git-base-one-entrance-dungeons-20

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

• Loss: 0.0115
• Wer Score: 0.2812

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: 2
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine_with_restarts
• lr_scheduler_warmup_steps: 100
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Score
0.0112	0.6061	10	0.0108	0.2812
0.0094	1.2121	20	0.0105	0.25
0.0109	1.8182	30	0.0114	0.2656
0.0112	2.4242	40	0.0103	0.25
0.0114	3.0303	50	0.0108	0.2812
0.0107	3.6364	60	0.0113	0.2812
0.0119	4.2424	70	0.0108	0.2344
0.0121	4.8485	80	0.0106	0.2344
0.0115	5.4545	90	0.0112	0.25
0.0126	6.0606	100	0.0107	0.25
0.0118	6.6667	110	0.0119	0.25
0.0116	7.2727	120	0.0105	0.2188
0.0122	7.8788	130	0.0105	0.2656
0.0103	8.4848	140	0.0109	0.2812
0.0102	9.0909	150	0.0107	0.25
0.0099	9.6970	160	0.0118	0.25
0.0091	10.3030	170	0.0113	0.2656
0.0095	10.9091	180	0.0109	0.2656
0.0093	11.5152	190	0.0114	0.25
0.0088	12.1212	200	0.0119	0.2812
0.0091	12.7273	210	0.0123	0.2812
0.009	13.3333	220	0.0119	0.2969
0.0092	13.9394	230	0.0112	0.25
0.0084	14.5455	240	0.0116	0.2812
0.009	15.1515	250	0.0118	0.2969
0.0077	15.7576	260	0.0120	0.2656
0.008	16.3636	270	0.0116	0.2344
0.0079	16.9697	280	0.0115	0.2812
0.0077	17.5758	290	0.0115	0.2812
0.0079	18.1818	300	0.0116	0.2812
0.0084	18.7879	310	0.0115	0.2812
0.0085	19.3939	320	0.0115	0.2812

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1