关于ou-mvlp数据集训练了完准确率只有83.4，参数都保持了一致，可能是数据加载的问题吗？还是其他的问题？

[xusong-20]

250000
Transforming...
Evaluating...
Evaluation complete. Cost: 0:49:03.915056
===Rank-1 (Include identical-view cases)===
00: 84.270
---Rank-1 (Exclude identical view cases)---
00: 83.407
---Rank-1 of each angle (Exclude identical-view cases)---
00: [75.20 82.97 86.66 87.04 81.76 85.32 82.52 74.29 83.71 87.39 88.19 83.34 86.12 83.19]

config
'pid_num': 5153,
'pid_shuffle': False,
'hidden_dim': 256,
'lr': 1e-4,
'hard_or_full_trip': 'full',
'batch_size': (32, 16),
'restore_iter': 0,
'total_iter': 250000,
'margin': 0.2,
'num_workers': 2,
'frame_num': 30,
'model_name': 'GaitSet',
test

Exclude identical-view cases

def de_diag13(acc, each_angle=False):
result = np.sum(acc - np.diag(np.diag(acc)), 1) / 13.0
if not each_angle:
result = np.mean(result)
return result

[mapei319]

训练oumvlp得到83%，也很困惑

[w1998w]

老哥，请问在OUMVLP上训练的时候 除了按照README里面修改config里面的参数和basic_blocks、gaitset文件外还需要修改哪些地方，我其他地方没动然后运行一直提示我ValueError: Sample larger than population or is negative这个问题，感谢老哥！