Dialect_accent_identification_10dialects_3

This model is a fine-tuned version of nadsoft/Dialect_accent_identification_10dialects_2_new on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0959
• Accuracy: 96.3504
• Dialect Accuracy: 97.4121
• Accent Accuracy: 95.2887

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
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Dialect Accuracy	Accent Accuracy
0.3518	0.02	200	0.4182	82.4818	74.5189	90.4446
0.3456	0.04	400	0.3949	83.0790	75.5143	90.6437
0.3598	0.06	600	0.3653	83.9416	76.6423	91.2409
0.3164	1.0157	800	0.3564	84.9701	78.9648	90.9754
0.3456	1.0357	1000	0.3444	85.3351	79.1639	91.5063
0.3374	1.0557	1200	0.3320	85.0365	78.8321	91.2409
0.3026	2.0114	1400	0.3219	85.9987	80.8228	91.1745
0.3094	2.0314	1600	0.2968	86.9277	81.6855	92.1699
0.3345	2.0514	1800	0.2964	86.9277	81.5528	92.3026
0.2707	3.0072	2000	0.2878	87.9562	84.0743	91.8381
0.2733	3.0272	2200	0.2712	88.1553	84.2734	92.0372
0.2929	3.0472	2400	0.2634	88.3875	85.3351	91.4399
0.2504	4.0029	2600	0.2587	88.7857	85.7332	91.8381
0.2611	4.0229	2800	0.2640	88.2880	85.2687	91.3072
0.243	4.0429	3000	0.2309	89.9137	87.7903	92.0372
0.2458	4.0629	3200	0.2228	90.4114	88.3875	92.4353
0.2195	5.0186	3400	0.2216	90.3782	87.6576	93.0989
0.228	5.0386	3600	0.2133	90.7764	88.9184	92.6344
0.2331	5.0586	3800	0.1898	92.3358	91.3736	93.2979
0.2341	6.0143	4000	0.1847	92.5680	91.3072	93.8288
0.2074	6.0343	4200	0.1863	92.3026	91.6390	92.9662
0.2159	6.0543	4400	0.1708	93.1984	92.8998	93.4970
0.1688	7.0101	4600	0.1655	93.5302	93.4307	93.6297
0.1757	7.0301	4800	0.1603	93.4638	93.4307	93.4970
0.1888	7.0501	5000	0.1595	93.6961	93.5634	93.8288
0.1628	8.0059	5200	0.1511	94.0610	94.0279	94.0942
0.1605	8.0259	5400	0.1435	94.3928	94.6914	94.0942
0.1674	8.0459	5600	0.1404	94.4260	95.0232	93.8288
0.1467	9.0017	5800	0.1339	94.5587	95.3550	93.7624
0.1497	9.0217	6000	0.1288	95.1891	95.6204	94.7578
0.1526	9.0417	6200	0.1295	94.8242	96.3504	93.2979
0.1497	9.0617	6400	0.1213	95.1228	96.2177	94.0279
0.1436	10.0174	6600	0.1194	95.3218	96.0849	94.5587
0.1318	10.0374	6800	0.1133	95.5873	96.7485	94.4260
0.1322	10.0574	7000	0.1126	95.6868	96.8812	94.4924
0.1268	11.0131	7200	0.1102	95.9190	96.9476	94.8905
0.1266	11.0331	7400	0.1091	95.5873	96.8812	94.2933
0.1353	11.0531	7600	0.1052	95.9522	96.8812	95.0232
0.1106	12.0088	7800	0.1018	96.0518	97.4121	94.6914
0.1192	12.0288	8000	0.1021	96.1845	97.3457	95.0232
0.1193	12.0488	8200	0.0991	96.3172	97.3457	95.2887
0.1164	13.0046	8400	0.0988	96.1845	97.2130	95.1559
0.1069	13.0246	8600	0.0980	96.1181	97.2130	95.0232
0.1123	13.0446	8800	0.0983	96.4167	97.5448	95.2887
0.1163	14.0003	9000	0.0966	96.2840	97.2794	95.2887
0.1037	14.0203	9200	0.0964	96.4167	97.5448	95.2887
0.1071	14.0403	9400	0.0961	96.3172	97.4121	95.2223
0.1154	14.0603	9600	0.0960	96.2840	97.3457	95.2223
0.1147	15.0161	9800	0.0959	96.3835	97.4784	95.2887
0.1083	15.0361	10000	0.0959	96.3504	97.4121	95.2887

Framework versions

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