git-base-floors615images

This model is a fine-tuned version of microsoft/git-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0971
• Wer Score: 4.3530

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: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Score
7.0263	0.72	50	4.1468	2.9871
1.9906	1.45	100	0.2971	0.4260
0.1401	2.17	150	0.0827	2.2259
0.0772	2.9	200	0.0704	3.9564
0.0567	3.62	250	0.0677	3.6360
0.0501	4.35	300	0.0678	3.2940
0.0443	5.07	350	0.0664	3.3094
0.0386	5.8	400	0.0666	3.0884
0.035	6.52	450	0.0681	3.3818
0.0329	7.25	500	0.0667	3.2474
0.0297	7.97	550	0.0681	3.6648
0.0271	8.7	600	0.0700	3.9405
0.0257	9.42	650	0.0697	3.0393
0.0243	10.14	700	0.0701	3.5973
0.0233	10.87	750	0.0695	3.0565
0.0213	11.59	800	0.0721	3.6980
0.0203	12.32	850	0.0730	3.8883
0.0206	13.04	900	0.0718	4.9104
0.0187	13.77	950	0.0729	4.4850
0.0192	14.49	1000	0.0743	4.6366
0.0184	15.22	1050	0.0729	4.9141
0.0169	15.94	1100	0.0754	4.6446
0.0167	16.67	1150	0.0769	4.8232
0.0162	17.39	1200	0.0766	4.7569
0.0165	18.12	1250	0.0768	4.3266
0.0157	18.84	1300	0.0776	4.1375
0.0144	19.57	1350	0.0778	3.9724
0.015	20.29	1400	0.0790	4.7041
0.0146	21.01	1450	0.0780	4.5003
0.0137	21.74	1500	0.0794	4.2167
0.0141	22.46	1550	0.0792	4.4856
0.0137	23.19	1600	0.0805	4.2634
0.0137	23.91	1650	0.0817	4.4162
0.0127	24.64	1700	0.0804	4.0319
0.0127	25.36	1750	0.0829	4.3628
0.013	26.09	1800	0.0826	4.4211
0.0121	26.81	1850	0.0823	4.8932
0.0119	27.54	1900	0.0835	4.6636
0.012	28.26	1950	0.0842	3.8926
0.0118	28.99	2000	0.0844	3.9994
0.011	29.71	2050	0.0833	4.1743
0.0109	30.43	2100	0.0864	4.4217
0.0108	31.16	2150	0.0851	4.8029
0.0103	31.88	2200	0.0855	4.0694
0.01	32.61	2250	0.0871	4.3198
0.0102	33.33	2300	0.0863	4.4082
0.0099	34.06	2350	0.0871	4.2112
0.0094	34.78	2400	0.0872	4.1774
0.0092	35.51	2450	0.0887	3.9742
0.009	36.23	2500	0.0882	4.1958
0.0088	36.96	2550	0.0893	4.2591
0.0084	37.68	2600	0.0885	4.2983
0.0079	38.41	2650	0.0894	4.6550
0.008	39.13	2700	0.0904	4.1277
0.0076	39.86	2750	0.0908	3.6771
0.0072	40.58	2800	0.0912	4.1252
0.0072	41.3	2850	0.0908	4.5660
0.0069	42.03	2900	0.0917	3.9441
0.0062	42.75	2950	0.0924	4.2259
0.006	43.48	3000	0.0924	4.2167
0.0059	44.2	3050	0.0937	4.6047
0.0055	44.93	3100	0.0945	4.4408
0.0048	45.65	3150	0.0950	3.9871
0.0048	46.38	3200	0.0952	4.2259
0.0046	47.1	3250	0.0962	4.2204
0.0042	47.83	3300	0.0963	4.2750
0.0037	48.55	3350	0.0971	4.1891
0.0039	49.28	3400	0.0970	4.3100
0.0036	50.0	3450	0.0971	4.3530

Framework versions

• Transformers 4.38.2
• Pytorch 2.0.1
• Datasets 2.18.0
• Tokenizers 0.15.2