[07/08/2024 00:03:41] {train.py:87} INFO - Config file loaded from src/configs/training/swin3d_t.yaml
[07/08/2024 00:03:41] {train.py:93} INFO - Config file saved to experiments/test
[07/08/2024 00:03:41] {train.py:39} INFO - DataConfig(dataset='vsl_400', modality='rgb', subset='cam_3', data_dir='data/processed/vsl_400', transform=TransformConfig(horizontal_flip_prob=0.5, aug_type='augmix', aug_paras={'alpha': 1.0, 'depth': -1, 'magnitude': 3, 'width': 5}, sample_rate=2, normalization=True, random_choose=False, random_shift=False, random_move=False, window_size=-1, random_mirror=False, random_mirror_p=0.5), fps=30, debug=True, use_mmap=True, is_vector=False)
[07/08/2024 00:03:41] {train.py:41} INFO - ModelConfig(arch='swin3d_t', pretrained='DEFAULT', num_frozen_layers=0, num_frames=16, num_points=27, num_people=1, groups=8, block_size=41, in_channels=3, labeling_mode='spatial')
[07/08/2024 00:03:41] {train.py:43} INFO - TrainingConfig(output_dir='experiments/test', remove_unused_columns=False, do_train=True, eval_strategy='epoch', logging_strategy='epoch', save_strategy='epoch', logging_steps=1, save_steps=1, eval_steps=1, learning_rate=0.0001, weight_decay=0.01, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, warmup_ratio=0.1, num_train_epochs=1, per_device_train_batch_size=1, per_device_eval_batch_size=1, dataloader_num_workers=0, load_best_model_at_end=True, metric_for_best_model='accuracy', resume_from_checkpoint=None, run_name='test', report_to=None, push_to_hub=True, hub_model_id='vsltranslation/test', hub_strategy='checkpoint', hub_private_repo=True)
[07/08/2024 00:03:42] {train.py:46} INFO - VSL_400 dataset loaded
[07/08/2024 00:03:44] {train.py:53} INFO - swin3d_t model loaded from DEFAULT
[07/08/2024 00:03:44] {train.py:57} INFO - Splits created for training and evaluation
[07/08/2024 00:03:44] {train.py:58} INFO - Number of samples in training set: 10
[07/08/2024 00:03:44] {train.py:59} INFO - Number of samples in validation set: 10
[07/08/2024 00:03:44] {train.py:76} INFO - Trainer created
[07/08/2024 00:03:44] {train.py:78} INFO - Training started
[07/08/2024 00:04:43] {loggers.py:9} INFO - {'loss': 8.2304, 'grad_norm': 28.548147201538086, 'learning_rate': 0.0, 'epoch': 1.0}
[07/08/2024 00:04:59] {loggers.py:9} INFO - {'eval_loss': 6.429128170013428, 'eval_accuracy': 0.0, 'eval_f1': 0.0, 'eval_recall': 0.0, 'eval_precision': 0.0, 'eval_runtime': 15.7113, 'eval_samples_per_second': 1.273, 'eval_steps_per_second': 1.273, 'epoch': 1.0}
[07/08/2024 00:05:06] {loggers.py:9} INFO - {'train_runtime': 81.714, 'train_samples_per_second': 0.122, 'train_steps_per_second': 0.122, 'total_flos': 0.0, 'train_loss': 8.230411529541016, 'epoch': 1.0}
[07/08/2024 00:05:12] {train.py:80} INFO - Training completed
[07/08/2024 00:06:58] {train.py:87} INFO - Config file loaded from src/configs/training/swin3d_t.yaml
[07/08/2024 00:06:58] {train.py:93} INFO - Config file saved to experiments/test
[07/08/2024 00:06:58] {train.py:39} INFO - DataConfig(dataset='vsl_400', modality='rgb', subset='cam_3', data_dir='data/processed/vsl_400', transform=TransformConfig(horizontal_flip_prob=0.5, aug_type='augmix', aug_paras={'alpha': 1.0, 'depth': -1, 'magnitude': 3, 'width': 5}, sample_rate=2, normalization=True, random_choose=False, random_shift=False, random_move=False, window_size=-1, random_mirror=False, random_mirror_p=0.5), fps=30, debug=True, use_mmap=True, is_vector=False)
[07/08/2024 00:06:58] {train.py:41} INFO - ModelConfig(arch='swin3d_t', pretrained='DEFAULT', num_frozen_layers=0, num_frames=16, num_points=27, num_people=1, groups=8, block_size=41, in_channels=3, labeling_mode='spatial')
[07/08/2024 00:06:58] {train.py:43} INFO - TrainingConfig(output_dir='experiments/test', remove_unused_columns=False, do_train=True, eval_strategy='epoch', logging_strategy='epoch', save_strategy='epoch', logging_steps=1, save_steps=1, eval_steps=1, learning_rate=0.0001, weight_decay=0.01, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, warmup_ratio=0.1, num_train_epochs=1, per_device_train_batch_size=1, per_device_eval_batch_size=1, dataloader_num_workers=0, load_best_model_at_end=True, metric_for_best_model='accuracy', resume_from_checkpoint=None, run_name='test', report_to=None, push_to_hub=True, hub_model_id='vsltranslation/test', hub_strategy='checkpoint', hub_private_repo=True)
[07/08/2024 00:06:59] {train.py:46} INFO - VSL_400 dataset loaded
[07/08/2024 00:06:59] {tools.py:101} INFO - Registering swin3d_t classes
[07/08/2024 00:21:41] {train.py:87} INFO - Config file loaded from src/configs/training/swin3d_t.yaml
[07/08/2024 00:21:41] {train.py:93} INFO - Config file saved to experiments/test
[07/08/2024 00:21:41] {train.py:39} INFO - DataConfig(dataset='vsl_400', modality='rgb', subset='cam_3', data_dir='data/processed/vsl_400', transform=TransformConfig(horizontal_flip_prob=0.5, aug_type='augmix', aug_paras={'alpha': 1.0, 'depth': -1, 'magnitude': 3, 'width': 5}, sample_rate=2, normalization=True, random_choose=False, random_shift=False, random_move=False, window_size=-1, random_mirror=False, random_mirror_p=0.5), fps=30, debug=True, use_mmap=True, is_vector=False)
[07/08/2024 00:21:41] {train.py:41} INFO - ModelConfig(arch='swin3d_t', pretrained='DEFAULT', num_frozen_layers=0, num_frames=16, num_points=27, num_people=1, groups=8, block_size=41, in_channels=3, labeling_mode='spatial')
[07/08/2024 00:21:41] {train.py:43} INFO - TrainingConfig(output_dir='experiments/test', remove_unused_columns=False, do_train=True, eval_strategy='epoch', logging_strategy='epoch', save_strategy='epoch', logging_steps=1, save_steps=1, eval_steps=1, learning_rate=0.0001, weight_decay=0.01, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, warmup_ratio=0.1, num_train_epochs=1, per_device_train_batch_size=1, per_device_eval_batch_size=1, dataloader_num_workers=0, load_best_model_at_end=True, metric_for_best_model='accuracy', resume_from_checkpoint=None, run_name='test', report_to=None, push_to_hub=True, hub_model_id='vsltranslation/test', hub_strategy='checkpoint', hub_private_repo=True)
[07/08/2024 00:21:42] {train.py:46} INFO - VSL_400 dataset loaded
[07/08/2024 00:21:43] {tools.py:101} INFO - Registering swin3d_t classes
[07/08/2024 00:21:44] {train.py:53} INFO - swin3d_t model loaded from DEFAULT
[07/08/2024 00:21:44] {train.py:57} INFO - Splits created for training and evaluation
[07/08/2024 00:21:44] {train.py:58} INFO - Number of samples in training set: 10
[07/08/2024 00:21:44] {train.py:59} INFO - Number of samples in validation set: 10
[07/08/2024 00:21:45] {train.py:76} INFO - Trainer created
[07/08/2024 00:21:45] {train.py:78} INFO - Training started
[07/08/2024 00:22:41] {loggers.py:9} INFO - {'loss': 8.2304, 'grad_norm': 28.548147201538086, 'learning_rate': 0.0, 'epoch': 1.0}
[07/08/2024 00:22:57] {loggers.py:9} INFO - {'eval_loss': 6.429128170013428, 'eval_accuracy': 0.0, 'eval_f1': 0.0, 'eval_recall': 0.0, 'eval_precision': 0.0, 'eval_runtime': 16.6224, 'eval_samples_per_second': 1.203, 'eval_steps_per_second': 1.203, 'epoch': 1.0}
[07/08/2024 00:23:04] {loggers.py:9} INFO - {'train_runtime': 79.4118, 'train_samples_per_second': 0.126, 'train_steps_per_second': 0.126, 'total_flos': 0.0, 'train_loss': 8.230411529541016, 'epoch': 1.0}
[07/08/2024 00:23:10] {train.py:80} INFO - Training completed
[07/08/2024 00:23:49] {train.py:87} INFO - Config file loaded from src/configs/training/swin3d_t.yaml
[07/08/2024 00:23:49] {train.py:93} INFO - Config file saved to experiments/test
[07/08/2024 00:23:49] {train.py:39} INFO - DataConfig(dataset='vsl_400', modality='rgb', subset='cam_3', data_dir='data/processed/vsl_400', transform=TransformConfig(horizontal_flip_prob=0.5, aug_type='augmix', aug_paras={'alpha': 1.0, 'depth': -1, 'magnitude': 3, 'width': 5}, sample_rate=2, normalization=True, random_choose=False, random_shift=False, random_move=False, window_size=-1, random_mirror=False, random_mirror_p=0.5), fps=30, debug=True, use_mmap=True, is_vector=False)
[07/08/2024 00:23:49] {train.py:41} INFO - ModelConfig(arch='swin3d_t', pretrained='DEFAULT', num_frozen_layers=0, num_frames=16, num_points=27, num_people=1, groups=8, block_size=41, in_channels=3, labeling_mode='spatial')
[07/08/2024 00:23:49] {train.py:43} INFO - TrainingConfig(output_dir='experiments/test', remove_unused_columns=False, do_train=True, eval_strategy='epoch', logging_strategy='epoch', save_strategy='epoch', logging_steps=1, save_steps=1, eval_steps=1, learning_rate=0.0001, weight_decay=0.01, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, warmup_ratio=0.1, num_train_epochs=2, per_device_train_batch_size=1, per_device_eval_batch_size=1, dataloader_num_workers=0, load_best_model_at_end=True, metric_for_best_model='accuracy', resume_from_checkpoint=None, run_name='test', report_to=None, push_to_hub=True, hub_model_id='vsltranslation/test', hub_strategy='checkpoint', hub_private_repo=True)
[07/08/2024 00:23:50] {train.py:46} INFO - VSL_400 dataset loaded
[07/08/2024 00:23:50] {tools.py:101} INFO - Registering swin3d_t classes
[07/08/2024 00:23:51] {train.py:53} INFO - swin3d_t model loaded from DEFAULT
[07/08/2024 00:23:51] {train.py:57} INFO - Splits created for training and evaluation
[07/08/2024 00:23:51] {train.py:58} INFO - Number of samples in training set: 10
[07/08/2024 00:23:51] {train.py:59} INFO - Number of samples in validation set: 10
[07/08/2024 00:23:51] {train.py:76} INFO - Trainer created
[07/08/2024 00:23:51] {train.py:78} INFO - Training started
[07/08/2024 00:24:48] {loggers.py:9} INFO - {'loss': 8.3615, 'grad_norm': 22.119415283203125, 'learning_rate': 5.555555555555556e-05, 'epoch': 0.5}
[07/08/2024 00:25:04] {loggers.py:9} INFO - {'eval_loss': 6.366983890533447, 'eval_accuracy': 0.0, 'eval_f1': 0.0, 'eval_recall': 0.0, 'eval_precision': 0.0, 'eval_runtime': 16.3982, 'eval_samples_per_second': 1.22, 'eval_steps_per_second': 1.22, 'epoch': 0.5}
[07/08/2024 00:26:05] {loggers.py:9} INFO - {'loss': 5.1607, 'grad_norm': 23.26081085205078, 'learning_rate': 0.0, 'epoch': 1.5}
[07/08/2024 00:26:23] {loggers.py:9} INFO - {'eval_loss': 6.410016059875488, 'eval_accuracy': 0.0, 'eval_f1': 0.0, 'eval_recall': 0.0, 'eval_precision': 0.0, 'eval_runtime': 18.4764, 'eval_samples_per_second': 1.082, 'eval_steps_per_second': 1.082, 'epoch': 1.5}
[07/08/2024 00:26:30] {loggers.py:9} INFO - {'train_runtime': 158.637, 'train_samples_per_second': 0.126, 'train_steps_per_second': 0.126, 'total_flos': 0.0, 'train_loss': 6.76112995147705, 'epoch': 1.5}
[07/08/2024 00:27:27] {train.py:80} INFO - Training completed
[07/08/2024 00:34:50] {train.py:88} INFO - Config file loaded from src/configs/training/swin3d_t.yaml
[07/08/2024 00:34:50] {train.py:94} INFO - Config file saved to experiments/test
[07/08/2024 00:34:50] {train.py:39} INFO - DataConfig(dataset='vsl_400', modality='rgb', subset='cam_3', data_dir='data/processed/vsl_400', transform=TransformConfig(horizontal_flip_prob=0.5, aug_type='augmix', aug_paras={'alpha': 1.0, 'depth': -1, 'magnitude': 3, 'width': 5}, sample_rate=2, normalization=True, random_choose=False, random_shift=False, random_move=False, window_size=-1, random_mirror=False, random_mirror_p=0.5), fps=30, debug=True, use_mmap=True, is_vector=False)
[07/08/2024 00:34:50] {train.py:41} INFO - ModelConfig(arch='swin3d_t', pretrained='DEFAULT', num_frozen_layers=0, num_frames=16, num_points=27, num_people=1, groups=8, block_size=41, in_channels=3, labeling_mode='spatial')
[07/08/2024 00:34:50] {train.py:43} INFO - TrainingConfig(output_dir='experiments/test', remove_unused_columns=False, do_train=True, eval_strategy='epoch', logging_strategy='epoch', save_strategy='epoch', logging_steps=1, save_steps=1, eval_steps=1, learning_rate=0.0001, weight_decay=0.01, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, warmup_ratio=0.1, num_train_epochs=2, per_device_train_batch_size=1, per_device_eval_batch_size=1, dataloader_num_workers=0, load_best_model_at_end=True, metric_for_best_model='accuracy', resume_from_checkpoint=None, run_name='test', report_to=None, push_to_hub=True, hub_model_id='vsltranslation/test', hub_strategy='checkpoint', hub_private_repo=True)
[07/08/2024 00:34:51] {train.py:46} INFO - VSL_400 dataset loaded
[07/08/2024 00:34:51] {tools.py:101} INFO - swin3d_t classes registered
[07/08/2024 00:34:53] {train.py:53} INFO - swin3d_t model loaded from DEFAULT
[07/08/2024 00:34:53] {train.py:57} INFO - Splits created for training and evaluation
[07/08/2024 00:34:53] {train.py:58} INFO - Number of samples in training set: 10
[07/08/2024 00:34:53] {train.py:59} INFO - Number of samples in validation set: 10
[07/08/2024 00:34:53] {train.py:77} INFO - Trainer created
[07/08/2024 00:34:53] {train.py:79} INFO - Training started
[07/08/2024 00:36:00] {loggers.py:9} INFO - {'loss': 8.3615, 'grad_norm': 22.119415283203125, 'learning_rate': 5.555555555555556e-05, 'epoch': 0.5}
[07/08/2024 00:36:27] {loggers.py:9} INFO - {'eval_loss': 6.366983890533447, 'eval_accuracy': 0.0, 'eval_f1': 0.0, 'eval_recall': 0.0, 'eval_precision': 0.0, 'eval_runtime': 27.291, 'eval_samples_per_second': 0.733, 'eval_steps_per_second': 0.733, 'epoch': 0.5}
[07/08/2024 00:37:49] {loggers.py:9} INFO - {'loss': 5.1607, 'grad_norm': 23.26081085205078, 'learning_rate': 0.0, 'epoch': 1.5}
[07/08/2024 00:38:09] {loggers.py:9} INFO - {'eval_loss': 6.410016059875488, 'eval_accuracy': 0.0, 'eval_f1': 0.0, 'eval_recall': 0.0, 'eval_precision': 0.0, 'eval_runtime': 20.5102, 'eval_samples_per_second': 0.975, 'eval_steps_per_second': 0.975, 'epoch': 1.5}
[07/08/2024 00:38:18] {loggers.py:9} INFO - {'train_runtime': 205.3053, 'train_samples_per_second': 0.097, 'train_steps_per_second': 0.097, 'total_flos': 0.0, 'train_loss': 6.76112995147705, 'epoch': 1.5}
[07/08/2024 00:38:21] {train.py:81} INFO - Training completed
[07/08/2024 00:55:48] {train.py:88} INFO - Config file loaded from src/configs/training/videomae_s.yaml
[07/08/2024 00:55:48] {train.py:94} INFO - Config file saved to experiments/test
[07/08/2024 00:55:48] {train.py:39} INFO - DataConfig(dataset='vsl_400', modality='rgb', subset='cam_3', data_dir='data/processed/vsl_400', transform=TransformConfig(horizontal_flip_prob=0.5, aug_type='augmix', aug_paras={'alpha': 1.0, 'depth': -1, 'magnitude': 3, 'width': 5}, sample_rate=4, normalization=True, random_choose=False, random_shift=False, random_move=False, window_size=-1, random_mirror=False, random_mirror_p=0.5), fps=30, debug=True, use_mmap=True, is_vector=False)
[07/08/2024 00:55:48] {train.py:41} INFO - ModelConfig(arch='videomae', pretrained='MCG-NJU/videomae-small-finetuned-kinetics', num_frozen_layers=0, num_frames=16, num_points=27, num_people=1, groups=8, block_size=41, in_channels=3, labeling_mode='spatial')
[07/08/2024 00:55:48] {train.py:43} INFO - TrainingConfig(output_dir='experiments/test', remove_unused_columns=False, do_train=True, eval_strategy='epoch', logging_strategy='epoch', save_strategy='epoch', logging_steps=1, save_steps=1, eval_steps=1, learning_rate=5e-05, weight_decay=0, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, warmup_ratio=0.1, num_train_epochs=1, per_device_train_batch_size=1, per_device_eval_batch_size=1, dataloader_num_workers=0, load_best_model_at_end=True, metric_for_best_model='accuracy', resume_from_checkpoint=None, run_name='test', report_to=None, push_to_hub=True, hub_model_id='vsltranslation/test', hub_strategy='checkpoint', hub_private_repo=True)
[07/08/2024 00:55:49] {train.py:46} INFO - VSL_400 dataset loaded
[07/08/2024 00:55:49] {tools.py:101} INFO - videomae classes registered
[07/08/2024 00:55:53] {train.py:53} INFO - videomae model loaded from MCG-NJU/videomae-small-finetuned-kinetics
[07/08/2024 00:55:53] {train.py:57} INFO - Splits created for training and evaluation
[07/08/2024 00:55:53] {train.py:58} INFO - Number of samples in training set: 10
[07/08/2024 00:55:53] {train.py:59} INFO - Number of samples in validation set: 10
[07/08/2024 00:55:53] {train.py:77} INFO - Trainer created
[07/08/2024 00:55:53] {train.py:79} INFO - Training started
[07/08/2024 00:57:09] {loggers.py:9} INFO - {'loss': 7.4362, 'grad_norm': 37.577880859375, 'learning_rate': 0.0, 'epoch': 1.0}
[07/08/2024 00:57:25] {loggers.py:9} INFO - {'eval_loss': 6.594984531402588, 'eval_accuracy': 0.0, 'eval_f1': 0.0, 'eval_recall': 0.0, 'eval_precision': 0.0, 'eval_runtime': 16.241, 'eval_samples_per_second': 0.616, 'eval_steps_per_second': 0.616, 'epoch': 1.0}
[07/08/2024 00:57:31] {loggers.py:9} INFO - {'train_runtime': 97.6475, 'train_samples_per_second': 0.102, 'train_steps_per_second': 0.102, 'total_flos': 0.0, 'train_loss': 7.436226654052734, 'epoch': 1.0}