git-base-pokemon

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: 0.0340
• Wer Score: 2.1498

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: 10
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Score
7.321	2.13	50	4.4679	21.5557
2.2294	4.26	100	0.3441	11.8745
0.1021	6.38	150	0.0283	0.5672
0.0187	8.51	200	0.0251	0.6018
0.0086	10.64	250	0.0272	3.6786
0.0038	12.77	300	0.0288	6.7119
0.0019	14.89	350	0.0300	4.2023
0.0011	17.02	400	0.0308	4.0768
0.0009	19.15	450	0.0310	3.5980
0.0007	21.28	500	0.0315	3.5723
0.0007	23.4	550	0.0323	2.8835
0.0006	25.53	600	0.0325	2.8399
0.0006	27.66	650	0.0330	2.6274
0.0006	29.79	700	0.0331	2.5416
0.0006	31.91	750	0.0334	2.4213
0.0006	34.04	800	0.0335	2.3214
0.0006	36.17	850	0.0330	2.2330
0.0006	38.3	900	0.0337	2.2254
0.0006	40.43	950	0.0338	2.1652
0.0006	42.55	1000	0.0340	2.1447
0.0006	44.68	1050	0.0340	2.1767
0.0006	46.81	1100	0.0340	2.1536
0.0006	48.94	1150	0.0340	2.1498

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.0
• Datasets 2.12.0
• Tokenizers 0.13.3