尝试复现MSMT17 vit_base 结果时，遇到报错： RuntimeError: The size of tensor a (751) must match the size of tensor b (1041) at non-singleton dimension 0

[superTongTong]

您好！当我尝试复现MSMT17的结果遇到问题。下面是我尝试的命令：
python test.py --config_file configs/MSMT17/vit_base.yml MODEL.DEVICE_ID "('0')" TEST.WEIGHT 'vit_base_msmt.pth

遇到了报错：

File "D:\yao_pytorch\TransReID-main\model\make_model.py", line 206, in load_param
    self.state_dict()[i.replace('module.', '')].copy_(param_dict[i])
RuntimeError: The size of tensor a (751) must match the size of tensor b (1041) at non-singleton dimension 0

之后我尝试在 make_model.py 添加 print(i, self.state_dict()[i.replace('module.', '')].shape, param_dict[i].shape) 得到如下结果：

base.norm.weight torch.Size([768]) torch.Size([768])
base.norm.bias torch.Size([768]) torch.Size([768])
base.fc.weight torch.Size([1000, 768]) torch.Size([1000, 768])
base.fc.bias torch.Size([1000]) torch.Size([1000])
classifier.weight torch.Size([751, 768]) torch.Size([1041, 768])
Traceback (most recent call last):
File "D:\yao_pytorch\TransReID-main\test.py", line 46, in <module>
model.load_param(cfg.TEST.WEIGHT)
File "D:\yao_pytorch\TransReID-main\model\make_model.py", line 206, in load_param
self.state_dict()[i.replace('module.', '')].copy_(param_dict[i])
RuntimeError: The size of tensor a (751) must match the size of tensor b (1041) at non-singleton dimension 0

求教如何解决？ 非常感谢。

[kp97524]

Hi @superTongTong ,

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