hwz-zju / Instruct-ReID

A General-purpose Person Re-identification Task with Instructions
115 stars 4 forks source link

Warning: mismatch: classifier.weight torch.Size([2800, 768]) torch.Size([96, 768]) #20

Closed Enforcer03 closed 3 weeks ago

Enforcer03 commented 3 weeks ago

I have tried to test checkpoint li on the real2 dataset. I have gotten the above warning. Also can you provide with the file real2_v1.json. I have run the testing on runner_real1_v1_gpt.json

I have these results:

using stride: [16, 16], and patch number is num_y16 * num_x8
=> Loaded checkpoint '/media2/harshm/part2/Instruct-ReID/checkpoint_li.pth.tar'
mismatch: classifier.weight torch.Size([2800, 768]) torch.Size([96, 768])
mismatch: classifier_f.weight torch.Size([2800, 768]) torch.Size([96, 768])
mismatch: classifier_c.weight torch.Size([2800, 768]) torch.Size([96, 768])
missing keys in state_dict: {'classifier.weight', 'classifier_f.weight', 'classifier_c.weight'}
Test:
Extract Features: [50/75]   Time 1.526 (1.527)  Data 0.028 (0.036)  
/media2/harshm/part2/Instruct-ReID/reid/evaluation/evaluators.py:152: UserWarning: This overload of addmm_ is deprecated:
    addmm_(Number beta, Number alpha, Tensor mat1, Tensor mat2)
Consider using one of the following signatures instead:
    addmm_(Tensor mat1, Tensor mat2, *, Number beta, Number alpha) (Triggered internally at  ../torch/csrc/utils/python_arg_parser.cpp:1050.)
  dist_m.addmm_(1, -2, x, y.t())
================================================================================
Mean AP: 36.4%
CMC Scores:
  top-1          82.7%
  top-5          93.7%
  top-10         96.9%
  top-20         98.4%
  top-50         99.5%