git-base-isg

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

• Loss: 0.0942
• Wer Score: 2.2938

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 150

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Score
7.115	9.0909	50	4.3128	8.375
2.1356	18.1818	100	0.3729	6.7062
0.1103	27.2727	150	0.0698	0.6562
0.0158	36.3636	200	0.0788	0.6813
0.0063	45.4545	250	0.0859	1.4
0.0016	54.5455	300	0.0894	2.3062
0.0009	63.6364	350	0.0916	2.2875
0.0007	72.7273	400	0.0914	2.2875
0.0005	81.8182	450	0.0923	2.475
0.0005	90.9091	500	0.0928	2.3438
0.0004	100.0	550	0.0934	2.35
0.0004	109.0909	600	0.0939	2.3375
0.0003	118.1818	650	0.0941	2.275
0.0003	127.2727	700	0.0942	2.275
0.0003	136.3636	750	0.0942	2.2938

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3