nadsoft/Dialect_accent_identification_10dialects_2

Dialect_accent_identification_10dialects

This model is a fine-tuned version of nadsoft/Dialect_accent_identification_5dialects_tokens on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.8913
• Accuracy: 73.6969
• Dialect Accuracy: 63.3377
• Accent Accuracy: 84.0561

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: 2e-06
• train_batch_size: 128
• eval_batch_size: 4
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 500
• num_epochs: 35
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Dialect Accuracy	Accent Accuracy
3.5487	0.9569	200	2.5199	63.6662	61.9799	65.3526
3.1001	1.9139	400	2.1966	63.3377	61.4104	65.2650
2.1611	2.8708	600	1.6260	62.3303	59.3079	65.3526
1.3811	3.8278	800	1.1704	64.1042	59.9212	68.2873
1.0108	4.7847	1000	0.9685	64.6518	61.1038	68.1997
0.9075	5.7416	1200	0.9193	64.1042	59.7021	68.5064
0.7821	6.6986	1400	0.8516	64.7832	60.8410	68.7254
0.7074	7.6555	1600	0.8293	64.9146	61.3666	68.4625
0.6458	8.6124	1800	0.7995	64.4985	60.5344	68.4625
0.6108	9.5694	2000	0.8176	65.1336	58.4757	71.7915
0.5215	10.5263	2200	0.7979	70.6307	59.9212	81.3403
0.4769	11.4833	2400	0.7472	72.8647	62.8121	82.9172
0.4384	12.4402	2600	0.7528	72.7332	62.0675	83.3990
0.4102	13.3971	2800	0.7500	72.7551	61.8922	83.6180
0.3728	14.3541	3000	0.7640	73.5655	63.8195	83.3114
0.3336	15.3110	3200	0.7871	72.6456	62.3303	82.9610
0.3198	16.2679	3400	0.7685	73.7188	63.8195	83.6180
0.2938	17.2249	3600	0.7727	73.3246	63.2501	83.3990
0.2801	18.1818	3800	0.7891	73.7188	63.7757	83.6618
0.2744	19.1388	4000	0.7939	73.0837	62.6369	83.5304
0.2465	20.0957	4200	0.8146	73.8721	64.2138	83.5304
0.2539	21.0526	4400	0.8218	73.0399	62.1113	83.9685
0.2336	22.0096	4600	0.8273	73.4779	64.0823	82.8734
0.2283	22.9665	4800	0.8472	73.4122	63.0749	83.7495
0.2225	23.9234	5000	0.8676	72.8208	62.2865	83.3552
0.2089	24.8804	5200	0.8618	73.2808	62.8997	83.6618
0.2182	25.8373	5400	0.8781	73.1275	62.7245	83.5304
0.1948	26.7943	5600	0.8914	72.7989	62.0237	83.5742
0.2003	27.7512	5800	0.8767	73.3246	63.1187	83.5304
0.1909	28.7081	6000	0.8878	73.1932	63.1625	83.2238
0.1852	29.6651	6200	0.8900	73.7845	63.6881	83.8809
0.1976	30.6220	6400	0.8904	73.2589	63.5129	83.0048
0.1813	31.5789	6600	0.8883	73.8940	63.6005	84.1875
0.1774	32.5359	6800	0.8911	73.6093	63.3377	83.8809
0.1857	33.4928	7000	0.8916	73.6531	63.2939	84.0123
0.1845	34.4498	7200	0.8913	73.6969	63.3377	84.0561

Framework versions

• Transformers 4.57.3
• Pytorch 2.11.0+cu128
• Datasets 2.18.0
• Tokenizers 0.22.2