git-base-one-entrance-dungeons

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: 0.0219
• Wer Score: 1.2

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: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Score
0.025	5.0	10	0.0412	18.6
0.0225	10.0	20	0.0377	18.8
0.0182	15.0	30	0.0377	20.6
0.0135	20.0	40	0.0337	21.2
0.0095	25.0	50	0.0314	17.0
0.0067	30.0	60	0.0256	17.0
0.0054	35.0	70	0.0177	5.8
0.006	40.0	80	0.0240	37.8
0.0063	45.0	90	0.0213	0.2
0.0032	50.0	100	0.0222	1.8
0.0021	55.0	110	0.0210	1.6
0.0015	60.0	120	0.0203	2.6
0.0012	65.0	130	0.0211	1.4
0.0011	70.0	140	0.0217	1.4
0.001	75.0	150	0.0218	1.2
0.0009	80.0	160	0.0219	1.2
0.0009	85.0	170	0.0219	1.2
0.0009	90.0	180	0.0219	1.2
0.0009	95.0	190	0.0219	1.2
0.0009	100.0	200	0.0219	1.2

Framework versions

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