RuntimeError: The size of tensor a (173) must match the size of tensor b (160) at non-singleton dimension 0

[berylyellow]

想问一下，我自己写了veri wild 的数据读入部分，我想直接用在veri数据集上训练的模型测试一下在veri wild上的结果，然后报错了，请问还要修改哪里呢？
完整的报错信息是这样的：
Traceback (most recent call last): File "test.py", line 53, in model.load_param(cfg.TEST.WEIGHT) File "E:\Win_Pycharm_Project\TransReID-main\TransReID-main\model\make_model.py", line 376, in load_param self.state_dict()[i.replace('module.', '')].copy_(param_dict[i]) RuntimeError: The size of tensor a (173) must match the size of tensor b (160) at non-singleton dimension 0

麻烦大佬了~~~

[heshuting555]

是不是fc的大小不一致，可以把parameter的key打印出来就知道是哪个地方出错啦

[YuzhouPeng]

想问一下，我自己写了veri wild 的数据读入部分，我想直接用在veri数据集上训练的模型测试一下在veri wild上的结果，然后报错了，请问还要修改哪里呢？ 完整的报错信息是这样的： Traceback (most recent call last): File "test.py", line 53, in model.load_param(cfg.TEST.WEIGHT) File "E:\Win_Pycharm_Project\TransReID-main\TransReID-main\model\make_model.py", line 376, in load_param self.state_dict()[i.replace('module.', '')].copy_(param_dict[i]) RuntimeError: The size of tensor a (173) must match the size of tensor b (160) at non-singleton dimension 0

麻烦大佬了~~~

请问你解决了吗？我这边也遇到类似错误

[hamid-mp]

Hi, Thank you for sharing the code
I have same problem
here is part of defaults.py config:

_C.MODEL.NAME = 'transformer'
_C.MODEL.LAST_STRIDE = 1
_C.MODEL.PRETRAIN_PATH = './jx_vit_base_p16_224-80ecf9dd.pth'
_C.MODEL.PRETRAIN_CHOICE = 'imagenet'
...
...
_C.MODEL.TRANSFORMER_TYPE = 'vit_base_patch16_224_TransReID'
....
....
_C.TEST.WEIGHT = "./vit_transreid_vehicleID.pth"

I want to use your pretrained model on Vehicle ReID datasets, so I'll be gratefull if you mention where I'm wrong. Thanks

[kp97524]

Hi all ,

Did you figure out the fix for this issue? Am facing the same issue and there's no update for a while :(

Github issue: #70

[KingH12138]

Obviously,there will be some problems in "base.pos_embed" if you use vit_base to do inference on market1501.
em.....let me try to dealing with it first.Authors may take wrong weight file in readme.md's table.

[KingH12138]

Obviously,there will be some problems in "base.pos_embed" if you use vit_base to do inference on market1501. em.....let me try to dealing with it first.Authors may take wrong weight file in readme.md's table.

Alright, I can only use a kind of stupied method......First,this is kind of mismatch between model and loaded model(aka: weight).
Then,I rewrite the load_param function in codes/TransReID-main/model/make_model.py as followed:
`def load_param(self, trained_path):

    param_dict = torch.load(trained_path)

    for i in param_dict:

        # print(self.state_dict()[i.replace('module.', '')].shape,param_dict[i].shape)

        cur_model_key = i.replace('module.', '')

        if cur_model_key in self.state_dict():

            self.state_dict()[i.replace('module.', '')].copy_(param_dict[i])

        else:

            print(f"Warning:{cur_model_key} not found in current model but in loaded dictionary.")

    print('Loading pretrained model from {}'.format(trained_path))`

The order of this function is to match the "params" that can be matched and jump over that can't be matched.
This is my config:
`MODEL:
PRETRAIN_CHOICE: 'imagenet'
PRETRAIN_PATH: '/data/jhb_data/checkpoints/jx_vit_base_p16_224-80ecf9dd.pth'
METRIC_LOSS_TYPE: 'triplet'
IF_LABELSMOOTH: 'off'
IF_WITH_CENTER: 'no'
NAME: 'transformer'
NO_MARGIN: True
DEVICE_ID: ('2')
TRANSFORMER_TYPE: 'vit_base_patch16_224_TransReID'
STRIDE_SIZE: [12, 12]
SIE_CAMERA: True
SIE_COE: 3.0
JPM: False
RE_ARRANGE: True

INPUT:
SIZE_TRAIN: [256, 128]
SIZE_TEST: [256, 128]
PROB: 0.5 # random horizontal flip
RE_PROB: 0.5 # random erasing
PADDING: 10
PIXEL_MEAN: [0.5, 0.5, 0.5]
PIXEL_STD: [0.5, 0.5, 0.5]

DATASETS:
NAMES: ('market1501')
ROOT_DIR: ('/data/jhb_data/datasets')

DATALOADER:
SAMPLER: 'softmax_triplet'
NUM_INSTANCE: 4
NUM_WORKERS: 8

SOLVER:
OPTIMIZER_NAME: 'SGD'
MAX_EPOCHS: 120
BASE_LR: 0.008
IMS_PER_BATCH: 64
WARMUP_METHOD: 'linear'
LARGE_FC_LR: False
CHECKPOINT_PERIOD: 120
LOG_PERIOD: 50
EVAL_PERIOD: 120
WEIGHT_DECAY: 1e-4
WEIGHT_DECAY_BIAS: 1e-4
BIAS_LR_FACTOR: 2

TEST:
EVAL: True
IMS_PER_BATCH: 256
RE_RANKING: False
WEIGHT: '/data/jhb_data/checkpoints/vit_transreid_market.pth'
NECK_FEAT: 'before'
FEAT_NORM: 'yes'

OUTPUT_DIR: '../logs/market_vit_transreid_stride'

2024-10-21 07:45:40,986 transreid INFO: Running with config:
DATALOADER:
NUM_INSTANCE: 4
NUM_WORKERS: 8
SAMPLER: softmax_triplet
DATASETS:
NAMES: market1501
ROOT_DIR: /data/jhb_data/datasets
INPUT:
PADDING: 10
PIXEL_MEAN: [0.5, 0.5, 0.5]
PIXEL_STD: [0.5, 0.5, 0.5]
PROB: 0.5
RE_PROB: 0.5
SIZE_TEST: [256, 128]
SIZE_TRAIN: [256, 128]
MODEL:
ATT_DROP_RATE: 0.0
COS_LAYER: False
DEVICE: cuda
DEVICE_ID: 2
DEVIDE_LENGTH: 4
DIST_TRAIN: False
DROP_OUT: 0.0
DROP_PATH: 0.1
ID_LOSS_TYPE: softmax
ID_LOSS_WEIGHT: 1.0
IF_LABELSMOOTH: off
IF_WITH_CENTER: no
JPM: False
LAST_STRIDE: 1
METRIC_LOSS_TYPE: triplet
NAME: transformer
NECK: bnneck
NO_MARGIN: True
PRETRAIN_CHOICE: imagenet
PRETRAIN_PATH: /data/jhb_data/checkpoints/jx_vit_base_p16_224-80ecf9dd.pth
RE_ARRANGE: True
SHIFT_NUM: 5
SHUFFLE_GROUP: 2
SIE_CAMERA: True
SIE_COE: 3.0
SIE_VIEW: False
STRIDE_SIZE: [12, 12]
TRANSFORMER_TYPE: vit_base_patch16_224_TransReID
TRIPLET_LOSS_WEIGHT: 1.0
OUTPUT_DIR: ../logs/market_vit_transreid_stride
SOLVER:
BASE_LR: 0.008
BIAS_LR_FACTOR: 2
CENTER_LOSS_WEIGHT: 0.0005
CENTER_LR: 0.5
CHECKPOINT_PERIOD: 120
COSINE_MARGIN: 0.5
COSINE_SCALE: 30
EVAL_PERIOD: 120
GAMMA: 0.1
IMS_PER_BATCH: 64
LARGE_FC_LR: False
LOG_PERIOD: 50
MARGIN: 0.3
MAX_EPOCHS: 120
MOMENTUM: 0.9
OPTIMIZER_NAME: SGD
SEED: 1234
STEPS: (40, 70)
WARMUP_EPOCHS: 5
WARMUP_FACTOR: 0.01
WARMUP_METHOD: linear
WEIGHT_DECAY: 0.0001
WEIGHT_DECAY_BIAS: 0.0001
TEST:
DIST_MAT: dist_mat.npy
EVAL: True
FEAT_NORM: yes
IMS_PER_BATCH: 256
NECK_FEAT: before
RE_RANKING: False
WEIGHT: /data/jhb_data/checkpoints/vit_transreid_market.pth
=> Market1501 loaded
Dataset statistics:

subset | # ids | # images | # cameras

train | 751 | 12936 | 6
query | 750 | 3368 | 6
gallery | 751 | 15913 | 6

using Transformer_type: vit_base_patch16_224_TransReID as a backbone
using stride: [12, 12], and patch number is num_y21 * num_x10
camera number is : 6
using SIE_Lambda is : 3.0
using drop_out rate is : 0.0
using attn_drop_out rate is : 0.0
using drop_path rate is : 0.1`

And thera are so many the mismatched:

Warning:b1.0.norm1.weight not found in current model but in loaded dictionary. Warning:b1.0.norm1.bias not found in current model but in loaded dictionary. Warning:b1.0.attn.qkv.weight not found in current model but in loaded dictionary. Warning:b1.0.attn.qkv.bias not found in current model but in loaded dictionary. Warning:b1.0.attn.proj.weight not found in current model but in loaded dictionary. Warning:b1.0.attn.proj.bias not found in current model but in loaded dictionary. Warning:b1.0.norm2.weight not found in current model but in loaded dictionary. Warning:b1.0.norm2.bias not found in current model but in loaded dictionary. Warning:b1.0.mlp.fc1.weight not found in current model but in loaded dictionary. Warning:b1.0.mlp.fc1.bias not found in current model but in loaded dictionary. Warning:b1.0.mlp.fc2.weight not found in current model but in loaded dictionary. Warning:b1.0.mlp.fc2.bias not found in current model but in loaded dictionary. Warning:b1.1.weight not found in current model but in loaded dictionary. Warning:b1.1.bias not found in current model but in loaded dictionary. Warning:b2.0.norm1.weight not found in current model but in loaded dictionary. Warning:b2.0.norm1.bias not found in current model but in loaded dictionary. Warning:b2.0.attn.qkv.weight not found in current model but in loaded dictionary. Warning:b2.0.attn.qkv.bias not found in current model but in loaded dictionary. Warning:b2.0.attn.proj.weight not found in current model but in loaded dictionary. Warning:b2.0.attn.proj.bias not found in current model but in loaded dictionary. Warning:b2.0.norm2.weight not found in current model but in loaded dictionary. Warning:b2.0.norm2.bias not found in current model but in loaded dictionary. Warning:b2.0.mlp.fc1.weight not found in current model but in loaded dictionary. Warning:b2.0.mlp.fc1.bias not found in current model but in loaded dictionary. Warning:b2.0.mlp.fc2.weight not found in current model but in loaded dictionary. Warning:b2.0.mlp.fc2.bias not found in current model but in loaded dictionary. Warning:b2.1.weight not found in current model but in loaded dictionary. Warning:b2.1.bias not found in current model but in loaded dictionary. Warning:classifier_1.weight not found in current model but in loaded dictionary. Warning:classifier_2.weight not found in current model but in loaded dictionary. Warning:classifier_3.weight not found in current model but in loaded dictionary. Warning:classifier_4.weight not found in current model but in loaded dictionary. Warning:bottleneck_1.weight not found in current model but in loaded dictionary. Warning:bottleneck_1.bias not found in current model but in loaded dictionary. Warning:bottleneck_1.running_mean not found in current model but in loaded dictionary. Warning:bottleneck_1.running_var not found in current model but in loaded dictionary. Warning:bottleneck_1.num_batches_tracked not found in current model but in loaded dictionary. Warning:bottleneck_2.weight not found in current model but in loaded dictionary. Warning:bottleneck_2.bias not found in current model but in loaded dictionary. Warning:bottleneck_2.running_mean not found in current model but in loaded dictionary. Warning:bottleneck_2.running_var not found in current model but in loaded dictionary. Warning:bottleneck_2.num_batches_tracked not found in current model but in loaded dictionary. Warning:bottleneck_3.weight not found in current model but in loaded dictionary. Warning:bottleneck_3.bias not found in current model but in loaded dictionary. Warning:bottleneck_3.running_mean not found in current model but in loaded dictionary. Warning:bottleneck_3.running_var not found in current model but in loaded dictionary. Warning:bottleneck_3.num_batches_tracked not found in current model but in loaded dictionary. Warning:bottleneck_4.weight not found in current model but in loaded dictionary. Warning:bottleneck_4.bias not found in current model but in loaded dictionary. Warning:bottleneck_4.running_mean not found in current model but in loaded dictionary. Warning:bottleneck_4.running_var not found in current model but in loaded dictionary. Warning:bottleneck_4.num_batches_tracked not found in current model but in loaded dictionary.

However,in the end,I got the nice rank test as following:

My output is very similar to authors'.Well,is it a kind of luck?Wish my answer can help you.

[czhuo-57]

@KingH12138 您好，我在训练duke数据集以及自己采集的数据集，出现这个问题，尝试采用您发表那个解决方法以及各种方法都解决不掉，您可以为我解答解答吗？RuntimeError: The size of tensor a (128) must match the size of tensor b (768) at non-singleton dimension 3

[KingH12138]

@KingH12138 您好，我在训练duke数据集以及自己采集的数据集，出现这个问题，尝试采用您发表那个解决方法以及各种方法都解决不掉，您可以为我解答解答吗？RuntimeError: The size of tensor a (128) must match the size of tensor b (768) at non-singleton dimension 3

It's a usal mistake when you bulid your model. However, I can't judge reasons beacause of your inadequate mistake information. To my best, I think there's either something wrong internally with the forward function in your model, or there's something wrong with your logic for reading in the data. Of course, I think the former is more likely.

[czhuo-57]

@KingH12138 ，Ok,I know it,i will try to solve it,thank you~