student_videomobilevit_dist_kl_temp_1_alpha_0.6_teacher_mvit_v2_s_RWF2000

This model is a fine-tuned version of Video-MobileVit on an Real World Fight 2000 (RWF2000). It achieves the following results on the evaluation set:

• Loss: 0.6388
• Accuracy: 0.7488
• F1: 0.7471
• Precision: 0.7555

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 20
• eval_batch_size: 20
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 180
• training_steps: 1800
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision
0.5041	2.02	180	0.4884	0.7625	0.7604	0.7721
0.3788	4.04	360	0.4183	0.8094	0.8091	0.8114
0.3226	7.02	540	0.4292	0.8187	0.8175	0.8274
0.2425	9.04	720	0.4234	0.8094	0.8073	0.8231
0.243	12.02	900	0.3590	0.8313	0.8311	0.8325
0.1914	14.04	1080	0.3708	0.8375	0.8373	0.8388
0.1832	17.02	1260	0.3755	0.8469	0.8469	0.8470
0.1624	19.04	1440	0.4323	0.8375	0.8366	0.8453

Framework versions

• Transformers 4.45.2
• Pytorch 2.0.1+cu118
• Datasets 3.0.1
• Tokenizers 0.20.0