cvt-13-384-22k-fv-finetuned-memes

This model is a fine-tuned version of microsoft/cvt-13-384-22k on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.5761
• Accuracy: 0.8315
• Precision: 0.8302
• Recall: 0.8315
• F1: 0.8292

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
1.3821	0.99	20	1.2780	0.4969	0.5083	0.4969	0.4458
1.0785	1.99	40	0.8633	0.6669	0.6658	0.6669	0.6500
0.8862	2.99	60	0.7110	0.7218	0.7258	0.7218	0.7013
0.665	3.99	80	0.5515	0.8045	0.8137	0.8045	0.8050
0.6056	4.99	100	0.5956	0.7960	0.8041	0.7960	0.7846
0.4779	5.99	120	0.6229	0.7937	0.7945	0.7937	0.7857
0.4554	6.99	140	0.5355	0.8099	0.8126	0.8099	0.8086
0.4249	7.99	160	0.5447	0.8269	0.8275	0.8269	0.8236
0.4313	8.99	180	0.5530	0.8153	0.8140	0.8153	0.8132
0.423	9.99	200	0.5346	0.8238	0.8230	0.8238	0.8223
0.3997	10.99	220	0.5413	0.8338	0.8347	0.8338	0.8338
0.4095	11.99	240	0.5999	0.8207	0.8231	0.8207	0.8177
0.3979	12.99	260	0.5632	0.8284	0.8255	0.8284	0.8250
0.3408	13.99	280	0.5725	0.8207	0.8198	0.8207	0.8196
0.3828	14.99	300	0.5631	0.8277	0.8258	0.8277	0.8260
0.3595	15.99	320	0.6005	0.8308	0.8297	0.8308	0.8275
0.3789	16.99	340	0.5840	0.8300	0.8271	0.8300	0.8273
0.3545	17.99	360	0.5983	0.8246	0.8226	0.8246	0.8222
0.3472	18.99	380	0.5795	0.8416	0.8382	0.8416	0.8390
0.355	19.99	400	0.5761	0.8315	0.8302	0.8315	0.8292

Framework versions

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