videomae-base-finetuned-sign-subset

This model is a fine-tuned version of MCG-NJU/videomae-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.3672
• Accuracy: 0.1905

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: 5e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 270
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.04	11	2.4220	0.0870
2.3491	1.04	22	2.6315	0.0
2.3491	2.04	33	2.6680	0.0435
2.2285	3.04	44	2.8487	0.1304
2.2285	4.04	55	3.0361	0.0870
1.996	5.04	66	3.0258	0.1304
1.996	6.04	77	3.2125	0.1304
1.6956	7.04	88	3.2063	0.1304
1.6956	8.04	99	3.1919	0.1304
1.5088	9.04	110	3.1940	0.1304
1.3777	10.04	121	3.3180	0.1739
1.3777	11.04	132	3.3112	0.1304
1.1509	12.04	143	3.3400	0.1304
1.1509	13.04	154	3.2550	0.1739
0.9036	14.04	165	3.3682	0.1304
0.9036	15.04	176	3.3775	0.1304
0.8303	16.04	187	3.4701	0.1304
0.8303	17.04	198	3.4340	0.1739
0.6683	18.04	209	3.4843	0.1304
0.5126	19.04	220	3.3552	0.2174
0.5126	20.04	231	3.3702	0.2609
0.3728	21.04	242	3.3871	0.2609
0.3728	22.04	253	3.3565	0.2609
0.3291	23.04	264	3.3861	0.3043
0.3291	24.02	270	3.3876	0.3043

Framework versions

• Transformers 4.26.1
• Pytorch 1.13.1+cu116
• Datasets 2.9.0
• Tokenizers 0.13.2