git-base-new-one-entrance-dungeons-35

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: 5.6255
• Wer Score: 19.8

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: 2e-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: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 35

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Score
7.2058	0.8	2	7.1464	69.7
7.1938	1.6	4	7.1157	69.7
7.1572	2.4	6	7.0644	69.7
7.0951	3.2	8	6.9847	69.7
7.0118	4.0	10	6.9004	69.7
6.928	4.8	12	6.8197	69.7
6.8465	5.6	14	6.7396	69.65
6.7675	6.4	16	6.6637	69.4
6.6913	7.2	18	6.5899	69.65
6.6176	8.0	20	6.5193	65.0
6.5466	8.8	22	6.4509	69.7
6.4786	9.6	24	6.3847	21.9
6.4126	10.4	26	6.3221	21.0
6.3502	11.2	28	6.2613	69.25
6.2899	12.0	30	6.2033	19.65
6.2323	12.8	32	6.1486	19.7
6.1779	13.6	34	6.0961	19.8
6.1258	14.4	36	6.0462	19.85
6.0761	15.2	38	5.9986	19.85
6.0298	16.0	40	5.9544	19.9
5.9861	16.8	42	5.9129	19.8
5.9451	17.6	44	5.8742	19.8
5.9073	18.4	46	5.8384	19.8
5.8716	19.2	48	5.8049	19.8
5.8391	20.0	50	5.7745	19.8
5.8095	20.8	52	5.7473	19.8
5.7823	21.6	54	5.7224	19.8
5.7583	22.4	56	5.7002	19.8
5.7369	23.2	58	5.6811	19.8
5.7185	24.0	60	5.6649	19.8
5.7031	24.8	62	5.6513	19.8
5.6902	25.6	64	5.6405	19.8
5.6802	26.4	66	5.6326	19.8
5.6733	27.2	68	5.6276	19.8
5.6688	28.0	70	5.6255	19.8

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.2
• Tokenizers 0.19.1