mit-b2-fv-finetuned-memes

This model is a fine-tuned version of nvidia/mit-b2 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.5984
• Accuracy: 0.8323
• Precision: 0.8312
• Recall: 0.8323
• F1: 0.8315

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: 0.00012
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
1.3683	0.99	20	1.1798	0.5703	0.4914	0.5703	0.4915
1.0113	1.99	40	1.0384	0.6159	0.6813	0.6159	0.6274
0.7581	2.99	60	0.8348	0.6808	0.7377	0.6808	0.6840
0.6241	3.99	80	0.6034	0.7713	0.7864	0.7713	0.7735
0.4999	4.99	100	0.5481	0.7944	0.8000	0.7944	0.7909
0.3981	5.99	120	0.5253	0.8022	0.8091	0.8022	0.8000
0.3484	6.99	140	0.4688	0.8238	0.8147	0.8238	0.8146
0.3142	7.99	160	0.6245	0.7867	0.8209	0.7867	0.7920
0.2339	8.99	180	0.5053	0.8362	0.8426	0.8362	0.8355
0.2284	9.99	200	0.5070	0.8230	0.8220	0.8230	0.8187
0.1824	10.99	220	0.5780	0.8006	0.8138	0.8006	0.8035
0.1561	11.99	240	0.5429	0.8253	0.8197	0.8253	0.8218
0.1229	12.99	260	0.5325	0.8331	0.8296	0.8331	0.8303
0.1232	13.99	280	0.5595	0.8277	0.8290	0.8277	0.8273
0.118	14.99	300	0.5974	0.8292	0.8345	0.8292	0.8299
0.11	15.99	320	0.5796	0.8253	0.8228	0.8253	0.8231
0.0948	16.99	340	0.5581	0.8346	0.8358	0.8346	0.8349
0.0985	17.99	360	0.5700	0.8338	0.8301	0.8338	0.8318
0.0821	18.99	380	0.5756	0.8331	0.8343	0.8331	0.8335
0.0813	19.99	400	0.5984	0.8323	0.8312	0.8323	0.8315

Framework versions

• Transformers 4.24.0.dev0
• Pytorch 1.11.0+cu102
• Datasets 2.6.1.dev0
• Tokenizers 0.13.1