PaddleVideo代码与训练测试数据

[10/22 19:10:39] DALI is not installed, you can improve performance if use DALI [10/22 19:10:39] DATASET : [10/22 19:10:39] batch_size : 16 [10/22 19:10:39] num_workers : 4 [10/22 19:10:39] test : [10/22 19:10:39] file_path : /home/aistudio/data/data113093/test_data_joint_motion.npy [10/22 19:10:39] format : SkeletonDataset [10/22 19:10:39] test_mode : True [10/22 19:10:39] test_batch_size : 1 [10/22 19:10:39] test_num_workers : 0 [10/22 19:10:39] train : [10/22 19:10:39] file_path : /home/aistudio/data/data113093/train_data_joint_motion.npy [10/22 19:10:39] format : SkeletonDataset [10/22 19:10:39] label_path : /home/aistudio/data/data113093/train_label.npy [10/22 19:10:39] ------------------------------------------------------------ [10/22 19:10:39] INFERENCE : [10/22 19:10:39] name : STGCN_Inference_helper [10/22 19:10:39] num_channels : 2 [10/22 19:10:39] person_nums : 1 [10/22 19:10:39] vertex_nums : 25 [10/22 19:10:39] window_size : 1500 [10/22 19:10:39] ------------------------------------------------------------ [10/22 19:10:39] METRIC : [10/22 19:10:39] name : SkeletonMetric [10/22 19:10:39] out_file : submission.csv [10/22 19:10:39] ------------------------------------------------------------ [10/22 19:10:39] MIX : [10/22 19:10:39] alpha : 0.5 [10/22 19:10:39] name : Mixup [10/22 19:10:39] ------------------------------------------------------------ [10/22 19:10:39] MODEL : [10/22 19:10:39] backbone : [10/22 19:10:39] name : AGCN [10/22 19:10:39] framework : RecognizerGCN [10/22 19:10:39] head : [10/22 19:10:39] ls_eps : 0.1 [10/22 19:10:39] name : STGCNHead [10/22 19:10:39] num_classes : 30 [10/22 19:10:39] ------------------------------------------------------------ [10/22 19:10:39] OPTIMIZER : [10/22 19:10:39] learning_rate : [10/22 19:10:39] cosine_base_lr : 0.05 [10/22 19:10:39] iter_step : True [10/22 19:10:39] max_epoch : 100 [10/22 19:10:39] name : CustomWarmupCosineDecay [10/22 19:10:39] warmup_epochs : 10 [10/22 19:10:39] warmup_start_lr : 0.005 [10/22 19:10:39] momentum : 0.9 [10/22 19:10:39] name : Momentum [10/22 19:10:39] weight_decay : [10/22 19:10:39] name : L2 [10/22 19:10:39] value : 0.0001 [10/22 19:10:39] ------------------------------------------------------------ [10/22 19:10:39] PIPELINE : [10/22 19:10:39] test : [10/22 19:10:39] sample : [10/22 19:10:39] name : AutoPadding [10/22 19:10:39] window_size : 1500 [10/22 19:10:39] transform : [10/22 19:10:39] SkeletonNorm : None [10/22 19:10:39] train : [10/22 19:10:39] sample : [10/22 19:10:39] name : AutoPadding [10/22 19:10:39] window_size : 1500 [10/22 19:10:39] transform : [10/22 19:10:39] SkeletonNorm : None [10/22 19:10:39] ------------------------------------------------------------ [10/22 19:10:39] PRECISEBN : [10/22 19:10:39] num_iters_preciseBN : 200 [10/22 19:10:39] preciseBN_interval : 5 [10/22 19:10:39] ------------------------------------------------------------ [10/22 19:10:39] epochs : 100 [10/22 19:10:39] log_interval : 10 [10/22 19:10:39] model_name : AGCNJOINT [10/22 19:10:45] Loading data, it will take some moment... [10/22 19:10:47] Data Loaded! [10/22 19:10:50] epoch:[ 1/100] train step:0  loss: 3.40256 lr: 0.005000 top1: 0.06250 top5: 0.23423 batch_cost: 2.98685 sec, reader_cost: 1.06565 sec, ips: 5.35681 instance/sec. [10/22 19:11:09] epoch:[ 1/100] train step:10  loss: 3.40841 lr: 0.005241 top1: 0.00000 top5: 0.06250 batch_cost: 1.90822 sec, reader_cost: 0.00034 sec, ips: 8.38477 instance/sec. [10/22 19:11:28] epoch:[ 1/100] train step:20  loss: 3.27396 lr: 0.005481 top1: 0.00000 top5: 0.37425 batch_cost: 1.90838 sec, reader_cost: 0.00041 sec, ips: 8.38406 instance/sec. [10/22 19:11:47] epoch:[ 1/100] train step:30  loss: 3.21615 lr: 0.005722 top1: 0.09308 top5: 0.34308 batch_cost: 1.90380 sec, reader_cost: 0.00034 sec, ips: 8.40423 instance/sec. [10/22 19:12:06] epoch:[ 1/100] train step:40  loss: 3.18168 lr: 0.005962 top1: 0.06250 top5: 0.43353 batch_cost: 1.91365 sec, reader_cost: 0.00035 sec, ips: 8.36097 instance/sec. [10/22 19:12:25] epoch:[ 1/100] train step:50  loss: 3.14297 lr: 0.006203 top1: 0.11630 top5: 0.40507 batch_cost: 1.90744 sec, reader_cost: 0.00040 sec, ips: 8.38822 instance/sec. [10/22 19:12:44] epoch:[ 1/100] train step:60  loss: 3.26410 lr: 0.006443 top1: 0.00000 top5: 0.24817 batch_cost: 1.92309 sec, reader_cost: 0.00043 sec, ips: 8.31996 instance/sec. [10/22 19:13:04] epoch:[ 1/100] train step:70  loss: 3.13893 lr: 0.006684 top1: 0.17970 top5: 0.53911 batch_cost: 1.91474 sec, reader_cost: 0.00035 sec, ips: 8.35622 instance/sec. [10/22 19:13:23] epoch:[ 1/100] train step:80  loss: 3.11981 lr: 0.006924 top1: 0.09465 top5: 0.48661 batch_cost: 1.91216 sec, reader_cost: 0.00042 sec, ips: 8.36751 instance/sec. [10/22 19:13:42] epoch:[ 1/100] train step:90  loss: 3.05547 lr: 0.007165 top1: 0.05487 top5: 0.52437 batch_cost: 1.92082 sec, reader_cost: 0.00034 sec, ips: 8.32978 instance/sec. [10/22 19:14:01] epoch:[ 1/100] train step:100  loss: 3.20427 lr: 0.007405 top1: 0.01033 top5: 0.43233 batch_cost: 1.91370 sec, reader_cost: 0.00035 sec, ips: 8.36075 instance/sec. [10/22 19:14:20] epoch:[ 1/100] train step:110  loss: 2.97483 lr: 0.007646 top1: 0.18749 top5: 0.49998 batch_cost: 1.91005 sec, reader_cost: 0.00037 sec, ips: 8.37675 instance/sec. [10/22 19:14:39] epoch:[ 1/100] train step:120  loss: 2.98267 lr: 0.007886 top1: 0.16796 top5: 0.50773 batch_cost: 1.92207 sec, reader_cost: 0.00035 sec, ips: 8.32435 instance/sec. [10/22 19:14:58] epoch:[ 1/100] train step:130  loss: 3.12564 lr: 0.008127 top1: 0.19422 top5: 0.38844 batch_cost: 1.91350 sec, reader_cost: 0.00034 sec, ips: 8.36164 instance/sec. [10/22 19:15:18] epoch:[ 1/100] train step:140  loss: 3.08088 lr: 0.008367 top1: 0.12272 top5: 0.55680 batch_cost: 1.92860 sec, reader_cost: 0.00035 sec, ips: 8.29618 instance/sec. [10/22 19:15:37] epoch:[ 1/100] train step:150  loss: 2.96247 lr: 0.008608 top1: 0.18329 top5: 0.43329 batch_cost: 1.91313 sec, reader_cost: 0.00036 sec, ips: 8.36325 instance/sec. [10/22 19:15:56] epoch:[ 1/100] tra