-
Couldn't load subscription status.
- Fork 386
Description
由于EuroCup2016数据集下载链接有问题,我便仿照标签格式自己做了一个小样本数据集,格式格式如下:
{
"fps": 5,
"gts": [
{
"url": "aaa.mp4",
"total_frames": 4369,
"actions": [
{
"label_ids": [
1
],
"label_names": [
"挥拍"
],
"start_id": 0,
"end_id": 598
},
{
"label_ids": [
2
],
"label_names": [
"接球"
],
"start_id": 598,
"end_id": 1508
},
{
"label_ids": [
2
],
"label_names": [
"接球"
],
"start_id": 1508,
"end_id": 2279
},
{
"label_ids": [
1
],
"label_names": [
"挥拍"
],
"start_id": 2279,
"end_id": 3147
},
{
"label_ids": [
2
],
"label_names": [
"接球"
],
"start_id": 3147,
"end_id": 4369
}
]
},
...
]
}
处理好数据后进行训练出现如下报错:
Loading bb_3_2.conv2._batch_norm._variance:
100%|██████████| 275/275 [00:02<00:00, 94.68it/s]
Loading bb_3_2.conv2._batch_norm._variance:
Traceback (most recent call last):
File "main.py", line 124, in
main()
File "main.py", line 120, in main
profiler_options=args.profiler_options)
File "/home/aistudio/PaddleVideo/paddlevideo/tasks/train.py", line 249, in train_model
record_list["batch_time"].sum)
ZeroDivisionError: division by zero
INFO 2025-01-27 11:47:44,217 launch_utils.py:341] terminate all the procs
ERROR 2025-01-27 11:47:44,217 launch_utils.py:604] ABORT!!! Out of all 1 trainers, the trainer process with rank=[0] was aborted. Please check its log.
INFO 2025-01-27 11:47:48,221 launch_utils.py:341] terminate all the procs
INFO 2025-01-27 11:47:48,221 launch.py:311] Local processes completed.