Wav2vec2-xlsr-Shemo-Ravdess

This model is a fine-tuned version of makhataei/emotion_recog on the minoosh/shEMO dataset. It achieves the following results on the evaluation set:

• Loss: 1.1736
• Accuracy: 0.6080

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: 0.003
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.1422	1.0	224	1.1661	0.6246
1.0217	2.0	449	1.1937	0.6257
1.0559	3.0	674	1.2977	0.5856
1.0936	4.0	899	1.3626	0.5939
1.1228	5.0	1123	1.2982	0.5880
1.0318	6.0	1348	1.2040	0.6175
1.0672	7.0	1573	1.2251	0.6045
1.0632	8.0	1798	1.4452	0.5797
1.1499	9.0	2022	1.3601	0.5868
1.0777	10.0	2247	1.3385	0.5832
1.0781	11.0	2472	1.1732	0.5915
1.0678	12.0	2697	1.2847	0.5821
1.1707	13.0	2921	1.3588	0.6033
1.1455	14.0	3146	1.3306	0.5632
1.1142	15.0	3371	1.5357	0.5289
1.1206	16.0	3596	1.4326	0.5702
1.1422	17.0	3820	1.7375	0.5018
1.0812	18.0	4045	1.5080	0.5325
1.1203	19.0	4270	1.3482	0.5809
1.1119	20.0	4495	1.2645	0.5632
1.1494	21.0	4719	1.4513	0.5608
1.1657	22.0	4944	1.4032	0.5632
1.1381	23.0	5169	1.2126	0.5773
1.0927	24.0	5394	1.3380	0.5655
1.015	25.0	5618	1.2219	0.5927
1.0204	26.0	5843	1.2501	0.5939
1.008	27.0	6068	1.2044	0.5891
1.0234	28.0	6293	1.2707	0.5832
0.9675	29.0	6517	1.1381	0.6009
0.9978	30.0	6742	1.2326	0.5797
1.0122	31.0	6967	1.2166	0.5797
0.9453	32.0	7192	1.1262	0.6104
1.0142	33.0	7416	1.1846	0.6104
0.9738	34.0	7641	1.1193	0.6175
0.9176	35.0	7866	1.1634	0.5950
1.0274	36.0	8091	1.1692	0.6187
1.0281	37.0	8315	1.1404	0.6045
0.9439	38.0	8540	1.1726	0.5927
0.9435	39.0	8765	1.0951	0.6234
0.9281	40.0	8990	1.1170	0.6068
0.9538	41.0	9214	1.3379	0.5785
0.9625	42.0	9439	1.2126	0.5915
0.912	43.0	9664	1.3407	0.5679
0.9173	44.0	9889	1.1140	0.6269
0.8946	45.0	10113	1.2278	0.5998
0.9263	46.0	10338	1.4053	0.5490
0.9285	47.0	10563	1.2031	0.6116
0.9143	48.0	10788	1.3764	0.5927
0.9136	49.0	11012	1.2281	0.6033
0.8833	50.0	11237	1.1829	0.5962
0.9111	51.0	11462	1.2174	0.5797
0.9125	52.0	11687	1.2228	0.6068
0.9345	53.0	11911	1.1174	0.6187
0.9006	54.0	12136	1.2127	0.5998
0.9047	55.0	12361	1.1860	0.5974
0.8726	56.0	12586	1.3472	0.5608
0.8383	57.0	12810	1.3870	0.5596
0.8635	58.0	13035	1.2173	0.5962
0.8488	59.0	13260	1.1971	0.5962
0.9105	60.0	13485	1.2144	0.5986
0.8562	61.0	13709	1.2294	0.5880
0.8683	62.0	13934	1.1234	0.6092
0.8944	63.0	14159	1.2112	0.5868
0.8509	64.0	14384	1.5180	0.5537
0.8356	65.0	14608	1.2713	0.5726
0.854	66.0	14833	1.2937	0.5797
0.8977	67.0	15058	1.2866	0.5844
0.8984	68.0	15283	1.2556	0.5880
0.8935	69.0	15507	1.3198	0.5868
0.8521	70.0	15732	1.2408	0.5986
0.8971	71.0	15957	1.2107	0.5986
0.8621	72.0	16182	1.2558	0.5797
0.8549	73.0	16406	1.2017	0.6021
0.8362	74.0	16631	1.3402	0.5915
0.8826	75.0	16856	1.2788	0.6009
0.8518	76.0	17081	1.2311	0.6198
0.8495	77.0	17305	1.1504	0.6340
0.8547	78.0	17530	1.2002	0.6080
0.8794	79.0	17755	1.2842	0.5821
0.893	80.0	17980	1.2378	0.5962
0.8451	81.0	18204	1.1525	0.6175
0.8368	82.0	18429	1.2184	0.6068
0.836	83.0	18654	1.1771	0.6033
0.8364	84.0	18879	1.2725	0.5844
0.7702	85.0	19103	1.2220	0.6057
0.8539	86.0	19328	1.1542	0.6104
0.8075	87.0	19553	1.2943	0.5927
0.7758	88.0	19778	1.2257	0.5868
0.782	89.0	20002	1.1406	0.6246
0.7952	90.0	20227	1.2678	0.5998
0.7911	91.0	20452	1.1540	0.6222
0.7845	92.0	20677	1.1804	0.6187
0.8066	93.0	20901	1.1764	0.6210
0.8042	94.0	21126	1.1706	0.6068
0.7981	95.0	21351	1.1516	0.6175
0.7585	96.0	21576	1.1980	0.5998
0.7576	97.0	21800	1.1870	0.6092
0.7908	98.0	22025	1.1770	0.6080
0.7646	99.0	22250	1.1745	0.6092
0.7791	99.67	22400	1.1736	0.6080

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.1+cu117
• Datasets 2.12.0
• Tokenizers 0.13.3