Something Wrong and i got 100% accuracy🥵

as a beginner of cv, i trained this model on a tesla P4 sever.

in training, i use umd dataset , and reset the parameter after copying configs.
```from easydict import EasyDict as edict
config = edict()
config.margin_list = (1.0, 0.5, 0.0)
config.network = "r100"
config.resume = False
config.output = None
config.embedding_size = 512
config.sample_rate = 1.0
config.fp16 = True
config.momentum = 0.9
config.weight_decay = 5e-4
config.batch_size = 6
config.lr = 0.1
config.verbose = 2000
config.dali = False
config.rec = "../_datasets_/faces_umd"
config.num_classes = 10575
config.num_image = 494414
config.num_epoch = 5
config.warmup_epoch = 0
config.val_targets = ['lfw', 'cfp_fp', "agedb_30"]
```

the log file show the result like this:[training.log](https://dropover.cloud/79bdec)
and the model is here: [model](https://dropover.cloud/859387)

when i want to use the model on few pics selected from IJBC, the result came 100%.
```CUDA_VISIBLE_DEVICES=0 python eval_ijbc.py --model-prefix work_dirs/face_umd_r100/model.pt --image-path IJBC --result-dir work_dirs/face_umd_r100/result --batch-size 100 --job face_umd_r100 --target IJBC --network r100
Time: 0.00 s.
Time: 0.00 s.
files: 300
batch 0
batch 1
batch 2
Time: 9.90 s.
Feature Shape: (300 , 1024) .
(300, 512) (300,)
Finish Calculating 0 template features.
Time: 0.01 s.
Finish 0/1 pairs.
Time: 0.00 s.
+-----------+--------+--------+--------+--------+--------+--------+
| Methods | 1e-06 | 1e-05 | 0.0001 | 0.001 | 0.01 | 0.1 |
+-----------+--------+--------+--------+--------+--------+--------+
| ijbc-IJBC | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 |
+-----------+--------+--------+--------+--------+--------+--------+```

i couldn't figure it out , and hope someone had solutions.