kurosekurose/wav2vec2-base-960h-EMOPIA-10sec-100epoc

wav2vec2-base-960h-EMOPIA-10sec-full-100epoc

This model is a fine-tuned version of facebook/wav2vec2-base-960h on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.5358
• Accuracy: 0.8737

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: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.2276	1.0	2248	1.3142	0.5730
1.4493	2.0	4496	1.5339	0.5907
1.4328	3.0	6744	1.5229	0.6584
1.3839	4.0	8992	1.3336	0.6922
1.4591	5.0	11240	1.4407	0.6762
1.3609	6.0	13488	1.1341	0.7117
1.222	7.0	15736	1.5204	0.7028
1.1713	8.0	17984	1.3222	0.7473
1.1414	9.0	20232	1.3233	0.7562
1.0977	10.0	22480	1.3587	0.7384
0.9768	11.0	24728	1.3482	0.7633
0.9219	12.0	26976	1.3743	0.7580
0.8636	13.0	29224	1.1195	0.8167
0.8543	14.0	31472	1.1716	0.8043
0.8053	15.0	33720	1.2033	0.8114
0.7718	16.0	35968	1.2491	0.7989
0.6904	17.0	38216	1.0851	0.8363
0.6545	18.0	40464	1.2963	0.8007
0.6858	19.0	42712	1.3231	0.8078
0.6444	20.0	44960	1.1918	0.8238
0.6166	21.0	47208	1.1358	0.8345
0.5437	22.0	49456	1.2446	0.8256
0.4719	23.0	51704	1.4120	0.8149
0.4802	24.0	53952	1.2611	0.8203
0.484	25.0	56200	1.2840	0.8363
0.3649	26.0	58448	1.2421	0.8434
0.4146	27.0	60696	1.3465	0.8292
0.3998	28.0	62944	1.2309	0.8505
0.4113	29.0	65192	1.1663	0.8523
0.3385	30.0	67440	1.2567	0.8470
0.3188	31.0	69688	1.2581	0.8434
0.3203	32.0	71936	1.2454	0.8541
0.2766	33.0	74184	1.2542	0.8523
0.2505	34.0	76432	1.5897	0.8149
0.2777	35.0	78680	1.3483	0.8363
0.2816	36.0	80928	1.2510	0.8523
0.2728	37.0	83176	1.4422	0.8327
0.255	38.0	85424	1.2928	0.8488
0.2172	39.0	87672	1.4022	0.8452
0.2204	40.0	89920	1.4114	0.8381
0.2232	41.0	92168	1.4324	0.8416
0.2301	42.0	94416	1.3528	0.8488
0.1751	43.0	96664	1.4649	0.8434
0.1982	44.0	98912	1.2216	0.8754
0.1803	45.0	101160	1.4569	0.8452
0.1582	46.0	103408	1.3650	0.8665
0.1837	47.0	105656	1.2877	0.8541
0.1458	48.0	107904	1.7389	0.8310
0.1664	49.0	110152	1.4001	0.8541
0.1473	50.0	112400	1.2979	0.8701
0.1341	51.0	114648	1.5705	0.8470
0.1603	52.0	116896	1.6043	0.8381
0.1133	53.0	119144	1.6194	0.8452
0.107	54.0	121392	1.4173	0.8630
0.116	55.0	123640	1.5268	0.8541
0.0988	56.0	125888	1.6092	0.8523
0.139	57.0	128136	1.4312	0.8648
0.0798	58.0	130384	1.7888	0.8327
0.0776	59.0	132632	1.5457	0.8665
0.1288	60.0	134880	1.4554	0.8630
0.0828	61.0	137128	1.7078	0.8559
0.0823	62.0	139376	1.4734	0.8754
0.0803	63.0	141624	1.6007	0.8594
0.0947	64.0	143872	1.4467	0.8701
0.0916	65.0	146120	1.4410	0.8737
0.0814	66.0	148368	1.7116	0.8470
0.0938	67.0	150616	1.5838	0.8630
0.066	68.0	152864	1.6458	0.8559
0.096	69.0	155112	1.6926	0.8559
0.0638	70.0	157360	1.5233	0.8630
0.063	71.0	159608	1.5641	0.8594
0.0758	72.0	161856	1.6767	0.8505
0.0579	73.0	164104	1.5338	0.8630
0.0379	74.0	166352	1.6348	0.8630
0.0351	75.0	168600	1.7037	0.8559
0.0472	76.0	170848	1.5682	0.8754
0.0253	77.0	173096	1.7067	0.8559
0.073	78.0	175344	1.4460	0.8754
0.049	79.0	177592	1.5897	0.8594
0.0503	80.0	179840	1.6017	0.8648
0.0497	81.0	182088	1.5319	0.8683
0.0553	82.0	184336	1.5479	0.8612
0.0416	83.0	186584	1.5556	0.8577
0.0641	84.0	188832	1.5675	0.8594
0.0425	85.0	191080	1.6854	0.8559
0.0311	86.0	193328	1.4628	0.8737
0.0456	87.0	195576	1.5069	0.8701
0.0224	88.0	197824	1.6130	0.8665
0.0345	89.0	200072	1.5750	0.8701
0.041	90.0	202320	1.5230	0.8719
0.0165	91.0	204568	1.6564	0.8594
0.0478	92.0	206816	1.5940	0.8630
0.032	93.0	209064	1.4741	0.8808
0.0433	94.0	211312	1.5333	0.8719
0.0243	95.0	213560	1.5165	0.8719
0.0165	96.0	215808	1.5775	0.8683
0.0177	97.0	218056	1.5302	0.8772
0.0253	98.0	220304	1.5424	0.8754
0.0224	99.0	222552	1.5462	0.8719
0.0213	100.0	224800	1.5358	0.8737

Framework versions

• Transformers 4.45.2
• Pytorch 2.4.1+cu118
• Datasets 3.0.1
• Tokenizers 0.20.0