbone.layers.2.1.3.ffn.layers.1.bias, backbone.layers.2.1.3.ffn.layers.4.weight, backbone.layers.2.1.3.ffn.layers.4.bias, backbone.layers.2.1.4.norm1.weight, backbone.layers.2.1.4.norm1.bias, backbone.layers.2.1.4.attn.attn.in_proj_weight, backbone.layers.2.1.4.attn.attn.in_proj_bias, backbone.layers.2.1.4.attn.attn.out_proj.weight, backbone.layers.2.1.4.attn.attn.out_proj.bias, backbone.layers.2.1.4.attn.sr.weight, backbone.layers.2.1.4.attn.sr.bias, backbone.layers.2.1.4.attn.norm.weight, backbone.layers.2.1.4.attn.norm.bias, backbone.layers.2.1.4.norm2.weight, backbone.layers.2.1.4.norm2.bias, backbone.layers.2.1.4.ffn.layers.0.weight, backbone.layers.2.1.4.ffn.layers.0.bias, backbone.layers.2.1.4.ffn.layers.1.weight, backbone.layers.2.1.4.ffn.layers.1.bias, backbone.layers.2.1.4.ffn.layers.4.weight, backbone.layers.2.1.4.ffn.layers.4.bias, backbone.layers.2.1.5.norm1.weight, backbone.layers.2.1.5.norm1.bias, backbone.layers.2.1.5.attn.attn.in_proj_weight, backbone.layers.2.1.5.attn.attn.in_proj_bias, backbone.layers.2.1.5.attn.attn.out_proj.weight, backbone.layers.2.1.5.attn.attn.out_proj.bias, backbone.layers.2.1.5.attn.sr.weight, backbone.layers.2.1.5.attn.sr.bias, backbone.layers.2.1.5.attn.norm.weight, backbone.layers.2.1.5.attn.norm.bias, backbone.layers.2.1.5.norm2.weight, backbone.layers.2.1.5.norm2.bias, backbone.layers.2.1.5.ffn.layers.0.weight, backbone.layers.2.1.5.ffn.layers.0.bias, backbone.layers.2.1.5.ffn.layers.1.weight, backbone.layers.2.1.5.ffn.layers.1.bias, backbone.layers.2.1.5.ffn.layers.4.weight, backbone.layers.2.1.5.ffn.layers.4.bias, backbone.layers.2.2.weight, backbone.layers.2.2.bias, backbone.layers.3.0.projection.weight, backbone.layers.3.0.projection.bias, backbone.layers.3.0.norm.weight, backbone.layers.3.0.norm.bias, backbone.layers.3.1.0.norm1.weight, backbone.layers.3.1.0.norm1.bias, backbone.layers.3.1.0.attn.attn.in_proj_weight, backbone.layers.3.1.0.attn.attn.in_proj_bias, backbone.layers.3.1.0.attn.attn.out_proj.weight, backbone.layers.3.1.0.attn.attn.out_proj.bias, backbone.layers.3.1.0.norm2.weight, backbone.layers.3.1.0.norm2.bias, backbone.layers.3.1.0.ffn.layers.0.weight, backbone.layers.3.1.0.ffn.layers.0.bias, backbone.layers.3.1.0.ffn.layers.1.weight, backbone.layers.3.1.0.ffn.layers.1.bias, backbone.layers.3.1.0.ffn.layers.4.weight, backbone.layers.3.1.0.ffn.layers.4.bias, backbone.layers.3.1.1.norm1.weight, backbone.layers.3.1.1.norm1.bias, backbone.layers.3.1.1.attn.attn.in_proj_weight, backbone.layers.3.1.1.attn.attn.in_proj_bias, backbone.layers.3.1.1.attn.attn.out_proj.weight, backbone.layers.3.1.1.attn.attn.out_proj.bias, backbone.layers.3.1.1.norm2.weight, backbone.layers.3.1.1.norm2.bias, backbone.layers.3.1.1.ffn.layers.0.weight, backbone.layers.3.1.1.ffn.layers.0.bias, backbone.layers.3.1.1.ffn.layers.1.weight, backbone.layers.3.1.1.ffn.layers.1.bias, backbone.layers.3.1.1.ffn.layers.4.weight, backbone.layers.3.1.1.ffn.layers.4.bias, backbone.layers.3.1.2.norm1.weight, backbone.layers.3.1.2.norm1.bias, backbone.layers.3.1.2.attn.attn.in_proj_weight, backbone.layers.3.1.2.attn.attn.in_proj_bias, backbone.layers.3.1.2.attn.attn.out_proj.weight, backbone.layers.3.1.2.attn.attn.out_proj.bias, backbone.layers.3.1.2.norm2.weight, backbone.layers.3.1.2.norm2.bias, backbone.layers.3.1.2.ffn.layers.0.weight, backbone.layers.3.1.2.ffn.layers.0.bias, backbone.layers.3.1.2.ffn.layers.1.weight, backbone.layers.3.1.2.ffn.layers.1.bias, backbone.layers.3.1.2.ffn.layers.4.weight, backbone.layers.3.1.2.ffn.layers.4.bias, backbone.layers.3.2.weight, backbone.layers.3.2.bias

missing keys in source state_dict: unet.init_conv.weight, unet.init_conv.bias, unet.time_mlp.1.weight, unet.time_mlp.1.bias, unet.time_mlp.3.weight, unet.time_mlp.3.bias, unet.downs.0.0.mlp.1.weight, unet.downs.0.0.mlp.1.bias, unet.downs.0.0.block1.proj.weight, unet.downs.0.0.block1.proj.bias, unet.downs.0.0.block1.norm.weight, unet.downs.0.0.block1.norm.bias, unet.downs.0.0.block2.proj.weight, unet.downs.0.0.block2.proj.bias, unet.downs.0.0.block2.norm.weight, unet.downs.0.0.block2.norm.bias, unet.downs.0.1.mlp.1.weight, unet.downs.0.1.mlp.1.bias, unet.downs.0.1.block1.proj.weight, unet.downs.0.1.block1.proj.bias, unet.downs.0.1.block1.norm.weight, unet.downs.0.1.block1.norm.bias, unet.downs.0.1.block2.proj.weight, unet.downs.0.1.block2.proj.bias, unet.downs.0.1.block2.norm.weight, unet.downs.0.1.block2.norm.bias, unet.downs.0.2.fn.fn.to_qkv.weight, unet.downs.0.2.fn.fn.to_out.0.weight, unet.downs.0.2.fn.fn.to_out.0.bias, unet.downs.0.2.fn.fn.to_out.1.g, unet.downs.0.2.fn.norm.g, unet.downs.0.3.weight, unet.downs.0.3.bias, unet.downs.1.0.mlp.1.weight, unet.downs.1.0.mlp.1.bias, unet.downs.1.0.block1.proj.weight, unet.downs.1.0.block1.proj.bias, unet.downs.1.0.block1.norm.weight, unet.downs.1.0.block1.norm.bias, unet.downs.1.0.block2.proj.weight, unet.downs.1.0.block2.proj.bias, unet.downs.1.0.block2.norm.weight, unet.downs.1.0.block2.norm.bias, unet.downs.1.1.mlp.1.weight, unet.downs.1.1.mlp.1.bias, unet.downs.1.1.block1.proj.weight, unet.downs.1.1.block1.proj.bias, unet.downs.1.1.block1.norm.weight, unet.downs.1.1.block1.norm.bias, unet.downs.1.1.block2.proj.weight, unet.downs.1.1.block2.proj.bias, unet.downs.1.1.block2.norm.weight, unet.downs.1.1.block2.norm.bias, unet.downs.1.2.fn.fn.to_qkv.weight, unet.downs.1.2.fn.fn.to_out.0.weight, unet.downs.1.2.fn.fn.to_out.0.bias, unet.downs.1.2.fn.fn.to_out.1.g, unet.downs.1.2.fn.norm.g, unet.downs.1.3.weight, unet.downs.1.3.bias, unet.downs.2.0.mlp.1.weight, unet.downs.2.0.mlp.1.bias, unet.downs.2.0.block1.proj.weight, unet.downs.2.0.block1.proj.bias, unet.downs.2.0.block1.norm.weight, unet.downs.2.0.block1.norm.bias, unet.downs.2.0.block2.proj.weight, unet.downs.2.0.block2.proj.bias, unet.downs.2.0.block2.norm.weight, unet.downs.2.0.block2.norm.bias, unet.downs.2.1.mlp.1.weight, unet.downs.2.1.mlp.1.bias, unet.downs.2.1.block1.proj.weight, unet.downs.2.1.block1.proj.bias, unet.downs.2.1.block1.norm.weight, unet.downs.2.1.block1.norm.bias, unet.downs.2.1.block2.proj.weight, unet.downs.2.1.block2.proj.bias, unet.downs.2.1.block2.norm.weight, unet.downs.2.1.block2.norm.bias, unet.downs.2.2.fn.fn.to_qkv.weight, unet.downs.2.2.fn.fn.to_out.0.weight, unet.downs.2.2.fn.fn.to_out.0.bias, unet.downs.2.2.fn.fn.to_out.1.g, unet.downs.2.2.fn.norm.g, unet.downs.2.3.weight, unet.downs.2.3.bias, unet.ups.0.0.mlp.1.weight, unet.ups.0.0.mlp.1.bias, unet.ups.0.0.block1.proj.weight, unet.ups.0.0.block1.proj.bias, unet.ups.0.0.block1.norm.weight, unet.ups.0.0.block1.norm.bias, unet.ups.0.0.block2.proj.weight, unet.ups.0.0.block2.proj.bias, unet.ups.0.0.block2.norm.weight, unet.ups.0.0.block2.norm.bias, unet.ups.0.0.res_conv.weight, unet.ups.0.0.res_conv.bias, unet.ups.0.1.mlp.1.weight, unet.ups.0.1.mlp.1.bias, unet.ups.0.1.block1.proj.weight, unet.ups.0.1.block1.proj.bias, unet.ups.0.1.block1.norm.weight, unet.ups.0.1.block1.norm.bias, unet.ups.0.1.block2.proj.weight, unet.ups.0.1.block2.proj.bias, unet.ups.0.1.block2.norm.weight, unet.ups.0.1.block2.norm.bias, unet.ups.0.1.res_conv.weight, unet.ups.0.1.res_conv.bias, unet.ups.0.2.fn.fn.to_qkv.weight, unet.ups.0.2.fn.fn.to_out.0.weight, unet.ups.0.2.fn.fn.to_out.0.bias, unet.ups.0.2.fn.fn.to_out.1.g, unet.ups.0.2.fn.norm.g, unet.ups.0.3.1.weight, unet.ups.0.3.1.bias, unet.ups.1.0.mlp.1.weight, unet.ups.1.0.mlp.1.bias, unet.ups.1.0.block1.proj.weight, unet.ups.1.0.block1.proj.bias, unet.ups.1.0.block1.norm.weight, unet.ups.1.0.block1.norm.bias, unet.ups.1.0.block2.proj.weight, unet.ups.1.0.block2.proj.bias, unet.ups.1.0.block2.norm.weight, unet.ups.1.0.block2.norm.bias, unet.ups.1.0.res_conv.weight, unet.ups.1.0.res_conv.bias, unet.ups.1.1.mlp.1.weight, unet.ups.1.1.mlp.1.bias, unet.ups.1.1.block1.proj.weight, unet.ups.1.1.block1.proj.bias, unet.ups.1.1.block1.norm.weight, unet.ups.1.1.block1.norm.bias, unet.ups.1.1.block2.proj.weight, unet.ups.1.1.block2.proj.bias, unet.ups.1.1.block2.norm.weight, unet.ups.1.1.block2.norm.bias, unet.ups.1.1.res_conv.weight, unet.ups.1.1.res_conv.bias, unet.ups.1.2.fn.fn.to_qkv.weight, unet.ups.1.2.fn.fn.to_out.0.weight, unet.ups.1.2.fn.fn.to_out.0.bias, unet.ups.1.2.fn.fn.to_out.1.g, unet.ups.1.2.fn.norm.g, unet.ups.1.3.1.weight, unet.ups.1.3.1.bias, unet.ups.2.0.mlp.1.weight, unet.ups.2.0.mlp.1.bias, unet.ups.2.0.block1.proj.weight, unet.ups.2.0.block1.proj.bias, unet.ups.2.0.block1.norm.weight, unet.ups.2.0.block1.norm.bias, unet.ups.2.0.block2.proj.weight, unet.ups.2.0.block2.proj.bias, unet.ups.2.0.block2.norm.weight, unet.ups.2.0.block2.norm.bias, unet.ups.2.0.res_conv.weight, unet.ups.2.0.res_conv.bias, unet.ups.2.1.mlp.1.weight, unet.ups.2.1.mlp.1.bias, unet.ups.2.1.block1.proj.weight, unet.ups.2.1.block1.proj.bias, unet.ups.2.1.block1.norm.weight, unet.ups.2.1.block1.norm.bias, unet.ups.2.1.block2.proj.weight, unet.ups.2.1.block2.proj.bias, unet.ups.2.1.block2.norm.weight, unet.ups.2.1.block2.norm.bias, unet.ups.2.1.res_conv.weight, unet.ups.2.1.res_conv.bias, unet.ups.2.2.fn.fn.to_qkv.weight, unet.ups.2.2.fn.fn.to_out.0.weight, unet.ups.2.2.fn.fn.to_out.0.bias, unet.ups.2.2.fn.fn.to_out.1.g, unet.ups.2.2.fn.norm.g, unet.ups.2.3.weight, unet.ups.2.3.bias, unet.mid_block1.mlp.1.weight, unet.mid_block1.mlp.1.bias, unet.mid_block1.block1.proj.weight, unet.mid_block1.block1.proj.bias, unet.mid_block1.block1.norm.weight, unet.mid_block1.block1.norm.bias, unet.mid_block1.block2.proj.weight, unet.mid_block1.block2.proj.bias, unet.mid_block1.block2.norm.weight, unet.mid_block1.block2.norm.bias, unet.mid_attn.fn.fn.to_qkv.weight, unet.mid_attn.fn.fn.to_out.weight, unet.mid_attn.fn.fn.to_out.bias, unet.mid_attn.fn.norm.g, unet.mid_block2.mlp.1.weight, unet.mid_block2.mlp.1.bias, unet.mid_block2.block1.proj.weight, unet.mid_block2.block1.proj.bias, unet.mid_block2.block1.norm.weight, unet.mid_block2.block1.norm.bias, unet.mid_block2.block2.proj.weight, unet.mid_block2.block2.proj.bias, unet.mid_block2.block2.norm.weight, unet.mid_block2.block2.norm.bias, unet.final_res_block.mlp.1.weight, unet.final_res_block.mlp.1.bias, unet.final_res_block.block1.proj.weight, unet.final_res_block.block1.proj.bias, unet.final_res_block.block1.norm.weight, unet.final_res_block.block1.norm.bias, unet.final_res_block.block2.proj.weight, unet.final_res_block.block2.proj.bias, unet.final_res_block.block2.norm.weight, unet.final_res_block.block2.norm.bias, unet.final_res_block.res_conv.weight, unet.final_res_block.res_conv.bias, unet.final_conv.weight, unet.final_conv.bias, conv_seg_new.weight, conv_seg_new.bias, embed.weight

2023-03-03 13:59:39,286 - mmseg - INFO - EncoderDecoderFreeze(
  (backbone): MixVisionTransformerCustomInitWeights(
    (layers): ModuleList(
      (0): ModuleList(
        (0): PatchEmbed(
          (projection): Conv2d(3, 64, kernel_size=(7, 7), stride=(4, 4), padding=(3, 3))
          (norm): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
        )
        (1): ModuleList(
          (0): TransformerEncoderLayer(
            (norm1): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=64, out_features=64, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(64, 64, kernel_size=(8, 8), stride=(8, 8))
              (norm): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (1): TransformerEncoderLayer(
            (norm1): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=64, out_features=64, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(64, 64, kernel_size=(8, 8), stride=(8, 8))
              (norm): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (2): TransformerEncoderLayer(
            (norm1): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=64, out_features=64, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(64, 64, kernel_size=(8, 8), stride=(8, 8))
              (norm): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
        )
        (2): LayerNorm((64,), eps=1e-06, elementwise_affine=True)
      )
      (1): ModuleList(
        (0): PatchEmbed(
          (projection): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))
          (norm): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
        )
        (1): ModuleList(
          (0): TransformerEncoderLayer(
            (norm1): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=128, out_features=128, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(128, 128, kernel_size=(4, 4), stride=(4, 4))
              (norm): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (1): TransformerEncoderLayer(
            (norm1): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=128, out_features=128, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(128, 128, kernel_size=(4, 4), stride=(4, 4))
              (norm): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (2): TransformerEncoderLayer(
            (norm1): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=128, out_features=128, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(128, 128, kernel_size=(4, 4), stride=(4, 4))
              (norm): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (3): TransformerEncoderLayer(
            (norm1): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=128, out_features=128, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(128, 128, kernel_size=(4, 4), stride=(4, 4))
              (norm): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
        )
        (2): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
      )
      (2): ModuleList(
        (0): PatchEmbed(
          (projection): Conv2d(128, 320, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))
          (norm): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
        )
        (1): ModuleList(
          (0): TransformerEncoderLayer(
            (norm1): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=320, out_features=320, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(320, 320, kernel_size=(2, 2), stride=(2, 2))
              (norm): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(320, 1280, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(1280, 1280, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1280)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(1280, 320, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (1): TransformerEncoderLayer(
            (norm1): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=320, out_features=320, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(320, 320, kernel_size=(2, 2), stride=(2, 2))
              (norm): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(320, 1280, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(1280, 1280, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1280)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(1280, 320, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (2): TransformerEncoderLayer(
            (norm1): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=320, out_features=320, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(320, 320, kernel_size=(2, 2), stride=(2, 2))
              (norm): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(320, 1280, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(1280, 1280, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1280)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(1280, 320, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (3): TransformerEncoderLayer(
            (norm1): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=320, out_features=320, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(320, 320, kernel_size=(2, 2), stride=(2, 2))
              (norm): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(320, 1280, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(1280, 1280, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1280)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(1280, 320, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (4): TransformerEncoderLayer(
            (norm1): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=320, out_features=320, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(320, 320, kernel_size=(2, 2), stride=(2, 2))
              (norm): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(320, 1280, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(1280, 1280, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1280)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(1280, 320, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (5): TransformerEncoderLayer(
            (norm1): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=320, out_features=320, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (sr): Conv2d(320, 320, kernel_size=(2, 2), stride=(2, 2))
              (norm): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            )
            (norm2): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(320, 1280, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(1280, 1280, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1280)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(1280, 320, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
        )
        (2): LayerNorm((320,), eps=1e-06, elementwise_affine=True)
      )
      (3): ModuleList(
        (0): PatchEmbed(
          (projection): Conv2d(320, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))
          (norm): LayerNorm((512,), eps=1e-06, elementwise_affine=True)
        )
        (1): ModuleList(
          (0): TransformerEncoderLayer(
            (norm1): LayerNorm((512,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
            (norm2): LayerNorm((512,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(2048, 2048, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=2048)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (1): TransformerEncoderLayer(
            (norm1): LayerNorm((512,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
            (norm2): LayerNorm((512,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(2048, 2048, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=2048)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
          (2): TransformerEncoderLayer(
            (norm1): LayerNorm((512,), eps=1e-06, elementwise_affine=True)
            (attn): EfficientMultiheadAttention(
              (attn): MultiheadAttention(
                (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
            (norm2): LayerNorm((512,), eps=1e-06, elementwise_affine=True)
            (ffn): MixFFN(
              (activate): GELU(approximate='none')
              (layers): Sequential(
                (0): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))
                (1): Conv2d(2048, 2048, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=2048)
                (2): GELU(approximate='none')
                (3): Dropout(p=0.0, inplace=False)
                (4): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1))
                (5): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
            )
          )
        )
        (2): LayerNorm((512,), eps=1e-06, elementwise_affine=True)
      )
    )
  )
  init_cfg={'type': 'Pretrained', 'checkpoint': 'pretrained/segformer_mit-b2_512x512_160k_ade20k_20220620_114047-64e4feca.pth'}
  (decode_head): SegformerHeadUnetFCHeadSingleStep(
    input_transform=multiple_select, ignore_index=0, align_corners=False
    (loss_decode): CrossEntropyLoss(avg_non_ignore=False)
    (conv_seg): None
    (dropout): Dropout2d(p=0.1, inplace=False)
    (convs): ModuleList(
      (0): ConvModule(
        (conv): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn): SyncBatchNorm(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (activate): ReLU(inplace=True)
      )
      (1): ConvModule(
        (conv): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn): SyncBatchNorm(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (activate): ReLU(inplace=True)
      )
      (2): ConvModule(
        (conv): Conv2d(320, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn): SyncBatchNorm(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (activate): ReLU(inplace=True)
      )
      (3): ConvModule(
        (conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn): SyncBatchNorm(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (activate): ReLU(inplace=True)
      )
    )
    (fusion_conv): ConvModule(
      (conv): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn): SyncBatchNorm(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (activate): ReLU(inplace=True)
    )
    (unet): Unet(
      (init_conv): Conv2d(272, 128, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3))
      (time_mlp): Sequential(
        (0): SinusoidalPosEmb()
        (1): Linear(in_features=128, out_features=512, bias=True)
        (2): GELU(approximate='none')
        (3): Linear(in_features=512, out_features=512, bias=True)
      )
      (downs): ModuleList(
        (0): ModuleList(
          (0): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Identity()
          )
          (1): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Identity()
          )
          (2): Residual(
            (fn): PreNorm(
              (fn): LinearAttention(
                (to_qkv): Conv2d(128, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)
                (to_out): Sequential(
                  (0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
                  (1): LayerNorm()
                )
              )
              (norm): LayerNorm()
            )
          )
          (3): Conv2d(128, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1))
        )
        (1): ModuleList(
          (0): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Identity()
          )
          (1): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Identity()
          )
          (2): Residual(
            (fn): PreNorm(
              (fn): LinearAttention(
                (to_qkv): Conv2d(128, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)
                (to_out): Sequential(
                  (0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
                  (1): LayerNorm()
                )
              )
              (norm): LayerNorm()
            )
          )
          (3): Conv2d(128, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1))
        )
        (2): ModuleList(
          (0): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Identity()
          )
          (1): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Identity()
          )
          (2): Residual(
            (fn): PreNorm(
              (fn): LinearAttention(
                (to_qkv): Conv2d(128, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)
                (to_out): Sequential(
                  (0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
                  (1): LayerNorm()
                )
              )
              (norm): LayerNorm()
            )
          )
          (3): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        )
      )
      (ups): ModuleList(
        (0): ModuleList(
          (0): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1))
          )
          (1): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1))
          )
          (2): Residual(
            (fn): PreNorm(
              (fn): LinearAttention(
                (to_qkv): Conv2d(128, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)
                (to_out): Sequential(
                  (0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
                  (1): LayerNorm()
                )
              )
              (norm): LayerNorm()
            )
          )
          (3): Sequential(
            (0): Upsample(scale_factor=2.0, mode=nearest)
            (1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          )
        )
        (1): ModuleList(
          (0): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1))
          )
          (1): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1))
          )
          (2): Residual(
            (fn): PreNorm(
              (fn): LinearAttention(
                (to_qkv): Conv2d(128, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)
                (to_out): Sequential(
                  (0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
                  (1): LayerNorm()
                )
              )
              (norm): LayerNorm()
            )
          )
          (3): Sequential(
            (0): Upsample(scale_factor=2.0, mode=nearest)
            (1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          )
        )
        (2): ModuleList(
          (0): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1))
          )
          (1): ResnetBlock(
            (mlp): Sequential(
              (0): SiLU()
              (1): Linear(in_features=512, out_features=256, bias=True)
            )
            (block1): Block(
              (proj): WeightStandardizedConv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (block2): Block(
              (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
              (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
              (act): SiLU()
            )
            (res_conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1))
          )
          (2): Residual(
            (fn): PreNorm(
              (fn): LinearAttention(
                (to_qkv): Conv2d(128, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)
                (to_out): Sequential(
                  (0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
                  (1): LayerNorm()
                )
              )
              (norm): LayerNorm()
            )
          )
          (3): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        )
      )
      (mid_block1): ResnetBlock(
        (mlp): Sequential(
          (0): SiLU()
          (1): Linear(in_features=512, out_features=256, bias=True)
        )
        (block1): Block(
          (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
          (act): SiLU()
        )
        (block2): Block(
          (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
          (act): SiLU()
        )
        (res_conv): Identity()
      )
      (mid_attn): Residual(
        (fn): PreNorm(
          (fn): Attention(
            (to_qkv): Conv2d(128, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (to_out): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
          )
          (norm): LayerNorm()
        )
      )
      (mid_block2): ResnetBlock(
        (mlp): Sequential(
          (0): SiLU()
          (1): Linear(in_features=512, out_features=256, bias=True)
        )
        (block1): Block(
          (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
          (act): SiLU()
        )
        (block2): Block(
          (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
          (act): SiLU()
        )
        (res_conv): Identity()
      )
      (final_res_block): ResnetBlock(
        (mlp): Sequential(
          (0): SiLU()
          (1): Linear(in_features=512, out_features=256, bias=True)
        )
        (block1): Block(
          (proj): WeightStandardizedConv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
          (act): SiLU()
        )
        (block2): Block(
          (proj): WeightStandardizedConv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          (norm): GroupNorm(8, 128, eps=1e-05, affine=True)
          (act): SiLU()
        )
        (res_conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1))
      )
      (final_conv): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1))
    )
    (conv_seg_new): Conv2d(256, 151, kernel_size=(1, 1), stride=(1, 1))
    (embed): Embedding(151, 16)
  )
  init_cfg={'type': 'Pretrained', 'checkpoint': 'pretrained/segformer_mit-b2_512x512_160k_ade20k_20220620_114047-64e4feca.pth'}
)
2023-03-03 13:59:40,184 - mmseg - INFO - Loaded 20210 images
2023-03-03 13:59:41,189 - mmseg - INFO - Loaded 2000 images
2023-03-03 13:59:41,192 - mmseg - INFO - Start running, host: laizeqiang@SH-IDC1-10-140-37-124, work_dir: /mnt/petrelfs/laizeqiang/mmseg-baseline/work_dirs/segformer_mit_b2_segformer_head_unet_fc_single_step_ade_pretrained_freeze_embed_80k_ade20k151
2023-03-03 13:59:41,192 - mmseg - INFO - Hooks will be executed in the following order:
before_run:
(VERY_HIGH   ) StepLrUpdaterHook                  
(NORMAL      ) CheckpointHook                     
(LOW         ) DistEvalHook                       
(VERY_LOW    ) TextLoggerHook                     
 -------------------- 
before_train_epoch:
(VERY_HIGH   ) StepLrUpdaterHook                  
(LOW         ) IterTimerHook                      
(LOW         ) DistEvalHook                       
(VERY_LOW    ) TextLoggerHook                     
 -------------------- 
before_train_iter:
(VERY_HIGH   ) StepLrUpdaterHook                  
(LOW         ) IterTimerHook                      
(LOW         ) DistEvalHook                       
 -------------------- 
after_train_iter:
(ABOVE_NORMAL) OptimizerHook                      
(NORMAL      ) CheckpointHook                     
(LOW         ) IterTimerHook                      
(LOW         ) DistEvalHook                       
(VERY_LOW    ) TextLoggerHook                     
 -------------------- 
after_train_epoch:
(NORMAL      ) CheckpointHook                     
(LOW         ) DistEvalHook                       
(VERY_LOW    ) TextLoggerHook                     
 -------------------- 
before_val_epoch:
(LOW         ) IterTimerHook                      
(VERY_LOW    ) TextLoggerHook                     
 -------------------- 
before_val_iter:
(LOW         ) IterTimerHook                      
 -------------------- 
after_val_iter:
(LOW         ) IterTimerHook                      
 -------------------- 
after_val_epoch:
(VERY_LOW    ) TextLoggerHook                     
 -------------------- 
after_run:
(VERY_LOW    ) TextLoggerHook                     
 -------------------- 
2023-03-03 13:59:41,192 - mmseg - INFO - workflow: [('train', 1)], max: 80000 iters
2023-03-03 13:59:41,192 - mmseg - INFO - Checkpoints will be saved to /mnt/petrelfs/laizeqiang/mmseg-baseline/work_dirs/segformer_mit_b2_segformer_head_unet_fc_single_step_ade_pretrained_freeze_embed_80k_ade20k151 by HardDiskBackend.