segformer-finetuned-food101

This model is a fine-tuned version of nvidia/mit-b0 on the food101 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3478
• Accuracy: 0.888

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.0272	0.98	23	1.8039	0.329
1.5806	2.0	47	1.2465	0.608
1.0564	2.98	70	0.7507	0.756
0.7358	4.0	94	0.6263	0.784
0.6482	4.98	117	0.5551	0.795
0.5692	6.0	141	0.5849	0.794
0.5552	6.98	164	0.4931	0.831
0.4956	8.0	188	0.5166	0.83
0.4748	8.98	211	0.4808	0.834
0.424	10.0	235	0.4238	0.852
0.4314	10.98	258	0.4858	0.838
0.4071	12.0	282	0.4304	0.858
0.3928	12.98	305	0.4621	0.851
0.3695	14.0	329	0.4398	0.859
0.3704	14.98	352	0.4172	0.855
0.3299	16.0	376	0.4225	0.856
0.3391	16.98	399	0.4165	0.855
0.3023	18.0	423	0.3828	0.869
0.3318	18.98	446	0.4190	0.861
0.2994	20.0	470	0.4190	0.861
0.323	20.98	493	0.4034	0.866
0.2883	22.0	517	0.4083	0.874
0.2959	22.98	540	0.4202	0.862
0.2665	24.0	564	0.3740	0.881
0.2765	24.98	587	0.4123	0.866
0.2728	26.0	611	0.3763	0.868
0.2817	26.98	634	0.3939	0.864
0.2467	28.0	658	0.3938	0.87
0.2772	28.98	681	0.4013	0.866
0.2243	29.36	690	0.3478	0.888

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0