git-base-instagram-captions

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: 2.8334
• Wer Score: 13.4415

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Score
4.3503	0.3556	20	4.0231	13.3087
3.2554	0.7111	40	3.0955	14.7696
2.7452	1.0667	60	2.5707	11.9736
1.916	1.4222	80	2.4606	14.1495
2.0349	1.7778	100	2.4337	15.4433
1.7522	2.1333	120	2.4274	13.5215
1.5496	2.4889	140	2.4298	13.0932
1.554	2.8444	160	2.4232	14.4336
1.3561	3.2	180	2.4537	13.2498
1.2081	3.5556	200	2.4563	14.3843
1.1593	3.9111	220	2.4378	14.3703
1.0503	4.2667	240	2.4849	14.6614
1.0292	4.6222	260	2.4937	14.5066
1.092	4.9778	280	2.4853	14.2252
0.8704	5.3333	300	2.5228	13.2709
0.8933	5.6889	320	2.5363	14.3369
0.8494	6.0444	340	2.5329	16.8566
0.7886	6.4	360	2.5424	13.8083
0.8338	6.7556	380	2.5585	13.6931
0.7756	7.1111	400	2.5856	14.4222
0.6575	7.4667	420	2.5890	13.6350
0.6793	7.8222	440	2.5833	14.1258
0.6439	8.1778	460	2.6046	15.0616
0.597	8.5333	480	2.6183	13.7511
0.6157	8.8889	500	2.6166	15.7696
0.6132	9.2444	520	2.6466	17.8857
0.5666	9.6	540	2.6472	16.8452
0.5691	9.9556	560	2.6456	13.7449
0.529	10.3111	580	2.6676	14.5945
0.5286	10.6667	600	2.6727	13.7344
0.5339	11.0222	620	2.6706	14.9833
0.4967	11.3778	640	2.6928	13.9587
0.459	11.7333	660	2.7035	14.5708
0.5117	12.0889	680	2.7064	14.3026
0.4424	12.4444	700	2.7304	14.0836
0.4829	12.8	720	2.7175	14.7573
0.4308	13.1556	740	2.7129	14.5928
0.3871	13.5111	760	2.7333	14.1328
0.4238	13.8667	780	2.7319	14.1759
0.4141	14.2222	800	2.7605	14.0580
0.3435	14.5778	820	2.7604	13.6966
0.3807	14.9333	840	2.7634	13.7836
0.3293	15.2889	860	2.7802	13.9120
0.3244	15.6444	880	2.7760	13.5101
0.3289	16.0	900	2.7884	13.5409
0.2768	16.3556	920	2.7946	13.7801
0.2825	16.7111	940	2.7988	14.1662
0.3059	17.0667	960	2.8092	13.6667
0.2279	17.4222	980	2.8253	13.7704
0.2283	17.7778	1000	2.8236	13.7256
0.2293	18.1333	1020	2.8174	13.1741
0.1934	18.4889	1040	2.8220	13.5479
0.1927	18.8444	1060	2.8225	13.6324
0.1794	19.2	1080	2.8343	13.4195
0.169	19.5556	1100	2.8345	13.3967
0.1724	19.9111	1120	2.8334	13.4415

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1