models

This model is a fine-tuned version of Salesforce/blip-image-captioning-base on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 1.4107
• Wer Score: 0.5495

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

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Score
9.4536	0.05	10	7.8217	41.7753
7.3267	0.11	20	6.6585	0.7753
6.2358	0.16	30	5.7758	0.5667
5.2862	0.22	40	4.7628	0.5419
4.3786	0.27	50	3.9203	0.6398
3.5554	0.33	60	3.1482	0.5613
2.849	0.38	70	2.5209	0.5548
2.3041	0.44	80	2.0561	0.5645
1.8999	0.49	90	1.7474	0.5645
1.658	0.55	100	1.5722	0.5548
1.5238	0.6	110	1.4836	0.5591
1.4726	0.66	120	1.4461	0.5538
1.4328	0.71	130	1.4285	0.5473
1.4211	0.77	140	1.4205	0.5559
1.4202	0.82	150	1.4156	0.5548
1.4098	0.88	160	1.4129	0.5505
1.4124	0.93	170	1.4113	0.5548
1.4075	0.99	180	1.4107	0.5495

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cpu
• Datasets 2.13.1
• Tokenizers 0.13.3