resnet-50-FV2-finetuned-memes

This model is a fine-tuned version of microsoft/resnet-50 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.9263
• Accuracy: 0.6453
• Precision: 0.5728
• Recall: 0.6453
• F1: 0.5964

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
1.5763	0.99	20	1.5575	0.4281	0.2966	0.4281	0.2669
1.4761	1.99	40	1.4424	0.4343	0.1886	0.4343	0.2630
1.3563	2.99	60	1.3240	0.4343	0.1886	0.4343	0.2630
1.2824	3.99	80	1.2636	0.4389	0.3097	0.4389	0.2734
1.2315	4.99	100	1.2119	0.4529	0.3236	0.4529	0.3042
1.1956	5.99	120	1.1764	0.4900	0.3731	0.4900	0.3692
1.1452	6.99	140	1.1424	0.5147	0.3963	0.5147	0.4090
1.1076	7.99	160	1.1190	0.5371	0.4121	0.5371	0.4392
1.0679	8.99	180	1.0825	0.5719	0.4465	0.5719	0.4831
1.0432	9.99	200	1.0482	0.5750	0.5404	0.5750	0.4930
0.9903	10.99	220	1.0275	0.5958	0.5459	0.5958	0.5241
0.9675	11.99	240	1.0145	0.6051	0.5350	0.6051	0.5379
0.9335	12.99	260	0.9860	0.6175	0.5537	0.6175	0.5527
0.9157	13.99	280	0.9683	0.6105	0.5386	0.6105	0.5504
0.8901	14.99	300	0.9558	0.6352	0.5686	0.6352	0.5833
0.8722	15.99	320	0.9382	0.6345	0.5657	0.6345	0.5807
0.854	16.99	340	0.9322	0.6376	0.5623	0.6376	0.5856
0.8494	17.99	360	0.9287	0.6422	0.6675	0.6422	0.5918
0.8652	18.99	380	0.9212	0.6399	0.5640	0.6399	0.5863
0.846	19.99	400	0.9263	0.6453	0.5728	0.6453	0.5964

Framework versions

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